Fixes
Fixes for errors in real apps
One page per error message: what causes it and how to fix it, checked against the vendor’s own docs. If you’d rather hand it over, each page shows how.
Not sure the rules are the problem? Check first, free and in your browser: Firebase Security Rules checker, Supabase RLS policy checker and Supabase grants and RLS checker.
Lovable, Bolt and Replit
- AI builder stuckLovable keeps breaking, and the AI can’t fix the same bugStop retrying the same prompt. Roll back to the version just before the bug appeared, ask the AI to investigate and plan before it builds again, and change one thing at a time. If the bug comes back after a rollback, or it’s in sign-in, payments or who can see which data, bring in a person who reads the code.
- Netlify and Vercel build errorThe app works in Bolt or Lovable, but the deploy build failsScroll up the build log to the first error, not the last line. Most often it’s a missing environment variable: Vite builds VITE_ variables into the code at build time, and Netlify doesn’t read .env files, so they must be set in the host’s settings. Next: a different Node version, a file name whose capital letters don’t match the import, or warnings treated as errors.
- Replit deployment bugUploaded files or saved data disappear after you redeploy your Replit appReplit says to avoid saving and relying on data written to a published app’s file system. A publish is a snapshot of your project’s files, so treat anything the running app writes as temporary, not as part of the next publish. Keep records in a database, such as Replit’s managed PostgreSQL, and uploaded files in object storage, such as Replit’s App Storage.
- Lovable publishing problemMy Lovable app works in the preview, but the published site is blankOpen the published site, open your browser’s developer tools, and read the first red error in the console. It points to the cause. Lovable’s debugging guide suggests pasting that exact error into the project chat. If the console is clean, check that you published your latest changes and that you’re on the right address.
- Lovable GitHub problemMy Lovable project stopped syncing with GitHubCheck the sync status before you push anything. If Lovable says the branch has diverged, your next push replaces Lovable’s version, and edits made only in Lovable drop out of the project. Lovable saves a backup first, but don’t rely on it: keep a copy of those edits before you push. Otherwise, find your symptom below; often it comes down to which branch is synced.
- Lovable security questionIs Lovable safe for a real app?It can be, but protecting your users’ data is your job, so check it before launch. A 2025 vulnerability record, CVE-2025-48757, describes generated sites whose row-level security let anyone read or write tables; Lovable disputes it, saying each customer is responsible for their app’s data. Either way the checks are the same: row-level security on every table, no secret key in the browser, a test with two accounts, and sign-in checked on the server.
- Bolt tokensBolt keeps running out of tokensMost tokens go on Bolt reading and syncing your project files, so bigger projects cost more per message. Ask about specific files, use Plan mode to think things through, and use buttons and Version history instead of prompts, since those don’t use tokens. Stop pressing Attempt fix over and over on the same error: Bolt says each attempt uses tokens and suggests stepping in manually or asking in Plan mode. If the bug is still there after that, a person reading the code can take it over at a fixed price.
- Lovable backendLovable: the edge function can’t read my API keyEdge functions read keys from Secrets by exact name, such as STRIPE_SECRET_KEY. Check the secret exists with the name the function reads, add any secret you skipped when Lovable asked, and don’t put server keys in VITE_ variables, which go to the browser. If two Lovable projects share one Supabase project, they overwrite each other’s secrets.
- Lovable email problemLovable app not sending emails: sign-ups, resets and contact formsFind which service sends the email. On your own Supabase project with no custom SMTP, Supabase sends auth emails only to your team’s addresses, 2 an hour, so add a custom SMTP server. With Resend, the test sender onboarding@resend.dev reaches only the Resend account owner, so add and verify your domain. If mail lands in spam, set SPF, DKIM and DMARC for the sending domain. For a contact form, check the backend function that sends it has its email API key.
- Lovable performance problemWhy is my Lovable app so slow?Measure first: test the published URL with PageSpeed Insights, and aim for the main content to appear within 2.5 seconds for at least 75% of visits. If data loads slowly, ask Lovable to find and fix the slowest database queries before you pay for a bigger database. If the page itself is slow, ask Lovable for a performance audit that reports duplicate network calls and large assets without changing code, then fix the worst first. Never lazy-load the main (LCP) image.
- Replit performance problemWhy is my Replit app so slow?Check the deployment type first. An Autoscale deployment drops to zero servers when idle, so the first visitor after a quiet spell may have to wait for a server to start; a Reserved VM never sleeps, at a fixed monthly cost. If it is slow under load, choose a more powerful machine configuration and publish again. Then, on a paid plan, run Replit’s Performance Agent to find slow code it can tie to your project.
Supabase
- Supabase setup questionHow do I get a Supabase staging environment and stop testing on production?Use Supabase Branching: a persistent branch for staging, or a preview branch for each pull request. A branch connected to GitHub builds its database structure from the migrations in your repository; every branch starts with no data by default and gets placeholder sign-in secrets. So get your schema into migrations first, then add a seed file and branch secrets. Watch the cost: branches bill by the hour and aren’t covered by the Spend Cap.
- Supabase StorageMy Supabase storage bucket is public. Are the files exposed?A public bucket skips access control for reading files: anyone with a file’s URL can open it. Uploads, deletes and moves are still checked by your policies. Keep public buckets for files meant to be public, move private files to a private bucket with row-level security policies, and serve them through signed URLs or signed-in downloads.
- Supabase RealtimeSupabase Realtime is subscribed, but no changes arriveFirst check the table is in the supabase_realtime publication, since Realtime ignores tables that aren’t. Then check row-level security: Realtime applies it as the subscribing user, so on a table with RLS it only sends changes that user is allowed to see, and SUBSCRIBED doesn’t mean they can. If events arrive but old values are empty, that’s replica identity. Last, check the table name, schema and filter in the subscription.
- Supabase databaseSupabase: too many connections, or remaining connection slots are reservedYour database has a connection limit set by its compute size, and something is opening more connections than it allows. In serverless or edge functions, connect through Supabase’s shared pooler in transaction mode, create the client once, and set its pool to 1 connection. If you’re already pooling and still at the limit, a larger compute size allows more connections.
- Supabase Auth session bugSupabase logs users out when they refresh the pageIn a Next.js or other server-rendered app, use the @supabase/ssr clients so the session lives in cookies, and add the middleware (proxy.ts on Next.js 16) that refreshes it on every request. In a browser-only app, create one Supabase client with persistSession on, and wait for the stored session to load before your route guard sends anyone to the login page.
- Supabase database errornew row violates row-level security policy for tableThe table has RLS on and no INSERT policy that this row passes. Add one whose WITH CHECK matches the row, usually user_id = auth.uid(), and make sure the insert sets user_id for a signed-in user. If you call .select() after the insert, or use upsert, the table also needs a SELECT policy the new row passes. Don’t fix it by turning RLS off.
- Supabase Auth email bugSupabase confirmation and reset emails link to localhost:3000Change the Site URL in Supabase under Authentication, URL Configuration, from http://localhost:3000 to your live address. Supabase uses it when your code doesn’t pass a redirect URL. If your code does pass one, that exact URL must also be on the Redirect URLs allow list.
- Supabase key leakMy Supabase service_role key is exposedTreat the key as compromised. Fix how it leaked first, or the new key leaks the same way. Then create a new secret key, replace the old one everywhere, confirm nothing still uses it, and retire the old key. If it was a legacy service_role key, also revoke the legacy JWT secret. If someone is actively using the leaked key, retire it right away and do the leak fix afterwards. Retiring legacy keys also switches off the anon key, so the site breaks until your apps move to a publishable key. Then check your Supabase logs and tables for changes you didn’t make.
- Supabase Auth email errorSupabase “email rate limit exceeded” on sign-upSupabase’s built-in email service is for trying things out. It only sends to members of your project’s team, and Supabase currently limits it to 2 messages an hour, a number it says can change without notice. Set up custom SMTP with an email provider. Supabase then starts you at 30 messages an hour, which you can raise on the Rate Limits page before a launch.
- Supabase Auth errorSupabase sign-in fails with “Email not confirmed”With Confirm Email on, which is the default on hosted Supabase projects, users have to click the link in the confirmation email before their first sign-in. If they never got the email, check sending first: the default email service allows 2 emails per hour on a best-effort basis, so set up custom SMTP for production. Make sure the link goes to your real site rather than localhost:3000, then resend the confirmation with supabase.auth.resend().
- Supabase Auth errorSupabase sign-up fails with “Database error saving new user”Supabase lists three common causes: a trigger or trigger function on auth.users that has an error, a constraint on auth.users that isn’t met, or Prisma breaking the permissions on that table. Open the Auth logs first, then the Postgres logs, for the exact error. A frequent one is a handle_new_user trigger inserting into a profiles table that doesn’t exist.
- Supabase Edge Functions errorSupabase: “Edge Function returned a non-2xx status code”Your function ran but returned an error status (4xx or 5xx). In supabase-js that is a FunctionsHttpError, and the function’s own error message is in error.context: read it with await error.context.json(). Then check the function’s logs in the Logs tab of the Supabase dashboard, fix what the function reports, and have it return a clear status code and message.
- Supabase Auth errorSupabase users logged out with “Invalid Refresh Token”Each Supabase refresh token can be exchanged only once, for a new access and refresh token pair. Reuse is allowed within a 10-second interval by default, or when the parent of the active token is used; otherwise Supabase treats the whole session as terminated and revokes all its refresh tokens. Look for two places refreshing with the same token, such as the server and the browser, or an app restoring an older saved session.
- Supabase Auth errorSupabase: “Email link is invalid or has expired” (otp_expired)Supabase says the most common cause is email prefetching: a mail client or security tool, such as Safe Links in Microsoft Defender for Office 365, opens the link to check it and uses up the one-time token. Stop the link from consuming the token: put the 6-digit code {{ .Token }} in the email template, or link to a page of your own where the user enters it.
