The boring part of ChatGPT Ads is the one that can cost you the most days. Not because the rules are strange, but because there are three separate reviews and most people assume there is only one. This lesson is the map, so you are never stuck without knowing why.
Three reviews, not one
| Review | What it looks at | What happens meanwhile |
|---|---|---|
| Account brand | Your identity as an advertiser: name, icon and website (official documentation) | Until it is approved you can have everything built and serve nothing |
| Account integrity | The account as a whole; separate from the one above | The documentation warns that an active account with approved branding can still have another review outstanding |
| Ad | Each creative: in review, approved or rejected | A rejected ad breaks a policy: you edit it for it to be re-reviewed |
📌 We measured this ourselves: in our account the 8 ads were approved, with no low-bid warning, and we still had days of zero impressions. The reason was the brand review, approved on 16/09/2026 at around 11:30. The campaign served the next day. If you see approved ads and no impressions, check the account reviews before touching the bid.
What reopens a review
The official documentation is explicit on one point: brand changes can trigger a review, and the brand review looks at name, icon and website. In other words, changing your trading name, your logo or the domain you send traffic to is not a cosmetic tweak — it is rejoining the queue.
On creatives, the only thing the documentation states clearly is that a rejected ad is edited so it gets re-reviewed. We found no official source detailing which edits to an already approved ad reopen its review and which do not, so the practical rule is the cautious one: assume that touching the title, text, image or destination means another review, and do not leave a creative change until the eve of a seasonal campaign.
Restricted sectors and content
Ads do not appear near personal health, mental health or politics, and political advertising is banned outright. Healthcare and financial services carry strict eligibility criteria and, at best, are assessed case by case — mostly in the US (Headcore manual, August 2026). The basic rules of the advertising policy: nothing deceptive, professional language, no discrimination, no imitating the ChatGPT interface, and a landing page that matches the advertiser and the offer.
Before building anything, the question is not "how do I word this so it passes?" but "is my category allowed in my country today?".
What the card promises must be behind the click
This one caught us out. In our own campaign we advertised two resources as a "free ebook" when they were in fact for subscribers. There was no bad faith: they had been open at another time. It made no difference — the card promised something the page did not deliver, so we pulled those ads.
⚠️ Trap: "free" with a sign-up in the middle is still free if you say so. What does not hold up is a click that leads to a different wall from the one you advertised. The check takes ten seconds: open your ad's link in a private window and ask whether a stranger gets exactly what was promised.
Other people's trademarks: in the description, never in the name
Two rules pointing the same way:
- "ChatGPT" and "GPT" belong to OpenAI. Its brand guidelines do not allow them inside the name of your product, app or company — the rule is also reported by the trade press, and OpenAI enforces it by asking offenders to drop the term.
- Google says exactly the same about its own: "Don't incorporate Google Brand Features into your own product names, service names, trademarks, logos, or company names."
The way out is not to stop naming them: it is to use them as description. Our own product is not called "Ninja GPT Ads"; it is "Ninja Scripts Suite · for ChatGPT Ads" (and its twin, "· for Google Ads"). It says exactly what it does, and the other brand stays where it is allowed to be: explaining which platform it serves, not forming part of the name.
Checklist before launching
- Is my category allowed in my country? Does it brush against health, mental health or politics?
- Are the advertiser's name and icon final? (If you plan to change them, do it first: it triggers a review.)
- Does the landing page match the advertiser, respond, and let OpenAI's crawlers through?
- Is what the card promises behind the click, with no conditions I am not announcing?
- Am I using someone else's trademark in my product's name?
- Is measurement in place before the first impression?
- Have I left days of slack for the reviews before the date I want to be live?
What to take away
- There are three reviews — brand, integrity and ad — and the brand one can keep you at zero impressions with everything approved.
- Changing name, icon or website triggers a review: settle them first.
- Ads stay away from health and politics; healthcare and finance are case by case and mostly in the US.
- Card and page must say the same thing: it cost us two pulled ads.
- Other brands go in the description, never in your product's name.