Who can advertise in ChatGPT: policies and eligible categories
As of September 2026, OpenAI allows advertising in ChatGPT mainly for consumer categories: home and lifestyle goods, local services, travel and experiences, digital products and training. Financial, health and legal services are reviewed manually, only for approved advertisers and in practice only in the USA — outside the USA these three categories are prohibited in principle, which includes Bulgaria. Completely prohibited everywhere are gambling, alcohol and tobacco, dating and adult content, political content, recreational substances, counterfeits, individual job and property listings, and any misleading or exaggerated claims. The policies apply to the whole path — text, visual and landing page must be consistent — and every advertiser goes through a manual check at registration. The list was updated on 31 August 2026 and OpenAI states that it will change. Below is a breakdown by category, aimed at the typical Bulgarian business.
I am Hristiyan Panov — a Performance Marketing consultant with over 5 years of experience and over €2,000,000 in managed ad budget for more than 90 companies on Google and Meta. Below is a breakdown of the ChatGPT advertising policies based on OpenAI’s official text — what is open, what is closed and what needs to be checked before a budget is planned. This is not legal advice; if in doubt, check the current text of the policies.
Which categories are currently eligible?
The official text says that in the initial period ads are limited mainly to consumer verticals: lifestyle and home goods, local services, travel and experiences, digital products and education. For a Bulgarian business this covers:
- online shops for furniture, electronics, tools, fashion, cosmetics without health claims, gifts and pet products without therapeutic claims;
- hotels, travel agencies, attractions and events;
- local services such as repairs, cleaning, car repair shops and salons;
- online courses, software and apps with no financial or medical character.
The category is determined by what is actually sold, not by the description of the business. The same shop can have an eligible and an ineligible part of its range: cosmetics without health claims sit in the consumer vertical, while the same brand with a treatment claim does not. That is why the review starts from the product pages and the text in the descriptions, not from the home page. When the range mixes the two, only the clean category is planned for advertising. OpenAI states that the list may expand.
Which categories are prohibited or restricted outside the USA?
Three categories are key for Bulgaria. Health services: ads for regulated medical products and services and any claims about prevention, diagnosis or treatment are restricted, and outside the USA prohibited in principle; in the USA, dietary supplements, dental services, medical devices and cosmetic procedures are approved manually. Financial services: loans, cards, insurance, investments and payment services are restricted and outside the USA prohibited in principle; budgeting apps and financial education without offering a service are allowed as a rule. Legal services: law firm and legal services outside the USA are prohibited, while legal education without offering a service is permitted.
The practical conclusion: a pharmacy, a supplement shop, an insurance broker, a credit intermediary and a law firm in Bulgaria do not have access to the channel at present. General healthy-lifestyle products without medical claims — fitness equipment, wearables — may be admitted; products with unregulated “wellness” claims are assessed under a separate policy for such claims.
Which categories are prohibited everywhere?
Gambling, including online betting, lotteries and games with cash prizes; casinos are allowed only when the ad is for a hotel or an experience with no focus on gambling. Alcohol above 0.5% and tobacco, including vapes and nicotine; incidental mentions are allowed only when alcohol is not the focus. Adult content, including dating apps; underwear and swimwear in a standard retail context are allowed. Political content for or against candidates, parties, elections and public policies. Recreational substances, including cannabis and THC; hemp products without an intoxicating effect may be admitted. Counterfeits and products imitating other brands. Individual job or property listings — listing platforms are allowed if the ad does not point to a specific listing.
What baseline standards apply to every ad?
Four standards apply regardless of category. No misleading or exaggerated claims about capabilities, prices, results, affiliation or comparisons with competitors — this includes “guaranteed” results. No offensive language or shocking visuals, even in product names. No discrimination or harassment. No imitation of the ChatGPT interface or voice — the ad must be clearly distinguishable from the product. OpenAI reserves the right to refuse to show an ad or an address for any reason.
The standards apply to the whole path: an approved ad cannot lead to a page with inadmissible content. OpenAI’s example is an ad for food delivery that leads to alcohol delivery. For a Bulgarian shop with a wide range of products, this means the landing page must be a category without prohibited products, not the home page.