- Supabase email problemWhy do my Supabase sign-up emails go to spam, and how do I fix it?Send through your own email provider (custom SMTP in Supabase, such as Resend or Twilio SendGrid) from your own domain, and verify that domain with the provider so SPF and DKIM pass. Add a DMARC record, starting with p=none, or edit the one you already have. Keep the From: address on the domain you verified, ideally the same domain as your website, and leave open and click tracking off for sign-up emails.
- Supabase project pausedMy Supabase project is paused and my app is downOpen the project in the Supabase dashboard and click Resume project. Supabase says a paused project can be restored for up to 1 year after it was paused. Once it’s back, check your tables and row counts and that the app connects again. For an app with real users, move the project to a paid plan: Supabase doesn’t pause paid projects.
- Supabase key questionIs it safe that my Supabase anon key is in my frontend code?Yes, if row-level security is on. Supabase calls the anon key the legacy version of the publishable key, which it says is safe to expose online, and anyone can read it in your code. Row-level security decides what that key can reach, so turn it on for every table and give each table policies. The key that must never be public is the secret or service_role key.
- Supabase security questionI turned on RLS in Supabase. Is my data safe now?Not necessarily. Turning on RLS is the first step, not the last. Check four things: that signed-out visitors don’t have permissions on tables they don’t need, that any view over a protected table follows the table’s rules, that no rule lets everyone read private data, and that the admin key (secret or service_role) never reaches the browser. Then test the rules by signing in as two different users.
- Supabase published app errorSupabase “TypeError: Failed to fetch” on a published appOpen the browser’s Network tab, reload, and look at the failed request. If its address contains undefined, the Supabase URL never reached the published build. If the address is right, check whether the project is paused, whether an Edge Function is failing its CORS check, and whether a browser extension is blocking the call.
- Supabase API errorSupabase “Could not find the table … in the schema cache” (PGRST205)Supabase’s API can’t see a table with that name. First check the table exists in the project your app’s URL points at, since a migration that never ran, or a staging project, would leave it missing. If it exists, reload the schema cache with NOTIFY pgrst, 'reload schema'; in the SQL Editor. If it lives outside the public schema, expose that schema or name it in the query.
- Supabase Auth errorSupabase “provider_disabled”: the sign-in provider is offEnable the provider on its page in the hosted project’s dashboard, with the client ID and secret from the provider. For local development, Supabase configures the provider in supabase/config.toml instead; Simon also restarts the local stack after editing it. For email or phone sign-up, check that those providers are on as well.
- Supabase Auth errorSupabase “Invalid login credentials” when users sign inCheck error.code. If it is invalid_credentials, Supabase doesn’t say which problem it is: it can mean a wrong password, no account with that email, or an account that only signs in through a social login. Send a password reset if the user has a password account, or point them to the button they signed up with. If the code is email_not_confirmed, it’s a different problem.
- Supabase query problemSupabase select returns an empty array, but the data is thereCheck that the table has a select policy: with RLS on and no policy, the API returns no data to the browser key. Then check who the request runs as: a signed-out user or an expired session matches no rows in a policy based on auth.uid(). If you get error 42501 instead of an empty array, the role is missing a grant.
- Supabase Auth errorSupabase: “Signups not allowed for this instance”In the Supabase dashboard, open the Authentication settings for sign-ups and turn Allow new users to sign up back on for the project your production app uses. For email sign-ups, check the email provider is enabled too. When you run Supabase locally, set enable_signup = true in supabase/config.toml and restart. Then test a fresh sign-up.
- Supabase RLS errorSupabase: “infinite recursion detected in policy for relation”A policy on the table queries a table with RLS on, and that table’s policy ends up running the first policy again. Move the lookup into a security definer function in a private schema, with search_path set to an empty string, and call it from both policies. Don’t fix it by turning RLS off or by writing a policy that lets everyone read.
- Supabase API errorSupabase: “permission denied for table” (42501)Run the grants query below to see what anon and authenticated may do on the table, then grant only what the app needs, such as select to anon or select, insert, update and delete to authenticated. Keep RLS on with policies, because the grant opens the table to the API. If the error names a sequence, column or schema, see the matching cause below.
- Supabase Auth errorSupabase Auth: over_request_rate_limit (429 Too Many Requests)Check the network tab for repeated auth calls and fix the loop first. If sign-ins go through your own server, every user shares its IP: turn on IP address forwarding and send the user’s IP, or sign in from the browser. Raise the limits in Authentication > Rate Limits only for real traffic.
- Supabase API errorSupabase: “Invalid API key”Copy the publishable key and the URL from the same project’s API settings, set them in your host’s environment variables, and rebuild. If legacy keys were disabled or a key was rotated, move the app to the current publishable key. Send keys on the apikey header.
- Supabase Auth errorSupabase Auth: Email address not authorizedSet up a custom SMTP provider in the Auth settings, with the host, port, user, password and a From address from your email service. After that, confirmation, magic link and reset emails go to any address. For a quick test before then, add the tester to the organization’s team.
- Supabase performance problemWhy is my Supabase app so slow?Open the Performance Advisor and the Query Performance report in the Supabase dashboard. They point at missing indexes and security policies that check every row. Then check that the app fetches only the rows and columns it shows, that the project region is near your users, that serverless code uses the shared pooler, and whether CPU is maxed out before paying for more compute.
- Supabase billingSupabase bill spike: find what you’re paying for, then cut itOpen the organization’s Usage page and the Upcoming Invoice on the billing page, then pick each project from the dropdown to see which one grew. Compute is billed per project by the hour and is not covered by the Spend Cap, so an extra project, a bigger compute size, branches or read replicas add to the bill on their own. For usage lines such as egress or realtime messages, find the query or listener sending too much data and fix that first. Then delete projects you don’t use.
- Supabase database problemSupabase database full: the app reads but won’t saveCheck disk usage on the project’s Infrastructure page. On the Free plan the disk is a fixed 2 GB, so delete data you don’t need or upgrade. On a paid plan the disk grows on its own at 90%, but only four times in a rolling 24 hours, so a fast-growing database can still hit 95%; increase the disk size yourself. To delete data while read-only, Supabase documents a session override that turns writes back on for that session.
Firebase
- Firebase and Google Cloud billingUnexpected Gemini API charges on a Firebase projectStop the spend first: in the Google Cloud console, check the API metrics to see which key made the calls, then rotate it and restrict or delete every API key that allows the Generative Language API and isn’t a server-side Gemini key. If you can’t find the key fast, you can disable billing on the project, after checking what that takes down. Then move Gemini calls behind your server, turn on App Check for the Firebase services your app uses, set spend cap budgets where Firebase supports them, and keep an alert budget on the whole project. Budget alerts only warn you; they don’t stop charges.
- Firebase Firestore errorFirebaseError: Missing or insufficient permissionsUsually the query asks for more than the rules allow. Firestore rules are not filters: if a query could return even one document the user can’t read, the whole query fails. Add the same conditions to the query that the rule checks. Next most likely: no user is signed in yet when the request runs.
- Firebase security alertFirebase says your Cloud Firestore database isn’t properly securedYour rules let people read or write data they shouldn’t, commonly test-mode rules left on after launch, or rules that only check that someone is signed in. Replace them with rules that check who owns each document, test them in the Rules Playground or the emulator, then deploy. Paste your rules into the free checker first to see which patterns you have.
- Cloud Firestore quota errorWhy does Firestore say I exceeded my quota?On the free quota, Firestore allows 50,000 document reads, 20,000 writes and 20,000 deletes a day, reset around midnight Pacific time. Hitting the limit usually means something reads the same documents again and again: listeners that reconnect, queries with no limit, offsets, security rules that read other documents, or counting by downloading a whole collection. Check the Usage tab, fix the biggest source, or enable billing and set a budget alert, knowing a budget doesn’t cap usage.
- Firebase Authentication errorFirebase: Error (auth/unauthorized-domain)The domain your app runs on isn’t in Firebase’s Authorized domains list. In the Firebase console, go to Authentication, then Settings, then Authorized domains, and add the exact domain, such as app.example.com, without https:// or a path. Every preview URL counts as its own domain, and projects created after April 28, 2025 no longer include localhost.
- Firebase Authentication errorWhy does Firebase sign-in fail with auth/invalid-credential?With email enumeration protection on, which is the default for projects created on or after September 15, 2023, Firebase returns auth/invalid-credential for any invalid email or password, so your app can’t tell which one was wrong. Show one message such as “Email or password is incorrect”, stop branching on user-not-found and wrong-password, and make the forgot-password flow say an email was sent if the account exists. With protection on, a mistyped password gets the same code, so check the password too.
- Android sign-in errorWhy does Google Sign-In fail on Android with error 10?Status 10 means Google says the app is misconfigured. If sign-in works in debug but fails from Google Play, register the app signing key fingerprint from Play Console, not just your upload key. Add the debug fingerprint too, pass the Web client ID as the server client ID, and if you use Firebase, download the updated google-services.json after adding fingerprints.
- Firebase App CheckFirebase App Check is blocking my app’s requestsOnce App Check is enforced, every request without a valid App Check token is rejected. Localhost and CI need a debug token registered in the console, older app versions without the App Check SDK fail until users update, and enforcement can take up to 15 minutes to apply. Check the App Check metrics before enforcing, and keep it on.
- Firebase Auth errorFirebase: auth/requires-recent-login when deleting an accountDeleting an account, setting a primary email address and changing a password need a recent sign-in. If the user signed in too long ago, Firebase fails the action with auth/requires-recent-login. Ask the user to sign in again, pass the new credential to reauthenticateWithCredential, then retry the action. Account deletion can also run on your server with the Admin SDK, which deletes users by uid.
- Firebase Storage errorFirebase Storage “storage/unauthorized” errorYour Cloud Storage Security Rules rejected the request. Check three things: the rules allow this action for this user, the user is signed in before the upload starts, and the file path matches the path in the rule, such as a folder named after the user’s uid. Don’t fix it by allowing everyone: Firebase strongly recommends restricting access.
