App Store
Rejected under App Store Guideline 4.3 (Spam) for an AI chat app: is it fixable, and what to change
A 4.3 rejection means the reviewer saw your app as one more copy of something the App Store already has plenty of. For an AI chat app it is usually fixable, but not by small tweaks: the app has to do a specific job that a general chat screen does not, and the listing has to show it. Guideline text below is quoted from the version last updated on 2026-10-16.
What you are seeing
The rejection arrives in App Store Connect, on the App Review page of your app. It cites Guideline 4.3 (Spam, in the Design section of the guidelines). Read the full message in App Store Connect rather than the notification email, and note which part it points to, because 4.3 has two:
- 4.3(a): several Bundle IDs of the same app. This applies if you or your account submitted more than one app from the same project - for example the same chat app with a different name, colour or system prompt for each audience.
- 4.3(b): the app is "indistinguishable from what's already widely available". This is the usual one for a general-purpose AI chat app.
4.3 often arrives together with, or instead of, related rejections. Check the message for these too, since each needs its own change:
- 4.2 Minimum Functionality: the app is close to a repackaged website.
- 4.1 Copycats: the app copies another app, or uses another developer's brand or product name in its icon or name.
- 4.2.6: the app was created from a commercial template or app generation service and submitted by the service rather than by the provider of the app's content.
How to check it yourself
Start with what Apple actually wrote. The relevant part of 4.3(b):
“Don't submit apps that are indistinguishable from what's already widely available. Opportunistically creating variants of existing app categories or popular apps degrades App Store discovery, reduces overall app quality, and harms both users and developers.”
The guideline names some categories it treats as already saturated (dating, flashlight, sound effects, wallpaper, simple timers, fortune telling). AI chat is not on that list, so a 4.3 rejection for an AI app is the reviewer's judgement about your app, not a blanket ban. Apple's App Review page also lists "repeated submission of similar apps" among common rejections and asks you to combine such apps into one. Now check your own app and listing the way a reviewer sees them:
- Write one sentence: who is the app for, and what job does it do for them? If the honest answer is "anyone, for anything you would ask an AI", that is what 4.3(b) describes.
- On an iPhone, open the App Store app and search for the words a user would type to find your app. Open the first results and compare them with your screenshots. If you cannot name something your app does that they do not, the reviewer probably could not either.
- Open your app, sign in with a fresh account, and look only at the first screen. Is it an empty text box with a send button? Then the first thing a reviewer sees is the same thing every general chat app shows.
- In App Store Connect, go to Apps and list every app in your account. Are any of them the same project with a different name, icon or prompt? That is 4.3(a), and Apple asks for one app with the variations inside it.
- Check your listing for other companies' names. In App Store Connect, open App Information (Name, Subtitle) and the version page (Description, Keywords, Promotional Text, screenshots). Apple's field reference says names of other apps or companies are not allowed in Keywords, and 4.1(c) forbids another developer's brand or product name in your app's icon or name without their approval.
- Search your project for the same names, so the app itself does not present as someone else's product. The command below runs in the project folder on your computer.
- If your backend is Supabase, list the tables your app has, with the query below in the SQL Editor of your project's dashboard. If the list is little more than profiles, conversations and messages, the app holds no data of its own beyond the chat - which is a useful, if uncomfortable, picture of how specific it is.
grep -rniE "chatgpt|openai|gpt-?[0-9]|gemini|claude" \
--include="*.ts" --include="*.tsx" --include="*.html" --include="*.json" \
--exclude-dir=node_modules --exclude-dir=dist .select table_name
from information_schema.tables
where table_schema = 'public'
and table_type = 'BASE TABLE'
order by table_name;Why it happens
AI builders make it quick to put a chat screen in front of a model provider's API. The result works, but from a reviewer's seat it looks like every other app built the same way: a text box, a send button and answers from a well-known model. The work that went into your prompt, your settings or your backend is invisible unless the app's screens show what it does differently.
A few things make this worse:
- The difference lives only in a system prompt. Two apps with different prompts can still look and behave the same, and a reviewer judges the app by what its screens show, not by its prompt.
- Several variants from one codebase - a "coach", a "tutor", a "writer" - submitted as separate apps. That is the pattern 4.3(a) is written against.
- The listing leans on a model's brand. Apple's copycat guidance says not to include references to popular app names or trademarked terms in any metadata field, and 5.2.1 forbids misleading or copycat names and metadata.
- Screenshots show a generic chat transcript instead of the specific task, so the product page itself reads as "another AI chat".
How to fix it
The aim is that a reviewer understands quickly, from the first screen and the listing, what your app does that a general chat app does not. In rough order:
1. Narrow the product to one job
- Pick one user and one task - for example drafting replies to customer reviews for a small shop, or practising a specific exam - and build the app around that task rather than around the chat box.
- Give the task its own structure: inputs that fit it (a form, a photo, a document, a choice of options), outputs saved in a form the user can come back to, history organised by what the user worked on, not by message.
- Store the data that makes the app specific - the user's projects, documents, results or progress - so the app is useful between conversations, not only during one.
- Make the first screen show the task, not an empty prompt. A reviewer who opens the app should see what it is for before typing anything.
2. Merge variants into one app
If you have several apps from the same project, keep one and offer the variants inside it - as modes, templates or topics the user chooses. 4.3(a) suggests exactly this, and mentions in-app purchase as a way to offer the variations. Stop submitting the duplicates: cancel or withdraw any pending submissions for them. If some are already live, plan how their users and purchases move to the app that stays before you take them off sale.
3. Clean up the listing
- Name and subtitle describe the job, in your own words, with no model or company names (Name and Subtitle are limited to 30 characters each).
- Keywords describe the task, not other products - Apple's field reference does not allow names of other apps or companies there.
