What can you advertise on ChatGPT? OpenAI's ad policies explained

Why the policy matters before you spend anything

Eligibility is checked when you open the account, not only when you submit an ad. OpenAI's verification review looks at whether your products or services are eligible under its ad policies. If your category is out, you learn it before your first campaign, so read the rules first.

The policy has three layers: which categories may advertise, what every ad must look like, and which conversations never carry ads.

Categories that are open now

OpenAI describes the current set as a limited group of consumer categories, expected to grow as its review systems mature:

• Household and consumer goods
• Local services
• Travel, entertainment and experiences
• Digital products and education

If your business is a store, a local service provider, a travel brand, a software product or a course, you are inside the core set.

Restricted: finance, health and legal

Three categories sit behind manual, case-by-case approval, and only in the United States.

• Financial services. Approved US advertisers may run ads for products such as credit cards, insurance, mortgages, personal loans or brokerage, and may need proof of licensure. Budgeting apps and financial education that do not sell financial products are generally allowed. Credit repair, debt settlement and bullion are not.
• Health. Approved US advertisers may run ads for areas such as dental services, vision, hospitals and urgent care, health insurance or consumer medical devices. Unapproved products, unsupported treatment claims and high-risk procedures are not allowed.
• Legal services. Allowed in the US only when the advertiser is licensed where the ad is shown. Legal education without services may be fine.

Outside the US, ads for financial, health and legal services are generally prohibited. That matters for advertisers in Europe, Israel or Latin America: a clinic, a law firm or a lender there should plan around this for now.

Disallowed categories

These are currently out, whatever the market:

• Adult content, sexual products and dating apps
• Alcohol above 0.5% ABV, tobacco, vaping and nicotine
• Gambling and real-money gaming (a casino resort may advertise the stay, not the betting)
• Recreational drugs, including cannabis and psychedelics
• Political ads, and ads that exploit contested social issues or tragedies
• Counterfeit goods, scams and impersonation
• Individual job listings and individual housing listings (a platform may advertise without naming a listing)
• Wellness products with unsupported health claims, such as diet pills, detox programs or health coaching

Rules every ad must meet

Even in an allowed category, an ad is rejected if it breaks the baseline standards:

• Truthful. No unfounded claims about results, pricing, affiliations or comparisons.
• Professional. No obscene or shocking language, including in brand names.
• Respectful. No discrimination or content targeting protected groups.
• Clearly an ad. Nothing that imitates the look or voice of ChatGPT.

The rules cover the whole path: title, copy, image and landing page. An approved ad cannot link to a page that introduces something disallowed. You also need to use only trademarks and brand assets you own or are licensed to use, and describe your business identity accurately.

Where ads never appear

OpenAI does not place ads near sensitive or brand-unsafe conversations. The list includes topics such as child safety, graphic violence, hate, misinformation, political content, suicide and self-harm, and weapons, as well as mental and personal health conversations and emotionally reliant exchanges.

Ads are also not shown to users on Plus, Pro or Business plans, or to accounts identified as belonging to people under 18. They appear below the answer, labeled as sponsored, and never change the answer itself.

A quick self-check

Before you open an account, answer three questions. Is my product in the open consumer set? If I am in finance, health or law, am I a licensed US advertiser? Does my landing page contain anything on the disallowed list? Three clean answers mean the policy will not be what holds you back.