- Firebase Auth errorFirebase “auth/too-many-requests”Firebase refused the request because a request limit was exceeded. New accounts are limited per IP address, and SMS codes are limited per IP address and per phone number. During testing, use fictional phone numbers from the Firebase console and stop scripts that create accounts in a loop; for real users, wait and retry later.
- Firebase Auth errorFirebase “auth/operation-not-allowed”The sign-in provider your code calls is disabled for this Firebase project. In the Firebase console, open Security, then Authentication, then the Sign-in method tab, and enable the provider: Email/Password, Email link, Phone, Anonymous or a social provider. For email link sign-in, Email/Password must be on as well.
- Firebase Auth errorFirebase: “auth/invalid-api-key”, but only in productionThe deployed build has no Firebase API key, or the wrong one. Add the key to your host’s production environment variables using the prefix your framework exposes to the browser (VITE_ for Vite, NEXT_PUBLIC_ for Next.js, EXPO_PUBLIC_ for Expo), then rebuild and redeploy, since the value is baked in at build time. Copy it from the web app’s config in the Firebase console.
- Firebase phone sign-in errorFirebase phone sign-in: “auth/invalid-app-credential”Firebase couldn’t verify that the request came from your app. On the web, add your live domain to Authorized domains and render the reCAPTCHA verifier before calling signInWithPhoneNumber. On Android, add the app’s SHA-256 fingerprints, including the Play App Signing one, in the Firebase console. On iOS, upload an APNs key and add the URL scheme so the reCAPTCHA fallback works.
- Firebase and Google Cloud billingFirebase bill spike: find what’s spending, then stop itOpen Usage and billing in the Firebase console settings to see which product grew, then that product’s Usage tab for the day it started. Look for a query with no limit, a listener that reloads data, or a Cloud Function that writes to the document that triggered it (for Gemini charges, see the API key page linked below). Fix that code first. Then set an alerts-only budget on the project and spend cap budgets on the services that support them, because an alert email doesn’t stop anything by itself.
- Firebase Auth errorFirebase: “auth/account-exists-with-different-credential”Keep the credential from the error, ask the user to sign in with the method they used before, then call linkWithCredential to add the new provider to that account. After that, both sign-in methods open the same user.
- Firebase Auth errorFirebase: “This operation is restricted to administrators only” (auth/admin-restricted-operation)In the Firebase console, open Authentication, Settings, and look for the user actions options (create and delete) to see whether they are turned off. Turn them back on if users should sign up themselves. If accounts are meant to be created by an admin only, keep it off, create users with the Admin SDK on your server, and show a clear message in the app.
- Firebase Auth errorFirebase: “The action code is invalid” (auth/invalid-action-code)Treat the code as single use. On your handler page, read oobCode from the URL, check it once with verifyPasswordResetCode or checkActionCode, and only then apply it. If the check fails, show a button that sends a new email instead of a raw error.
- Firebase Hosting errorFirebase Hosting: “Page Not Found” after deploy or on refreshSet "public" in firebase.json to the folder your build writes, such as dist or build/web, run the build, then deploy again. If the home page works but other routes fail on refresh, add a rewrite of "**" to "/index.html". If the screen says Site Not Found instead, the site or custom domain usually has no live deploy yet, or the domain isn’t connected.
- Firebase Auth errorFirebase Auth: auth/billing-not-enabled on phone sign-inMove the project to the Blaze plan in the Firebase console and set a budget alert first. A budget alert only notifies you; it doesn’t cap the bill. For development and app review, add fictional phone numbers with fixed codes, which send no SMS. Then retry the sign-in.
- Firebase Storage errorFirebase Storage: 402 or 403 because the project is on the Spark planUpgrade the project to the Blaze plan by linking a Cloud Billing account; no-cost usage still applies. Set a budget alert at the same time. Your files stay in the bucket and come back once the project is on Blaze.
- Firebase Auth errorFirebase Auth: “Unable to process request due to missing initial state”On Firebase Hosting with a custom domain or a web.app subdomain, set authDomain to that domain. On other hosts, proxy /__/auth/ to your firebaseapp.com domain, or switch to signInWithPopup. Add your domain to Authorized domains and https://your-domain/__/auth/handler to the provider’s redirect URIs, then redeploy.
- Firebase Auth errorFirebase Auth: auth/user-token-expiredCatch the error wherever the app gets an ID token or calls Firebase, sign the user out cleanly, and send them to the sign-in screen with a short message. Don’t retry the same session: it can’t be refreshed.
- Firebase Auth errorFirebase Auth: auth/unauthorized-continue-uriIn the Firebase console, open Authentication settings and add the continue URL’s domain under Authorized domains. Then check the url in ActionCodeSettings: it must be a full, valid URL on that domain, and each environment should send its own.
- Firebase Auth errorFirebase Auth: auth/quota-exceededCompare the app’s SMS volume with the Authentication limits page to see which limit was hit. Move testing and App Review to fictional phone numbers, block bots before the SMS is sent, and upgrade to Identity Platform if real traffic needs more SMS per day.
- Firebase migrationFirebase Dynamic Links stopped workingIf you only need to open the app from a link once it is installed, use Android App Links and iOS Universal Links on a domain you control. If you also need deferred deep links, click analytics or short links, Firebase points to other deep-linking providers. For email link sign-in in a mobile app, update the Firebase Auth SDK, remove dynamicLinkDomain, and set linkDomain if the project uses a custom Firebase Hosting link domain. Then ship app updates.
- Firebase Authentication problemWhy don’t users get the Firebase password reset email?Check four things: that the address has an account in Firebase Authentication (with email enumeration protection on, Firebase sends nothing for an unknown address and shows no error), the spam folder, your project’s daily limit for password reset emails, and the sender domain. Sending from your own verified domain, set in the Authentication Templates tab, puts your brand in the From field and lets you add the DNS records that show the email is yours.
- Firebase Cloud Messaging problemWhy aren’t my Firebase push notifications arriving?Check the phone first: on Android 13 and higher a new install has notifications off until the user allows them, and on iPhone the app has to ask and the user can say no. If no iPhone gets anything, check that an APNs authentication key is uploaded in the Firebase console. If messages arrive late, send user-facing ones at high priority. If they stopped for some users, the server may be sending to old device tokens.
- Firebase Authentication problemWhy do my Firebase users keep getting logged out?Firebase sessions are long lived: refresh tokens expire only when the user is deleted or disabled, or after a major account change such as a new password or email address. If users drop at other times, check how the web app stores the session (local, session or none), whether users switch between domains, whether the app reads the current user before Firebase has loaded it, and whether a server keeps the one-hour ID token without refreshing it.
- Firebase Hosting problemWhy does my Firebase site still show the old version after I deploy?Open Hosting in the Firebase console and check the Release history for the site your domain uses. If the latest release isn’t there, the deploy went to a preview channel, another site or another project. If it is there, check that the build ran before the deploy and that firebase.json deploys the build folder, then look for an old service worker or long browser cache headers.
- Firebase performance problemWhy is my Firebase app so slow?Add Performance Monitoring to see which screen or request is slow. Then check that each list query has a limit and pages through results with a cursor, that the Firestore location is near your users, and that slow buttons aren’t waiting on a Cloud Function cold start, which a minimum number of instances reduces. For slow writes to a busy collection, check for document IDs that increase in order.
Stripe
- Stripe webhook errorNo signatures found matching the expected signature for payloadUsually it’s the wrong signing secret. Each endpoint has its own, the Stripe CLI prints a different one, and the same endpoint has a different secret for test and live keys. Next most likely: your framework parsed or changed the request body before the signature check ran.
- Stripe fulfillment bugThe payment went through in Stripe, but the account was never upgradedMost often, the upgrade only runs on the success page after checkout. Stripe says you can’t rely on that page, because customers don’t always reach it. Grant access from a webhook instead, and check the webhook’s delivery log in the Dashboard for failed deliveries.
- Stripe Billing errorWhy is my Stripe subscription stuck as incomplete?A new subscription stays incomplete until its first invoice is paid, and the customer has 23 hours. If the PaymentIntent says requires_action, the card needs authentication: listen for invoice.payment_action_required, send the customer through it with stripe.handleNextAction, and give access only after invoice.paid or an active subscription. After 23 hours it becomes incomplete_expired and you need a new subscription.
- Stripe payment errorWhy does Stripe say the card was declined with generic_decline?Stripe’s docs say card issuers categorize most declines as generic, so the exact decline reason is often unclear. Check the payment in the Dashboard: if Radar blocked it, the outcome type is blocked and you can allow-list a legitimate payment; if the issuer declined it, read the decline code and the advice code, ask the customer to re-enter the card or use another one, and don’t keep retrying a charge the advice code says not to.
- Stripe BillingStripe subscription canceled, but access in the app is wrongWhen a customer cancels at period end, the subscription runs on until the period ends, cancel_at_period_end becomes true, and Stripe sends customer.subscription.updated. Access should continue until the period ends. When it ends, Stripe sends customer.subscription.deleted: that’s the event that should remove access. Removing access on the update cuts people off early; ignoring the deleted event never removes it.
- Stripe BillingThe Stripe customer portal isn’t workingSave the portal settings in the Dashboard for live mode as well as sandboxes, since each has its own. Create a new portal session on your server each time a signed-in customer clicks the button, because session links are short-lived, and pass a return_url unless a default is set. Then listen to subscription webhooks so the app follows what customers change.
- Stripe Billing errorStripe: “This customer has no attached payment source or default payment method”For a subscription invoice, Stripe uses the subscription’s default_payment_method; if that isn’t set, it uses the customer’s invoice_settings.default_payment_method or default_source. The error means none of them gave it a payment method to charge. Set the card as the customer’s invoice_settings.default_payment_method, and for new subscriptions set payment_settings.save_default_payment_method to on_subscription.
- Stripe go-live problemStripe checkout works in test mode but fails in live modeTest mode (Stripe now calls it a sandbox) and live mode have separate API keys, products, prices, customers and webhook endpoints, and nothing created in one exists in the other. Put the live keys in your production app’s settings, recreate your products and prices in live mode and switch their IDs in your code, add a live webhook endpoint with its own signing secret, then redeploy.