What needs to be in order with trademarks and visuals?
The advertiser must own or have permission to use the brands and materials they use. Ads for products that copy another brand, logo or distinctive features are prohibited as counterfeits. For shops with licensed products — music bands, sports teams, film characters — this means the campaign must be based on its own or explicitly licensed designs.
For visuals, the review goes through the following:
- other brands’ logos and packaging that end up in the frame or in a collage are removed;
- the text on the visual repeats the offer in the ad, without a new claim that is not on the page;
- the visual stays simple and relevant, so that approval does not depend on interpretation.
Separately, the landing page must be accessible to the OAI-AdsBot and OAI-SearchBot crawlers — a blocked crawler means a failed review. Permission in robots.txt is not enough: a firewall, CDN or hosting rule can return a refusal to the crawler itself. Access is confirmed by the server logs for these two agents, not by how the page looks in a browser.
Where are ads not shown, even for an eligible category?
The placement policies prohibit ads next to sensitive user contexts — personal health, mental health, emotionally vulnerable conversations — and next to brand-unsafe contexts: violence, hate, scams, politics, weapons, regulated goods and others. This applies even to a fully eligible advertiser: a tool shop will not be shown next to a conversation about self-harm. For the advertiser this is brand protection, but also a limit on reach — some conversations simply carry no ads.
In planning, the consequences are specific:
- the available inventory is smaller than the volume of the topic — the same question may be eligible in one conversation and blocked in another, depending on the direction it has taken;
- a blocked context is not opened with a higher bid, because that is a platform decision, not an outcome of the auction;
- impressions accumulate in conversations with commercial intent, so reach expectations rest on the eligible part of the topic.
That is why reach is assessed only after it is clear which contexts around the product fall into the sensitive lists.
How do you check eligibility before planning a budget?
Three questions. First, which policy category does the product or service fall into — consumer, restricted or prohibited. Second, does the ad, the visual or the landing page contain a claim that is misleading, exaggerated or health-related. Third, are the brands and materials clean.
The order is like this because each next question only makes sense if the previous one has passed. The answers are recorded against specific URLs and specific products, not against the name of the business — so at review it is clear exactly what was declared. The date of the last update of the policies is also checked, because the text changes. With a mixed range, it is decided in advance which part goes into the advertising and which stays outside it.
When in doubt, the category is assessed on the strictest reading, because account verification includes an assessment of whether the products are eligible, and a rejection at this stage is not appealed quickly. That is why, with me, the eligibility check is the first step of every campaign — before payment and before the intent map.
FAQ — Frequently asked questions about eligibility in ChatGPT Ads
Can a dietary supplement shop in Bulgaria advertise in ChatGPT?
Not at present. Health products and services outside the USA are prohibited in principle, and supplements in the USA are approved manually. Products without health claims are assessed separately.
Can an insurance broker or a credit intermediary?
Not at present. Financial services outside the USA are prohibited in principle. Financial education or a budgeting app without offering a service is permitted.
Can a law firm?
No. Legal services outside the USA are prohibited. Legal education without offering a service is permitted.
Can a hotel with a casino?
Yes, if the ad is for accommodation or an experience and gambling is not the focus.
Can a clothing shop with licensed designs?
Only with its own or explicitly licensed designs. Products imitating other brands are treated as counterfeits.
Do the policies change?
Yes. OpenAI states that the categories will expand and the list will change; the text carries a last-updated date and is checked before every campaign.
Conclusion
Advertising in ChatGPT is open to consumer categories — shops, travel, local services, training — and closed to health, finance and law outside the USA, as well as to gambling, alcohol, tobacco and politics everywhere. The baseline standards prohibit exaggerated claims and require consistency between the ad and the page. Eligibility is checked before any budget planning, because every advertiser goes through a manual review and the policies change.
Request a quote for ChatGPT ads →
Related articles: ChatGPT Ads Manager setup · Advertising in ChatGPT: how it works · How much advertising in ChatGPT costs
Sources: OpenAI — Ad policies (31.08.2026) · OpenAI Help Center — Ads in ChatGPT · OpenAI Help Center — Allowing OpenAI Web Crawlers