- Screenshots show the actual task in use: the input, the saved result, the history. Apple's copycat guidance asks for screenshots that show the actual app in use.
- The description says who the app is for and what it does, and can mention which AI service powers it as a plain fact, without presenting the app as that service.
4. Clear the AI-specific rules at the same time
These are not 4.3, but a reviewer looking at an AI app checks them, and fixing them now avoids the next rejection:
- 5.1.2(i): "You must clearly disclose where personal data will be shared with third parties, including with third-party AI, and obtain explicit permission before doing so." Ask before the first request, as in the example below.
- 1.2 User-Generated Content covers "AI content creation apps": they need filtering of objectionable material, a way to report content with timely responses, the ability to block abusive users, and published contact information.
- 4.7 applies if your app offers chatbots that are not embedded in the binary, for example bots that users create and share: you are responsible for them, and they must follow the privacy rules and have filtering, reporting and blocking.
- 1.4.1: if the app gives health or medical answers, expect greater scrutiny.
import { Preferences } from "@capacitor/preferences";
const CONSENT_KEY = "ai-consent-v1";
export async function hasAiConsent(): Promise<boolean> {
const { value } = await Preferences.get({ key: CONSENT_KEY });
return value === "granted";
}
export async function grantAiConsent(): Promise<void> {
await Preferences.set({ key: CONSENT_KEY, value: "granted" });
}
// Call before every request to the AI provider. If it returns false,
// show a screen that names the provider and what is sent (text, photos),
// with "Allow" and "Not now" buttons; call grantAiConsent() on "Allow".
export async function canSendToAi(): Promise<boolean> {
return hasAiConsent();
}5. Tell App Review what changed
Apple's guidance on a complete review asks you to describe the app's concept, purpose and audience in App Review Information, and to list the external services the app uses for its core functionality, AI services included. Put that in the Notes field when you resubmit. A short, factual note works better than an argument:
Who it is for: [one specific user group].
What it does: [the one job], which a general chat app does not do because [specific reason].
Where to see it: after sign-in, tap [screen] to [action]; saved results are under [tab].
Changes since the last review: [merged apps X and Y into this one / removed brand names from the listing / new first screen].
AI service used: [provider], for [what]. Users are asked for permission before any data is sent.
Demo account: entered in the Sign-in required fields (add any extra accounts here).6. Reply, resubmit or appeal
- If you think the reviewer misunderstood the app, reply first: Apple's contact guidance says to use Reply to App Review in App Store Connect, be specific, and add information that helps App Review understand the app's concept.
- If you changed the app, resubmit with the Notes above. Apple says not to file an appeal after you have addressed the issues and resubmitted.
- If the reply did not resolve it and you believe the app follows the guidelines, appeal to the App Review Board: cite the guideline and explain point by point how the app meets it. Apple asks for one appeal per rejected app.
- If you are unsure how 4.3 applies before you rebuild, you can request a 30-minute App Review appointment. Apple notes that it is informational and does not guarantee the outcome of a review.
When to get help
- You have been rejected for 4.3 more than once and do not know what to change next.
- You have several similar apps in your account and need to merge them without losing users or purchases.
- Narrowing the product means new screens, new tables and new rules in the backend, and your AI tool keeps breaking the parts that already work.
- There is a deadline - a launch, a client, a demo - and another rejection would miss it.
Sources
Primary documentation this guide was checked against. Platforms change - if something here no longer matches, the source wins.
- App Store Review Guidelines (last updated 2026-10-16)Apple Developer
- App Review - common rejectionsApple Developer
- 4.1 - Preventing copycat and impersonation rejectionsApple Developer
- Provide information for a complete reviewApple Developer
- Contact App Review for help with a submissionApple Developer
- Appeal to the App Review BoardApple Developer
- Request a one-on-one App Review appointmentApple Developer
- Platform version information (Keywords, Notes)App Store Connect Help
- App information (Name, Subtitle)App Store Connect Help
- Database overview (SQL Editor)Supabase docs
- information_schema.tablesPostgreSQL docs
- Preferences pluginCapacitor docs
Frequently asked questions
- Is a 4.3 rejection a dead end for an AI chat app?
- Not necessarily. AI chat is not one of the categories Guideline 4.3(b) names as saturated, so the rejection is a judgement about your app. It is fixable when the app is changed to do a specific job a general chat app does not, and the listing and Review Notes show that. Resubmitting the same app with a new name is not a fix.
- Does Apple reject every app that uses ChatGPT, Claude or Gemini?
- No. The guidelines do not forbid using a third-party AI model. They require that the app is not indistinguishable from what is already widely available (4.3), that it does not use another developer's brand or product name in its icon or name without approval (4.1(c)), and that users are told and asked before personal data goes to third-party AI (5.1.2(i)).
- Can I mention the AI model in my App Store listing?
- Saying which service powers a feature, as a plain fact in the description, is different from building the listing around another company's brand. Apple's field reference does not allow names of other apps or companies in Keywords, and 4.1(c) covers the icon and the app name.
- I have three similar AI apps. Should I resubmit them one by one?
- No. 4.3(a) and Apple's App Review page ask you to combine similar apps into one. Keep one app, offer the variants inside it, and resubmit that.
- Should I appeal or resubmit?
- If the reviewer was right that the app looks like every other chat app, change it and resubmit with Notes that explain what changed. If you believe the reviewer misunderstood the app, reply in App Store Connect first, and appeal to the App Review Board only if that does not resolve it. Apple asks for one appeal per rejected app.
Related guides
Fixed-price audit
Want someone to check the whole app, not just this?
A review of data security, production readiness and App Store readiness for apps built with AI tools. 900 USD, a ranked list of fixes within 5 business days.