- Stripe webhook problemStripe shows the webhook delivered, but my database never updatedA 200 only means your endpoint received the event. Stripe asks you to answer quickly, before the slow work, so if the database write fails after that, Stripe still records a successful delivery and has nothing to retry. Your own server logs hold the error. Check them first, then the event types your endpoint receives, repeated or out-of-order events, and the database key your handler writes with.
- Stripe Connect errorStripe Connect: the seller finished onboarding, but charges or payouts are still disabledComing back to your return URL doesn’t mean onboarding is complete. Retrieve the account and read its requirements: as long as requirements.currently_due isn’t empty, it can be restricted. Send the seller back through a new Account Link when anything is due, create a fresh link from your refresh URL when one expires, and listen to account.updated so new requirements don’t surprise you.
- Stripe payments problemStripe: Apple Pay or Google Pay doesn’t show in the payment formRegister every domain and subdomain that shows the payment form on the Payment method domains page, in live mode as well as each sandbox. If the form sits in an iframe, register both domains. Then confirm the wallets are on in your Payment Method Configurations, and test on a device with a real card in the wallet, outside a private window.
- Stripe go-live errorStripe: “Invalid API Key provided”Check the key your production server actually sends. It must be a secret key (sk_live_) or a restricted key (rk_live_) from the Stripe account you mean, copied whole, and not one that was rolled or deleted. Copy the current key from the Dashboard into your host’s production environment variables, redeploy, and never put it in front-end code.
- Stripe payment errorStripe: “Your card has insufficient funds” (insufficient_funds)If the decline_code is insufficient_funds, the customer’s bank said no. Nothing in your code is broken: show a clear message and let them pay with another card or method. If the error code is balance_insufficient, or a refund is stuck as pending, your own available Stripe balance is too low: wait for pending funds to become available (or top up if a refund is pending), then retry.
- Stripe decline codeStripe: “The card doesn’t support this type of purchase” (card_not_supported)Don’t retry the same card. Show a message that asks the customer to use another card or payment method, or to contact their bank, and keep checkout open for a second try. For a renewal, ask the customer to update the saved card. If many customers hit it, look for a pattern in the declined payments in the Dashboard.
- Stripe Checkout errorStripe: “This Checkout Session has expired”Create a new Checkout Session each time the customer clicks to pay, and redirect to its URL straight away instead of storing or emailing that URL. Set expires_at if you need a shorter window. To win back customers whose session ran out, enable after_expiration.recovery and listen for the checkout.session.expired webhook.
- Stripe API errorStripe: payment_intent_authentication_failure (3D Secure failed)Catch the error after confirmPayment, show a message that the bank couldn’t confirm the payment, and keep the same PaymentIntent so the customer can try again or enter another card. For redirect flows, read the PaymentIntent on your return page and check for requires_payment_method. Test it with card 4000000000003220 and the Failure button.
- Stripe decline codeStripe: card declined with authentication_requiredCatch the decline, keep the PaymentIntent it returns, and email or notify the customer with a link to a page that calls confirmPayment with that PaymentIntent’s client secret. The bank then shows its check and the payment can succeed. If the same code comes back after a passed check, the customer has to call the bank.
- Stripe decline codeStripe: card declined with do_not_honorShow the customer a plain message that the bank declined the card and ask them to contact the bank or try another card. In the Dashboard, check the payment’s CVC and address checks and the advice code. Don’t retry the same card in a loop, because card networks limit how often one charge can be retried.
- Stripe decline codeStripe: incorrect_cvc (the card’s security code is incorrect)Catch the error after confirmPayment, check for code incorrect_cvc, and show a message on the CVC field instead of a generic decline. Keep the same PaymentIntent so the retry goes through the same payment. In tests, enter a CVC, because Stripe skips the check when it is empty, and use test card 4000000000000127 to see the decline.
- Stripe decline codeStripe: expired_card (the card has expired)Check error.code for expired_card, show a message that the card has expired, and ask for another one, keeping the same PaymentIntent. For renewals, send the customer to a page where they can replace the card, such as the Stripe customer portal, then pay the open invoice with stripe.invoices.pay. A wrong expiry typed at checkout returns invalid_expiry_month or invalid_expiry_year instead, which a retry with the right date fixes.
- Stripe decline codeStripe: processing_error (an error occurred while processing the card)Show a message that the payment didn’t go through and offer a retry with the same PaymentIntent. If it fails a second time, suggest another card. On the server, reuse an idempotency key only when you repeat the same request after a timeout, and use a new key for a new attempt. For subscriptions, Smart Retries or your retry schedule will try the renewal again.
- Stripe decline codeStripe: card_velocity_exceeded (card limit reached)Show a clear message that asks for another card or for the customer to contact the bank. Don’t retry the same card in a loop. For subscriptions, let your retry schedule try again later and email the customer a link to update the card.
- Stripe decline codeStripe: incorrect_zip (billing postal code didn’t match)Show a message that asks for the billing ZIP of the card, not the shipping address. Check that your form sends the postal code field the customer typed, and doesn’t prefill it from the shipping address or a profile. Then let the customer try again.
- Stripe decline codeStripe: lost_card and stolen_card declinesShow a generic “Your card was declined” message and ask for another payment method. Don’t reveal the lost or stolen reason, and don’t retry the card. Log the decline code for your own review, and check whether the account shows other signs of card testing.
- Stripe decline codeStripe: card declined, decline code fraudulentShow the same message as a generic decline and don’t retry the same card. Open the payment in the Dashboard and read the Radar outcome and risk insights. If many fraudulent declines arrive in a short time, treat it as card testing and add rate limits or a CAPTCHA to checkout.
- Stripe decline codeStripe: card declined, currency_not_supportedTell the customer the card can’t pay in this currency and ask for another card or payment method. A bad currency code gives an API error, not this decline. If many customers in one country hit it, offer their local currency, if your account supports it, or another payment method.
- Stripe BillingStripe: subscription stuck in past_dueListen for invoice.payment_failed and email the customer a link to update their card. Update the payment method on the field Stripe charges, then pay the open invoice. In Dashboard settings, choose what happens after the last retry: cancel, mark unpaid, or stay past_due.
- Stripe BillingStripe: a payment was disputed (chargeback)Open the dispute in the Dashboard and note the reason and the due date. If the customer is right, accept it. If not, gather evidence that matches the reason, such as receipts, login records and emails, and submit it all at once. Handle charge.dispute.created so you never miss a deadline.
- Stripe card errorStripe: incorrect_number, the card number is incorrectShow Stripe’s error message next to the card field and let the customer type the number again or pick another card. Collect cards with Stripe Elements or Checkout, which check the number as it is typed. Don’t retry the same card automatically.
- Stripe BillingStripe: webhook events fail to deliver to your endpointOpen the endpoint’s Event deliveries tab and read the status: a redirect, a 4xx, a 5xx, a TLS error or a timeout. Fix that, return 200 before slow work, then list events with delivery_success set to false and process what you missed.
- Stripe payout errorStripe: payout failedOpen the payout in the Dashboard, or read failure_code and failure_message from the API. Re-enter the bank details, or add a new account if the old one is closed. Stripe tries again at the next scheduled payout. If the details are right, contact the bank, then click Resume Payouts.
- Stripe errorStripe: setup_intent_authentication_failureShow the customer that the card could not be verified, and let them try again or enter another card on the same SetupIntent. Don’t save the payment method or start the subscription until the SetupIntent succeeds. In test mode, check the flow with the always-authenticate test card.
- Stripe payment errorStripe: charge_expired_for_captureThe held money is gone, so ask the customer to pay again, or charge a saved payment method with a new PaymentIntent if the customer agreed to that. Then fix the timing: store capture_before for each authorized charge and capture before it, or request an extended authorization when creating the PaymentIntent, where the card is eligible.
- Stripe Payments errorStripe: card_decline_rate_limit_exceededFor one customer, ask them to check their card details or use another card; the same card can be tried again after 24 hours. If many cards fail at once, check for card testing: refund suspicious payments, move to Stripe Checkout or the latest Payment Element, and limit how often your own endpoint can create payments.
- Stripe Billing errorStripe: invoice_payment_intent_requires_actionGet the PaymentIntent on the invoice and have the customer complete it on your site, for example by confirming it with Stripe.js while they are present. For off-session renewals, turn on “Send a Stripe-hosted link for customers to confirm their payments when required” in the Billing email settings.
- Stripe API errorStripe: stripe_tax_inactiveOpen the tax settings page in the Dashboard for the same account the secret key belongs to, click Get started, confirm the head office address, preset tax code and default tax behavior, and add a registration for each place the business is registered to collect tax. Then retry the request with automatic_tax.enabled=true. Existing invoices, subscriptions and payment links need automatic tax turned on one by one.
- Stripe API errorStripe: capability_not_activeRead the capabilities array in the error response to see which capability is missing. Retrieve that capability for the account and check its status and requirements. Request it if it is unrequested, collect everything listed in currently_due and past_due from the connected account, and retry once the status is active.
- Stripe API errorStripe: payment_method_not_availableShow the customer a clear message and offer another payment method, or retry later with the same one. If the server sends an idempotency key, a retry with the same key can return Stripe’s cached response (look for the Idempotent-Replayed: true header) instead of attempting the payment again.
- Stripe Connect errorStripe: payouts_not_allowedRetrieve the connected account and check payouts_enabled and the requirements hash. If requirements are currently_due or past_due, collect that information from the account holder and update the account. Watch account.updated events so the platform knows when payouts are enabled again.
- Stripe payment errorStripe: charge_exceeds_transaction_limitThis is an account limit, not a bug in your code. Contact Stripe Support to request a higher processing limit for the payment method in the error. Until it is raised, offer customers another payment method, such as a card, for payments that hit the limit, and show them a clear message instead of a generic failure.
- Stripe Payments errorStripe: refund_disputed_paymentLook up the dispute for the payment. To give the customer the money back, accept the dispute, which closes it in their favor. To fight it, submit evidence. Before any automatic refund, check the charge’s disputed flag, and handle the refund failure reason that appears when a dispute lands while a refund is still pending.
