A Guideline 1.1 rejection for an AI chat app is rarely about one bad message a reviewer saw. It is about the app being able to produce objectionable content with nothing in place to catch it. The good news is that the fix is mostly a feature you build: moderation, plus a way to report and block. Here is what Apple expects and how to add it.
Short answer
Guideline 1.1 prohibits offensive or objectionable content, and for an interactive AI app Apple expects the same safeguards Guideline 1.2 requires for user-generated content: a method to filter objectionable material, a mechanism to report a response with a timely follow-up, the ability to block, and published contact information. You are responsible for your chatbot's output under Guideline 4.7, and you must disclose that users are talking to a bot. The fix is moderation paired with report and block, which is a feature set you can implement.
What you should know
- 1.1 bans objectionable content: offensive, insensitive, or upsetting material is not allowed.
- AI output is treated like UGC: Apple applies the 1.2 safeguards to content your app generates.
- You are responsible for the bot: Guideline 4.7 puts the chatbot's output on you.
- Four safeguards are expected: filter, report, block, and published contact information.
- Disclose the AI: make clear users are talking to a bot, not a human.
What does Guideline 1.1 require, and how does 1.2 apply to AI?
Guideline 1.1 states that apps should not include content that is offensive, insensitive, upsetting, intended to disgust, or in exceptionally poor taste. Guideline 1.2 then sets the concrete safeguards for apps where content appears that you did not write yourself: a method for filtering objectionable material, a mechanism to report offensive content with timely responses, the ability to block abusive users, and published contact information. Although 1.2 is written for user-generated content, an AI app generates content on demand, and Apple holds you responsible for it under Guideline 4.7, so reviewers expect the same protections. In practice, an AI chat or image app is reviewed as if its output were user-generated. It helps to read the two together: 1.1 says objectionable content is not allowed, and 1.2 says how an app that surfaces content it did not author is expected to keep that content out, which is exactly the position an AI app is in.
What safeguards does an AI app need?
The four from 1.2, adapted to AI output, plus disclosure. The table maps each one to what it means when the content comes from a model rather than another user.
| Safeguard | What it means for an AI app |
|---|---|
| Filter objectionable material | Moderate the model's output before showing it to the user |
| Report mechanism | Let users report a specific AI response, and respond to reports promptly |
| Block | Let users block content or stop an abusive session |
| Published contact info | A reachable way for users to raise concerns |
| AI disclosure | Make clear the user is talking to a bot, not a person |
Missing any of the first four is the usual reason a 1.1 or 1.2 rejection sticks, because the reviewer can produce objectionable output and finds no way to filter, report, or escalate it.
How do you implement Block and Report for AI text and images?
Build it around the model's output. First, run generated text and images through a moderation step before display, using a moderation endpoint or your own classifier, and refuse or replace content that fails. Second, add a report control on each AI response, so a user can flag it, and make sure those reports reach you and get a timely answer. Third, add a block or stop control that lets the user end an abusive exchange, and on your side suppress patterns of misuse. Fourth, publish contact information in the app and on your listing. Finally, label the assistant clearly as AI. Together these turn an open generator into a moderated one, which is what the guideline is really asking for. None of them require a perfect filter, which no system can promise; they require that a reasonable safeguard exists at each stage, so a reviewer who produces something offensive can see it caught, flag it, or stop it rather than hitting a dead end.
Why do AI chat apps get hit by Guideline 1.1 specifically?
Because a model can generate objectionable content on request, and a reviewer will try. Where a normal app only shows what you built, an AI app will produce whatever the prompt steers it toward, so a reviewer can often coax it into something offensive in a few messages. If that output appears with no filter and no way to report it, the app fails 1.1, since the content is both present and unguarded. The rejection is not a judgment that your app is bad; it is that an unmoderated generator is, by default, an app that can show objectionable content.
What to watch out for
The first trap is treating moderation as optional polish; for an AI app it is the core requirement, not an add-on. The second is adding a report button that goes nowhere, since the guideline asks for timely responses, so reports have to reach a person or process. The third is anonymous or random-chat framing, which draws extra scrutiny and can be removed without notice, so keep the experience clearly moderated and disclosed. Guideline 1.1 is a content and safety rule rather than a binary security finding, so it sits apart from what a pre-submission scan checks; a scan such as PTKD.com (https://ptkd.com) covers the binary layer against OWASP MASVS, while the 1.1 fix is the moderation and reporting you add to the app.
What to take away
- A Guideline 1.1 rejection for an AI app means it can produce objectionable content with no safeguards in place.
- Apply the four Guideline 1.2 protections to your model's output: filter, report with timely responses, block, and published contact info.
- Moderate generated text and images before display, add a report control on each response, and disclose that the assistant is AI.
- The fix is a moderation and reporting feature you build, separate from the binary checks a pre-submission scan such as PTKD.com performs.