- Stripe Payments errorStripe: testmode_charges_onlyOpen the Stripe Dashboard with the account that owns the live key and finish setting up the account: verify the business and complete the account checklist. Answer any extra verification request. Until that is done, use test keys and test cards, and keep live keys out of the app.
- Stripe Payments errorStripe: balance_insufficientRetrieve the balance and use the available amount, in the right currency, as the limit for the transfer or payout. Funds from recent payments sit in pending until they become available. For Connect transfers paid from a specific charge, set source_transaction: the transfer goes through right away, and the seller can use the money once the charge’s funds become available.
- Stripe Payments errorStripe: api_key_expiredOpen the API keys page in the Dashboard and get the current key. In live mode Stripe shows a secret or restricted key only once, so if nobody saved it, rotate or create a new one. Then replace the old key in every place that sends it: the server environment, background jobs, webhook handlers and any no-code tools. Redeploy so the new value loads. If the code is platform_api_key_expired, see the Connect platform cause below.
- Stripe Payments errorStripe: payment_method_unactivatedTurn the payment method on in the Dashboard payment methods settings, in the same mode (test or live) and on the same account the code uses. Better still, stop listing payment_method_types in code and let the Dashboard decide which methods show, so the two can’t drift apart.
- Stripe ACH errorStripe: bank_account_verification_failedCheck which failure it was. If the customer ran out of attempts, ask for a different bank account, because Stripe says that one can no longer be verified or reused. If the deposits never arrived or the 10 days ran out, collect the bank details again. To skip microdeposits where possible, verify instantly with Financial Connections.
- Stripe checkout problemWhy did Stripe charge my customer twice?If it is one order, your checkout created a second payment for it. This can happen when the customer presses Pay twice, reloads the payment page, or your server retries a request. Refund the duplicate in the Stripe Dashboard, then fix the cause: create one PaymentIntent per order and reuse it, send an idempotency key based on the order ID with every create request, and disable the Pay button while a payment is in progress.
- Stripe Dashboard statusWhy do my Stripe payments say “Incomplete”?Incomplete means the payment has not finished: it still needs a payment method, a confirmation, or an action from the customer such as a 3D Secure check. Some are expected, because Stripe advises creating the payment as soon as checkout starts, so customers who leave are recorded too. Open a few, check the exact status in Workbench Inspector, and look for a pattern: many stuck at requires_action, or many with the same decline, point to something worth fixing.
- Stripe Billing problemWhy did Stripe charge a customer after they cancelled?Open the customer in the Stripe Dashboard and check three things: that the subscription itself shows as canceled, that the customer has no second active subscription, and whether the charge is a final invoice for prorations or pending invoice items. If the cancel only happened inside your app, Stripe never heard about it. Refund the payment from the Payments page, then fix the cancel flow so it cancels in Stripe.
- Stripe payments problemWhy aren’t my customers getting Stripe receipts?In the Dashboard, open Settings, then Business, then Customer emails, and turn on Successful payments. Then check that each payment has an email address, either a receipt email on the payment or an email on the Stripe Customer. Stripe sends no automatic receipt for test payments, and none for payments that failed. To send one by hand, open a successful payment, go to Receipt history, click the overflow menu and select Send receipt.
- Stripe Billing problemWhy did Stripe charge a customer when their free trial ended?A Stripe trial moves to the regular price automatically when it ends, and a card saved at signup is charged. To stop surprises, turn on the trial ending reminder email, make sure cancelling in your app cancels in Stripe, and for trials that take no card, choose to cancel or pause the subscription at the end instead of invoicing. Refund the payment from the Payments page if the customer didn’t mean to stay.
Mobile apps: crashes, push and in-app purchases
Guide: Expo and Flutter build errors
- Expo runtime errorExpo app works in Expo Go, crashes in TestFlight or productionExpo calls this a runtime error: the build succeeded, but the app crashes or hangs when it runs. Start with the crash report, in the Google Play Console’s crashes section or Xcode’s Crashes Organizer for TestFlight and App Store builds. Then capture native logs from a device that crashes, and run the app locally in production mode with npx expo start --no-dev --minify. Simon also checks that the environment variables the app reads are set for the build, then narrows it down change by change.
- Expo push notification problemWhy are my Expo push notifications not arriving?Send one test notification and read the push receipt: it names the error. A notification that works in development but not in the store build usually means the production credentials are missing or wrong: on Android, the FCM V1 service account key uploaded to EAS; on iOS, the push key. Then check that the app still holds a valid token for that user, and send with high priority on Android.
- RevenueCat errorRevenueCat: “None of the products registered in the RevenueCat dashboard could be fetched”The store returned none of the products RevenueCat asked for. Check, in order: the store agreements and banking are complete, the product IDs and bundle ID or package name match exactly, the SDK uses the API key for this platform, and the products are ready to load where you test: a physical device, a sandbox account on iOS, and a closed track on Google Play.
- App Store Connect statusTestFlight build stuck on “Missing Compliance”Open the build in App Store Connect and answer the export compliance questions about encryption. To stop it on every upload, add the ITSAppUsesNonExemptEncryption key to Info.plist: NO if the app uses no encryption or only exempt encryption, YES if it uses non-exempt encryption, and then you typically also add the code Apple gives you after reviewing your documentation.
- In-app purchase problemWhy did a user pay in my app but not get premium?Check that the app completes every purchase after unlocking it (finish on iPhone, acknowledge on Android within three days), listens for purchases that finish while the app is closed or later, such as Ask to Buy and pending payments, has a working Restore Purchases button, and, with RevenueCat, that each product is attached to the entitlement the app checks and the user is signed in as the same customer.
Hosting, sign-in and Google Maps
- Vercel errorVercel function times out with a 504 (FUNCTION_INVOCATION_TIMEOUT)Vercel returns a 504 when a function doesn’t finish within its maximum duration, and that time includes sending the response. Find the slow call first: an AI or outside API with no timeout, a slow database query, or a database far from the function’s region. Then stream the response or move long work to a background job. Raise maxDuration last, within your plan’s limit.
- Netlify errorNetlify Function times outNetlify gives synchronous functions 60 seconds and scheduled functions 30 seconds, and neither limit is configurable. So find the slow part instead of looking for a setting: an outside call with no timeout, a database far from the function’s region, or heavy work inside the function. Work that takes longer belongs in a background function, which answers with 202 straight away and can run for up to 15 minutes.
- Vercel deploymentVercel environment variable is undefined in productionVercel applies a new or changed variable only to new deployments, so redeploy (a new build) after saving it. Check it’s enabled for the right environment, Production or Preview. And if the browser needs it, it must use your framework’s public prefix, such as NEXT_PUBLIC_ in Next.js, which also means it’s visible to anyone, so never put secrets there.
- Google sign-inGoogle sign-in: Error 400: redirect_uri_mismatchGoogle only sends users back to addresses listed under Authorized redirect URIs on your OAuth client in the Google Cloud console, and the match must be exact: http or https, upper and lower case, and the trailing slash. Copy the exact callback address your auth provider uses, add it to the OAuth client the app actually uses, and try again.
- Netlify and Vercel deploy error404 "Page not found" when you refresh a page or open a linkYour app is a single-page app: one index.html that handles every route in the browser. When you refresh /dashboard, the host looks for a file at that path, finds none and returns 404. Add a rewrite that serves index.html for every path: a _redirects file on Netlify, or a vercel.json rewrite on Vercel.
- Vercel functions errorVercel FUNCTION_INVOCATION_FAILED (500 Internal Server Error)The function crashed while handling the request. Open the deployment’s Logs tab and find the error from the failing request. Typical causes are an error your code didn’t catch, or a setting such as an environment variable that’s missing in production. Fix that, wrap the handler in try/catch so it returns a real error response, and redeploy.
- Vercel deployment errorVercel “404: NOT_FOUND” on a deployment that says ReadyFirst open the .vercel.app URL. If it works and your own domain doesn’t, fix the DNS records. If both 404, check that the Framework Preset isn’t set to Other and that the Output Directory matches where your build writes files, then redeploy. If only dynamic routes 404, set dynamicParams or render them dynamically.
- Google Maps errorGoogle Maps shows “for development purposes only” or won’t loadOpen the browser console on the page with the map and find the error ending in MapError. RefererNotAllowedMapError means your site’s address isn’t on the key’s allowed websites. ApiNotActivatedMapError means the Maps JavaScript API isn’t enabled. BillingNotEnabledMapError means the project has no billing. InvalidKeyMapError means the key in the page isn’t recognised.
- Vercel domain errorVercel “ERR_TOO_MANY_REDIRECTS” with Cloudflare in frontA common cause is Cloudflare’s Flexible mode: in Cloudflare’s SSL/TLS settings, change Flexible to Full (strict), which needs a valid certificate on the Vercel side. If the loop continues after that, look for app redirects, Vercel domain redirects and Cloudflare rules that send the same address back and forth.
- Vercel access problemVercel preview link asks visitors to log in to VercelSend people the production domain, or a Shareable Link for the preview deployment. For tools such as tests or webhooks, use Protection Bypass for Automation with its secret in a header. Turning protection off makes every existing deployment public, so Simon keeps it on.
- Google sign-in warning“Google hasn’t verified this app”If the app only needs sign-in (name, email, profile), request just those scopes and publish the app to In production: Google treats those scopes as non-sensitive, so they don’t need verification. If it needs Gmail, Drive or Calendar data, list the exact scopes on the consent screen and submit the app for verification. Restricted scopes, such as most Gmail access, also need a security assessment. Until then, users see the warning and the app is capped at 100 new users.
- Google sign-in errorGoogle sign-in: “Access blocked” Error 403: disallowed_useragentGoogle sign-in must open in the device’s real browser or through Google’s native sign-in SDK, not inside a WebView. In your own app, switch to the native Google Sign-In SDK, or open the sign-in URL with ASWebAuthenticationSession on iOS and Custom Tabs on Android. For visitors arriving from social apps, detect the in-app browser and ask them to open the page in their browser, and offer email sign-in as a fallback.
- Vercel billingVercel bill spike: find what’s using it, then cap itOpen the Usage page, pick each project from the dropdown, and review the breakdown by project and region. Then look for the cause behind the metric that grew: an app that polls or refetches data, functions called again and again even when they fail, images transformed without a long cache time, or attack traffic. A spend amount alone only sends alerts, so also turn on pausing production deployments when it is reached.
- Netlify billingNetlify credits ran out or the bill spiked: find what’s using themOpen the team’s usage page in the Netlify dashboard and see which line grew: web requests, compute (functions and similar), bandwidth or production deploys. If the sites are paused, a paid plan can buy a credit pack to bring them back, and a free plan has to upgrade or wait for the next billing cycle. Then fix the cause: an app that polls or refetches, functions that run long, or bot traffic you can block with Firewall Traffic Rules. Auto recharge is off by default, so a paused site means the plan limit was reached, not that a payment failed.
AI connectors and AI APIs
- OpenAI API errorOpenAI API 429: insufficient_quotaRead error.code in the response, not only the 429 status. credit_balance_exhausted means add credits; organization_spend_limit_exceeded or project_spend_limit_exceeded means raise or remove that limit. organization_usage_limit_exceeded means OpenAI set the limit, so request a higher approved limit or contact support. If the code points to a rate limit instead, slow down and follow Retry-After. In your app, treat billing codes as non-retryable and alert yourself, so users don’t sit through retries that can’t succeed.
- Claude custom connectorAuthorization with the MCP server failedWork from the outside in. First make sure Claude can reach the URL at all: a public address, no firewall or bot rule blocking Anthropic, and no redirect to another host. Then check the sign-in flow: an unauthenticated request must get a 401 that points to your OAuth metadata, your authorization server must support S256 PKCE and a way for Claude to register, and discovery, registration and token responses must each arrive within 10 seconds.
- ChatGPT MCP connectorChatGPT MCP connector: OAuth redirect or sign-in failsCopy the redirect URI from the MCP server’s page in ChatGPT and allowlist exactly that address on your authorization server. ChatGPT uses one of two addresses, depending on whether your authorization server supports issuer identification. Then check that your authorization server metadata advertises S256 PKCE and a way for ChatGPT to get a client ID: CIMD, dynamic client registration or a predefined client.
- MCP transport errorMCP server: 405 Method Not Allowed or 406 Not AcceptableServe one MCP endpoint that accepts POST and answers with JSON or an event stream. A 405 on GET is allowed by the spec for a server that doesn’t stream to the client, so only POST needs to work. A 406 comes from servers, often built on an SDK, that reject requests whose Accept header lacks application/json or text/event-stream. A 403 usually means your Origin or Host check rejected the request.
App Store rejections
Guide: App Store and Google Play rejections
- App Store rejectionGuideline 2.1 - Performance - App CompletenessTwo common causes of 2.1 rejections are a reviewer who couldn’t sign in and a build that crashed on their device. Add a demo account that works to the review notes, keep your back end running during review, and test the exact build on a real device before you resubmit.
- App Store rejectionGuideline 2.3 - Performance - Accurate MetadataGuideline 2.3 is about the listing, not the code: Apple wants the screenshots, previews, description and other metadata to match the app’s core experience. Common fixes are screenshots of the app in use rather than a splash or login screen, previews made only of screen captures, clear notes on what costs extra, and removing other platforms’ names and icons. Check which item the reviewer named, change it, then resubmit.
- App Store rejectionGuideline 4.3(a) - Design - SpamA 4.3 rejection means Apple sees your app as a copy: several near-identical apps under different Bundle IDs (4.3(a)), or an app indistinguishable from what’s already widely available (4.3(b)). The fix is usually a product change, such as merging the copies into one app, or making the app meaningfully different, not a code fix.
- App Store rejectionGuideline 5.1.1 - Legal - Privacy - Data Collection and StorageThe most common 5.1.1 rejections: no way to delete an account inside the app (5.1.1(v)), purpose strings that don’t explain why the app needs a permission (5.1.1(ii)), a missing privacy policy link (5.1.1(i)), or a permission demanded before the user can do something unrelated (5.1.1(iv)). Each is a fixable change in the app or its listing.
- App Store rejectionGuideline 4.2 - Design - Minimum FunctionalityA 4.2 rejection means Apple sees the app as a repackaged website, or not useful or app-like enough. The rule doesn’t ban web views; it asks for features, content and UI beyond the website. The fix is a product change: give people a reason to install it, shown clearly in the build and the review notes.
- App Store rejectionGuideline 3.1.1 - Business - Payments - In-App PurchaseIf people pay to unlock something inside the app, such as a subscription, premium content or credits, Apple requires in-app purchase. Physical goods and services used outside the app must use another payment method, such as Stripe or Apple Pay. Links to buy on your website are allowed in United States storefront apps; elsewhere they generally need one of Apple’s entitlements.
- App Store rejectionGuideline 3.1.2 - Business - Payments - SubscriptionsApple lists what a subscription sign-up screen must include: the subscription’s name and length, what people get during it, the full renewal price shown clearly, and a way to sign in or restore purchases. Your app and App Store metadata must also link to your Terms of Use and Privacy Policy, and the amount billed must be the most prominent price. Check which clause the rejection quotes first: 3.1.2(a) is about the subscription itself, not the screen.
- App Store rejectionGuideline 4.8 - Design - Login ServicesIf the main account in your app is set up through a third-party or social login, Apple requires an equivalent option from another login service that limits data to name and email, lets people keep their email private, and doesn’t collect app activity for ads without consent. The guideline names no service; it lists the features. Apps that only use their own email and password sign-in don’t need it.
- App Store Connect upload errorApp Store Connect: “ITMS-90725: SDK version issue”Build with Xcode 26 or later, which uses the iOS 26 SDK. Check with xcodebuild -version on the machine that makes the build, and on Codemagic, GitHub Actions, Bitrise or EAS pick an image with Xcode 26. Then increase the build number and upload again. Plan for the next step too: Apple says uploads need the iOS 27 SDK starting April 2027.
- App Store Connect upload errorApp Store Connect: “ITMS-91053: Missing API declaration”Find which code uses the API named in the email. If it is your own code, add a PrivacyInfo.xcprivacy file to the app target in Xcode and list the API category with an approved reason under NSPrivacyAccessedAPITypes. If it is a third-party SDK, update it to a version that ships its own privacy manifest. Then rebuild and upload.
- App Store Connect upload errorApp Store Connect: “ITMS-90683: Missing purpose string in Info.plist”Read the key names in the email. For each one (NSCameraUsageDescription, NSMicrophoneUsageDescription, NSPhotoLibraryUsageDescription, NSLocationWhenInUseUsageDescription, NSContactsUsageDescription and so on), add that key to the app target’s Info.plist, or in Xcode under the target’s Info tab, with one plain sentence saying what the app does with the access, then upload a new build. If the app never uses that resource, the code that touches it is usually inside a plugin or SDK: remove or replace that dependency, or add an honest purpose string anyway.
- App Store rejectionGuideline 2.5.1 - Performance - Software RequirementsFind the API or framework the rejection names. Remove any private API call, often one that comes from a third-party SDK, update dependencies that still use deprecated frameworks, and turn off capabilities the app doesn’t use. If the app uses HealthKit, HomeKit or a similar framework, say so in the App Store description. Then reply in App Store Connect and resubmit.
- App Store rejectionGuideline 5.1.2 - Legal - Privacy - Data Use and SharingShow the App Tracking Transparency prompt before any tracking starts, and stop tracking when the user says no. Before sending personal data to a third-party AI service, say so in the app and ask for permission. Don’t lock features, content or rewards behind turning on tracking, location or notifications, and use data only for the purpose you asked for it.
- App Store rejectionGuideline 4.0 - DesignRead the rejection note and open the screen and device it mentions. Fix that screen, not the whole app: stop asking for a name or email after Sign in with Apple, keep web logins inside the app with ASWebAuthenticationSession, and check every screen on an iPad. Then reply in App Store Connect with what changed and upload the new build.
- App Store rejectionGuideline 1.5 - Safety - Developer InformationOpen your Support URL in a private browser window on a phone. It should load without a login and show a way to contact you, such as an email address or a contact form. Fix the page or change the URL in App Store Connect on the rejected version, in every language you list, then reply to App Review. The URL can usually change without a new build.
- App Store rejectionApp Store Guideline 2.5.4: background modes used for the wrong purposeOpen Signing & Capabilities in Xcode and check each Background Mode. Remove any mode the app doesn’t need, including ones a plugin added. For a mode you keep, reply in App Store Connect with where the feature is and how to trigger it, and add a short video if it is hard to find.
- App Store Connect upload errorITMS-91061: missing privacy manifest for a third-party SDKRead the SDK names in Apple’s email. Update each one to a version that includes PrivacyInfo.xcprivacy, then update the Flutter, React Native or Capacitor plugin that bundles it. Run pod update or flutter pub upgrade, rebuild, and upload a new build. Remove any listed SDK the app no longer uses.
- App Store rejectionApp Store Guideline 1.2: user-generated content without moderation toolsAdd all four tools Apple lists: filter objectionable content before it is posted, let users report a post and act on reports quickly, let users block another user, and show contact details in the app. Then reply in App Store Connect with where each tool is and how to reach it.
- App Store rejectionApp Store Guideline 2.5.2: app downloads or runs code that changes its featuresFind what Apple saw: a remote update, downloaded scripts, or code the app generates and runs. Put every feature in the reviewed build, keep over-the-air updates to fixes, and if users run small HTML5 or JavaScript apps inside yours, follow Guideline 4.7. Then reply to App Review with what changed.
- App Store rejectionApp Store Guideline 4.5.4: push notifications required or used for marketingLet people use the app after they decline notifications. Add a separate marketing toggle with plain consent text, off until the user turns it on, and an in-app setting to turn it off again. Keep sensitive details out of the notification text. Then submit a new build with notes on what changed.
- App Store rejectionApp Store Guideline 5.2.1: protected third-party material used without permissionFind every place the other brand appears: the app name, subtitle, icon, screenshots, keywords and in-app screens. Replace it with your own name and art, or attach written permission from the owner in App Review notes. Then resubmit and reply with what changed.
- App Store rejectionApp Store Guideline 4.1: the app copies or impersonates another appCompare your name, icon, screenshots and main screens with the app Apple has in mind, and change whatever reads as a copy. Remove any brand you don’t own, or attach written approval from its owner in App Review notes. Then resubmit and reply in App Store Connect with what changed.
- App Store rejectionApp Store Guideline 4.2.3: the app needs another app, or downloads without askingMake the core features work after a fresh install with no other app present. Move anything that needs a companion app behind an optional step. If the app downloads models, levels or media on first launch, show the size and a button to start it. Then resubmit and explain the change in App Review notes.
- App Store rejectionApp Store Guideline 5.1.5: Location Services not justified or not explainedTie location to one visible feature, request it when the user starts that feature, and ask for When In Use unless the feature truly needs Always. Write a purpose string that names the feature. Remove location from SDKs that don’t need it, then resubmit and explain the change in App Review notes.
- App Store rejectionApp Store Guideline 2.3.8: metadata not appropriate for all audiencesReplace any icon, screenshot or preview frame with violence, mature themes or crude content with images a 4+ audience can see. Remove “For Kids” and “For Children” from the name, subtitle, icon, screenshots and description unless the app is in the Kids Category. Keep the app name and every icon similar, then resubmit and say what changed in App Review notes.
- App Store rejectionGuideline 1.3 - Safety - Kids CategoryPut every link out of the app and every purchase behind a parental gate. Remove third-party analytics and advertising SDKs unless they meet Apple’s limited exceptions, such as analytics that collect no IDFA or identifiable information about the child. Add a privacy policy. If the app is not really for kids, leave the Kids Category and take “for kids” wording out of the metadata.
- App Store rejectionGuideline 5.1.1(ix) - Legal - Privacy - Data Collection and StorageEnroll the company that provides the service in the Apple Developer Program as an organization, which needs a D-U-N-S Number, then transfer the app to that account or submit it from there. If you build the app for a client, the app should be submitted from the client’s organization account.
- App Store rejectionGuideline 5.1.1(v) - Legal - Privacy - Data Collection and StorageAdd a Delete account option in the app’s account settings that deletes the whole account record and its personal data. If deletion finishes on your website, link straight to that page. Tell people how long deletion takes and confirm when it’s done. If the app uses Sign in with Apple, revoke the user’s tokens with the Sign in with Apple REST API.
- App Store rejectionGuideline 2.3.3 - Performance - Accurate MetadataReplace the screenshots with real screens of the app doing its main job. Text and image overlays are allowed on top of real screens. Upload at least one screenshot for iPhone with Dynamic Island (medium display), plus an iPad 13-inch screenshot if the app supports iPad, then resubmit.
- App Store rejectionGuideline 2.2 - Performance - Beta TestingRemove beta, demo and trial wording from the app and its metadata, and submit the version you want the public to use. Keep testing in TestFlight: up to 100 internal testers and up to 10,000 external testers, with each build testable for up to 90 days. Never pay testers or offer TestFlight access as a crowdfunding reward.
- App Store rejectionGuideline 2.3.7 - Performance - Accurate MetadataRewrite the name and subtitle as a short, unique description of the app, each 30 characters or fewer. Take competitor names, other companies’ trademarks and prices out of the name, subtitle, keywords, screenshots and previews. Keep the keyword field to 100 bytes or less, without other apps’ or companies’ names, then resubmit.
- App Store rejectionGuideline 2.3.10 - Performance - Accurate MetadataSearch the description, What’s New text, screenshots, previews and in-app screens for other platforms’ names, logos and devices, then remove them. Shared code can be the source: a download badge or an Android phone frame built for the web or Android version shows up in the iOS build. Resubmit once the iOS app only talks about the Apple platforms it supports.
- App Store rejectionGuideline 5.1.1(i) - Legal - Privacy - Data Collection and StoragePut the policy URL in the Privacy Policy URL field in App Store Connect and add a link inside the app, for example in settings or on the sign-up screen. Then check the policy names the data collected, how and why, every third party that gets data such as analytics, ads and SDKs, how long data is kept, and how a user can withdraw consent or ask for deletion. Resubmit and say what changed in the review notes.
- App Store rejectionGuideline 2.3.1 - Performance - Accurate MetadataRewrite the Notes for Review as a specific list: each new feature, where to find it, and how to reach it. In App Review Information, add a valid demo account if features need sign-in, and the specifics of any special configuration. Be ready to provide a demo video or the hardware for anything that needs hard-to-replicate setup. Remove claims from the listing and marketing that the app can’t back up, then resubmit.
- App Store rejectionGuideline 2.3.12 - Performance - Accurate MetadataRewrite What’s New in this Version so it lists each significant change in plain words, for example new features, removed features, new prices or plans, and changed flows. Keep generic wording only for releases that really are just fixes. The field holds up to 4000 characters.
- App Store rejectionApp Store Guideline 1.4.1: medical app accuracy and safetyList every health number the app shows and how it is measured. Remove any reading taken from the phone’s own sensors that Apple names, such as blood pressure or blood oxygen, unless it comes from a connected, validated device. Publish the data and method behind each accuracy claim, add a reminder to check with a doctor, and attach any regulatory clearance in App Review notes.
- App Store rejectionApp Store Guideline 2.5.14: recording without consent or indicationFind every place the app records, logs or captures what the user does, including AI note takers, session replay SDKs and voice features. Ask for explicit consent before the first recording, show a visible or audible indicator for as long as it runs, and consider an easy way to stop. Then reply in App Review with where the consent and indicator appear.
- App Store rejectionGuideline 3.2.2 - Business - UnacceptableFind the sub-point the rejection quotes and remove that practice from the app. The ones a code change fixes: forcing a rating, review or download before people can use the app, ads built to inflate impressions or clicks, collecting donations in the app without approval, and blocking users by location or carrier. Then reply in App Store Connect with what changed.
Google Play launches
Guide: App Store and Google Play rejections
- Google Play production accessNew Google Play accounts need 12 testers for 14 days before going liveFinish your app setup, start a closed test and get at least 12 testers opted in. Keep them opted in for 14 days in a row, collect their feedback, then click Apply for production on the Play Console dashboard and answer the questions about your test. Google says its review usually takes seven days or less.
- Google Play policy issueGoogle Play rejected my app over the Data safety form. What do I fix?Google reviews the Data safety form with the app, and you alone are responsible for it being complete and accurate. List every kind of user data the app and its SDKs (third-party code libraries built into the app) collect or share, and answer the security questions truthfully. Link a working privacy policy, and if users can create an account, offer account deletion in the app and on the web. Then update the form and resubmit.
- Google Play policy issueGoogle Play rejected my app over Photo and Video Permissions. What do I fix?If your app targets Android 13 or later, it may only request READ_MEDIA_IMAGES and READ_MEDIA_VIDEO when a system picker such as the Android photo picker isn’t enough for its core functionality, as with a gallery app. If people only choose a photo or video to attach or upload, switch to the photo picker and remove both permissions. If a library adds them, remove them in your manifest with tools:node="remove". If the app really needs broad access, submit the declaration in Play Console.
- Google Play policy issueGoogle Play rejected my app over foreground service permissions. What do I fix?If your app targets Android 14 or later, declare every foreground service type it uses on the App content page in Play Console, with a description, the effect of delays or interruptions on users, the use case, and a link to a video of a user triggering the feature. In the manifest, give each service its type and the matching permission. If a task fits no type, Android recommends WorkManager or a user-initiated data transfer job instead.
- Google Play upload errorGoogle Play: “Your Android App Bundle is signed with the wrong key”Sign the bundle with the upload key Play Console expects. Compare the upload key certificate on the Play app signing page in Play Console with your keystores to find the right one. If it’s lost, create a new upload key, export its certificate, and request an upload key reset in Play Console. Google keeps the app signing key, so users keep getting updates.
- Google Play requirementGoogle Play: app must target API level 35 or 36Set targetSdk (and compileSdk) to 36 in the app’s Gradle file, build, and test the screens that Android 15 and 16 change, then upload the new version. An app that doesn’t ship an update needs at least API level 35 to stay available to new users on newer Android versions. In Flutter, Expo or React Native, change it where the project sets the Android SDK versions, then check the built bundle.
- Google Play policy issueGoogle Play Console warning: your app uses an unsupported Google Play Billing Library versionFind the billing dependency in your Android build (com.android.billingclient:billing, or the version pulled in by a plugin such as in_app_purchase or RevenueCat), and move it to version 8 or later. Version 9 is the newest in Google’s table, and its migration guide is linked below. Build, upload, and check that the warning clears. If you already upgraded and the warning stays, check that the merged AndroidManifest.xml contains com.google.android.play.billingclient.version.
- Google Play rejectionGoogle Play: QUERY_ALL_PACKAGES permission not allowed for your appOpen the merged manifest in Android Studio to see which library adds QUERY_ALL_PACKAGES. If your app only needs to open or check a few known apps, list them in a <queries> element and remove the permission with tools:node="remove". Keep it only if the app is a device search, antivirus, file manager or browser style app, and then fill in the Permissions Declaration Form.
- Google Play rejectionGoogle Play: USE_EXACT_ALARM permission not allowed for your appRemove USE_EXACT_ALARM from the manifest unless the app is an alarm, timer or calendar app. If a feature still needs exact timing, request SCHEDULE_EXACT_ALARM, check canScheduleExactAlarms() before you schedule, and send the user to the settings screen when it is off. Most reminders work with inexact alarms. Then upload a new release.
- Google Play rejectionGoogle Play: invalid or missing privacy policyPut the policy on a public web page with a valid HTTPS certificate that loads without a login, from any country. Make it about this app and what data it collects and shares. Add the URL under App content > Privacy policy in Play Console, add the same link inside the app, then send the update for review.
- Google Play rejectionGoogle Play: account deletion requirement not metAdd a Delete account option in the app that deletes the account and its data, or starts that request. Publish a web page where users can ask for deletion without installing the app. Add that URL in the Data deletion questions of the Data safety form, then send the update for review.
- Google Play rejectionGoogle Play: app access rejected, sign-in details missing or not workingCreate a review account that works without your help, and add it on the Sign-in details page in Play Console. Explain any one-time code, 2FA, invite code or location check in Any other instructions. Then send the changes for review from the Publishing overview page.
- Google Play rejectionGoogle Play: rejected for limited or broken functionalityPlay’s policy page is titled Functionality, Content, and User Experience, with limited and broken functionality sections. Read which one the email names. For limited functionality, add features that work inside the app, such as accounts, saved data, offline screens or push notifications. For broken functionality, install the release build on a real device, fix crashes and blank screens, and check the back end is live during review. Then resubmit.
- Google Play rejectionGoogle Play: prominent disclosure and consent requirement not metAdd a screen in the normal app flow, before the system permission dialog, that names the data, the feature it enables and any sharing. Give it a clear accept button and treat back or dismiss as no. Then send a new build for review.
- Google Play requirementGoogle Play: app doesn’t support 16 KB memory page sizesFind the .so files that fail in the Play Console warning or the APK Analyzer. Upgrade the Android Gradle Plugin to 8.5.1 or higher, build your own native code with NDK r28 or newer, and update SDKs that ship native libraries. Check an APK with zipalign and the bundle with bundletool before you upload.
- Google Play rejectionGoogle Play: app doesn’t meet the User Generated Content policyAdd terms of use that users accept before posting, a report option on posts and profiles, a way to block users, and tools to remove content and ban accounts. Then send a new release and reply to the rejection notice.
- Google Play rejectionGoogle Play: app violates the Impersonation policyFind every place another brand shows up: the title, icon, short description, screenshots, developer name and in-app screens. Replace it with your own, state that the app is independent, or get written permission from the owner. Then send a new release and reply to the policy notice.
- Google Play rejectionGoogle Play: background location permission declaration rejectedDecide whether the app truly needs location when it is closed. If not, remove ACCESS_BACKGROUND_LOCATION from every track and use foreground location. If it does, declare one core feature in the Location permissions form, add a short video of it, show a prominent disclosure before the system prompt, and resubmit.
- Google Play rejectionGoogle Play: SMS or Call Log permissions not allowedIf the app only reads one-time codes, remove READ_SMS, RECEIVE_SMS and READ_CALL_LOG from the manifest, including those added by SDKs, and use the SMS Retriever API or SMS User Consent API. If the app truly is the default SMS or Phone handler, submit the Permissions Declaration Form in Play Console, then upload the new build to every track.
- Google Play policy issueGoogle Play: Contacts Permission policy (READ_CONTACTS)Find every feature that reads contacts. If the user only picks one or a few contacts, such as to invite, share or pay someone, switch to the Contact Picker with Intent.ACTION_PICK_CONTACTS and remove READ_CONTACTS from the manifest. Keep the permission only for core features such as contact management or backup, and explain in the Play Console declaration why the picker is not enough.
- Google Play rejectionGoogle Play: AI-Generated Content policyAdd a report or flag control on every AI answer, image or video, so users can report it without leaving the app, and send those reports somewhere you review. Filter prompts and output for restricted content, such as content that facilitates the exploitation or abuse of children, and deceptive content. Then resubmit with a note on where the reporting control is.
- Google Play rejectionGoogle Play: Financial Services policy (personal loans)Complete the Financial features declaration in Play Console, set the app category to Finance, put the loan terms in the store listing (minimum and maximum repayment period, maximum APR and a representative total cost example), and remove access to contacts and photos. Loans must not require full repayment in 60 days or less, and US loans must stay under a 36% APR.
- Google Play rejectionGoogle Play: Payments policy (Play billing)Unless one of the policy’s exceptions applies, move every purchase of digital content, app features or subscriptions in the Android app to Google Play’s billing system, and remove the buttons, links, webviews and sign-up steps that send users to a web checkout. Keep physical goods and services on your own payment processor. Make the in-app prices match the Play billing screen, then send an update for review.
- Google Play rejectionGoogle Play: Subscriptions policy (paywall disclosure)Rebuild the paywall so the amount the user will actually be charged, how often, that it renews automatically, and when a free trial turns into a paid plan are all clear on the offer screen, in the user’s language and currency. Add a visible dismiss button when the app works without a subscription, and a way to cancel in account settings. Then submit an update.
- Google Play rejectionGoogle Play: Ads policy (disruptive and deceptive ads)Find where each ad network shows a full screen ad and move it to natural breaks, such as after a score screen, not at app start, on a button tap or at the start of a level. Make every interstitial closeable within 15 seconds, remove ads that look like app buttons or system notifications, and check the ads match the app’s content rating. Then submit an update.
- Google Play rejectionGoogle Play: Deceptive Behavior policyCompare the store listing with the app a new user actually gets: remove any feature, screenshot or claim the app can’t back up, drop “official” unless you have the rights, and check the category and content rating. Make sure the app behaves the same for reviewers as for users, with no features switched on remotely later. Then submit an update.
- Google Play rejectionGoogle Play: Device and Network Abuse policyRemove any in-app updater and any code path, in your code or an SDK, that downloads dex, JAR or .so files; ship those changes through a Play update instead. Check webviews with a JavaScript interface only load your own HTTPS pages, and update or replace any SDK that does either. Then submit an update.
- Google Play rejectionGoogle Play: Metadata policy (store listing)Shorten the title to 30 characters or less. Remove emojis and repeated special characters from the title, icon and developer name, and avoid ALL CAPS unless it is part of the brand name. Remove “#1”, “Popular”, award, discount or “free for limited time only” text from the title, icon and developer name. Cut keyword lists and unattributed testimonials from the description, keep it short and accurate, then resubmit the listing.
- Google Play policyGoogle Play: Content Ratings policy (IARC questionnaire)Open the content rating questionnaire in Play Console and answer it again for the app as it is now, including any content or features added since the last rating. Check that the ads shown are not significantly more mature than the app. Submit the new questionnaire, then the update. If you disagree with the rating, appeal through the link in the certificate email.
- Google Play rejectionGoogle Play: User Ratings, Reviews, and Installs policyRemove any reward, discount or other incentive tied to leaving a rating or review, and any pop-up that forces or tricks users into rating. Stop paid or fake reviews and any automated services that inflate installs or ratings. Ask for ratings in a clear, non-deceptive way, and keep review replies on the user’s issue without asking for a higher rating. Then submit an update.
- Google Play rejectionGoogle Play: AccessibilityService API use not allowed or not declaredDecide whether the app is an accessibility tool for people with disabilities. If it is, set isAccessibilityTool and complete the declaration with a video. If it isn’t, show a prominent disclosure and get consent before turning the service on, complete the declaration, and keep any automation narrow and rule-based. If the feature isn’t core, remove the service.
- Google Play rejectionGoogle Play: Health Connect and health permissions not approvedList every health permission in the manifest and the feature that reads or writes it. Remove the ones no feature needs, including history and background reads. In Play Console, request only the remaining data types, pick the use case that matches the app, and make the privacy policy name the health data, its use, retention and deletion. Then read the feedback in Play Console and resubmit.
- Google Play rejectionGoogle Play: REQUEST_INSTALL_PACKAGES permission rejectedCheck the merged manifest for the permission. If the app’s main purpose is not to send or receive app packages and let users install them, remove it, including when a library added it, and use Google Play for app updates. If it does qualify, fill in the Permissions Declaration Form in Play Console and describe the feature in the store listing.
- Google Play rejectionGoogle Play: Families policy and target audienceOpen Target audience and content in Play Console and check the age groups match who the app is really for. If children are included, show ads only from Families self-certified ads SDKs, without interest-based ads, use a neutral age screen for mixed audiences, and stop sending identifiers such as the advertising ID. If the app is not for children, remove listing art that suggests it is.
Moving and upgrading apps
- Lovable Cloud moveMove your Lovable app off Lovable Cloud to your own SupabaseExport the database from Lovable (it includes your users and their password hashes) and restore it into your own Supabase project. Then move storage files, edge functions, secrets and sign-in settings by hand, and switch the app over. Anyone signed in at the switch signs in once more. Keep the built-in backend until the new project works, because removing it can’t be undone.
- Leaving BoltMove your Bolt app off Bolt without breaking itDownload a copy, then connect a GitHub repository so the code lives in an account you own. Move the database by claiming your Bolt Database in Supabase, which needs a Pro or Teams plan and a Supabase org owner, rather than connecting a fresh Supabase database, which can lose data. Deploy the repository to your new host, add the environment variables there, and move the custom domain. Then test sign-in, emails, payments and webhooks before you switch users over.
- Lovable and Bolt to iOSTurn your Lovable or Bolt web app into an App Store appWrap the published web app in a native shell with Capacitor, build it in Xcode on a Mac, and submit it from your own Apple developer account. To pass review it has to be more than your website in a frame: plan native features such as push notifications, add Sign in with Apple if Google sign-in is the only option, and let users delete their account in the app.
- Framer and Webflow to an appAdd a real app to your Framer or Webflow siteKeep the marketing site where it is and build the app as a separate web app on a subdomain such as app.yourdomain.com, with its own sign-in, database and payments, linked from the site’s menu and buttons. Framer doesn’t offer an HTML export, and Webflow’s export leaves out the CMS, forms and user accounts, so moving the site rarely pays.
Not listed? Send the error
Email the full error text and what you were doing when it appeared to simon@app369.com. Simon replies by email, usually within one business day.
A different error?
Describe it and get a written quote
Send the error and what you were doing. You get a written scope, a fixed price and a timeline; Simon replies by email, usually within one business day.