What You Can Advertise on ChatGPT: the Launch Whitelist

At launch, advertising inside ChatGPT is a limited, allow-listed product, not an open marketplace. OpenAI opened it first to a short list of consumer categories — lifestyle & household goods, local services, travel & experiences, and digital products / education — the kinds of verticals where a contextual product suggestion at the bottom of a chat is low-risk. OpenAI has stated it “expect[s] to expand eligible categories over time,” so this list is a snapshot of a five-month-old policy, not a permanent boundary. For the wider picture, start with the flagship overview of ChatGPT Ads.

That whitelist is only the first of three gates every advertiser meets. Some categories are open now; some are restricted and gated to approved advertisers; and some are prohibited outright. The demand behind these questions is real but shallow — the head term “chatgpt ads” draws about 3,600 US searches a month, while the literal query “chatgpt ads policy” is near-zero volume, which is why a single trustworthy map of the rules matters more than a spec sheet. MB Adv Agency runs the paid-media programs — Google Ads, Meta, and the broader paid mix — that a new channel like ChatGPT Ads slots into, and for now this reads as a lifestyle, local, travel, and education product: the same brands that do well on Google’s shopping and local surfaces.

The table below sets out the full launch picture — allowed, restricted, and prohibited — sourced to OpenAI’s Ad policies page (published about Mar 20–22 2026) as reported by AdTechRadar. Treat it as a fast-moving snapshot and verify against the live policy page before you plan a campaign.

Key Takeaways

• ChatGPT Ads is allow-listed at launch. Four consumer categories — lifestyle & household goods, local services, travel & experiences, and digital products / education — are open now, and OpenAI expects to expand them over time.
• Prohibited, restricted, and context-excluded are three different gates. Prohibited categories cannot run at all; restricted verticals (healthcare, financial, legal) are approved case-by-case; and sensitive conversations are excluded regardless of category.
• The reported April-2026 refinement changed placement, not permission. Medical, legal, and financial advice contexts are reportedly no longer categorically ad-free — but placement (which chats can carry an ad) is not the same as permission (who is allowed to run one).
• Answer independence is OpenAI’s stated, unaudited position. OpenAI states advertisers cannot shape, rank, or alter ChatGPT’s responses — a policy it states, not an independently audited fact, and one a 42-state attorney-general coalition is now probing.
• Plan for manual review and a moving target. Check the whitelist, write ads that pass a per-ad review and land on a matching page, and expect the rules of a five-month-old product to keep changing.

What Is Banned and What Is Restricted

Two tiers sit below the whitelist, and they are not the same thing. Prohibited categories are fully disallowed: at launch OpenAI bans adult, dating, and sexual services; gambling; alcohol, tobacco, and drugs; and misleading, deceptive, or obscene ads. It also does not allow political ads. If your business sits in a prohibited category, there is no approval path during the test — you cannot run at all.

Restricted categories are different: they are allowed, but only from approved advertisers, reviewed case-by-case, and rolling out gradually. The restricted verticals are healthcare and medicine, financial services, and legal services — the regulated industries every ad platform gates. So a law firm or a clinic is not banned the way a casino is; it is queued behind verification and manual review. If you run paid search in one of these fields — the kind of work behind our legal PPC services — expect the same document-and-approve rhythm you already know from Google and Meta, applied to a much younger rulebook.

The line to hold in your head: prohibited means you cannot run; restricted means you can be approved, but you are gated and manually reviewed. The mechanics of that approval — verification, screening, and per-ad review — come later on this page. This structure is documented on OpenAI’s Ad Policies page as reported by AdTechRadar and Marketing Brew.

Brand Safety: Where ChatGPT Will Not Show Ads

Category rules decide what can be advertised. A separate set of context exclusions decides where ads appear at all — and this is the brand-safety promise that sets in-conversation ads apart from open display. As messaged at launch, OpenAI shows no ads in or near sensitive conversations: child-safety, explicit, or harmful contexts, and emotionally-vulnerable moments (OpenAI’s own example is mental-health “emotional reliance”). The sponsored unit is a single card labeled “Sponsored” at the bottom of a response, and it is withheld entirely when the conversation itself is sensitive.

There is a second, sharper exclusion: no ads to users predicted or declared to be under 18. If there is a “meaningful probability” that an account belongs to a minor, it is excluded from ads altogether. OpenAI also tells advertisers they receive only aggregate views and clicks — never chats, memories, name, email, precise location, or sensitive attributes. How that data boundary holds up, and where the 42-state attorney-general probe pushes on it, is the subject of the privacy and data pillar, and how personalization feeds ad matching is covered in targeting and audiences.

One tension is worth flagging before the next section. The launch also messaged “no ads near health, mental-health, or politics” — but that health-and-advice line has reportedly moved once already. The genuinely vulnerable contexts stay excluded and political ads stay prohibited; the advice-context default is what changed, and it is handled carefully next. Sources: OpenAI Help Center “Ads in ChatGPT” (article 20001047); the context-exclusion and minors framing is confirmed via reputable outlets quoting OpenAI’s pages.

The Reported April-2026 Refinement (Handle With Care)

This is the part of the rulebook most coverage gets wrong, so read it with the confidence label attached. It is reported — by multiple reputable outlets, not published the same primary way as the launch policy — that in April 2026 OpenAI refined its ad-placement policy to a “more precise” approach, so that medical, legal, and financial advice contexts are no longer categorically blocked from carrying ads by default. In practice, ChatGPT can show an ad in one of those verticals when a query is non-sensitive — a general exercise or diet question, or a broad money question that surfaces a budgeting tool.

Here is the distinction that makes or breaks this section: placement is not permission. Placement decides which conversations are eligible to show an ad — that is what the April change reportedly touched. Permission decides who is allowed to run one — and that did not change. Healthcare, finance, and legal are still restricted verticals that require case-by-case approval to advertise at all. So “medical, legal, and financial advice chats can now carry ads” does not mean “any doctor, lawyer, or adviser can now advertise.” The reported change widened the set of eligible conversations, not the set of eligible advertisers.

Two honesty notes. First, this refinement sits in tension with the launch “sensitive topics stay ad-free” messaging — it is a real narrowing of that promise, which is exactly why it deserves the “reported” label rather than a confident restatement. Second, the exact wording of the changelog is not asserted here: OpenAI’s live policy page returns limited access to automated retrieval, so the specifics are confirmed via reputable outlets rather than quoted verbatim. Verify against the live page before you rely on any single phrasing. Sources: Marketing Brew; Custom Legal Marketing on the placement-versus-permission distinction; and Paubox on the healthcare-governance side. Regulatory context lives in the privacy and data pillar.

Advertiser Eligibility and Ad Review (Reported Mechanics)

Getting approved is a gate, and its mechanics are best described as reported — documented by reputable outlets for a young product, not published as a fully itemized OpenAI checklist. The reported gate runs in stages: business verification (entity, tax ID, address, and an authorized representative), then account-quality and risk screening, then a per-ad review of the ad’s title, copy, and media, plus a landing-page match check — the destination has to match the ad and comply with policy. The full walkthrough of setting this up in Ads Manager lives in how to advertise on ChatGPT, and what the review looks at in the creative itself is covered in ChatGPT ad formats.

One myth worth retiring here, because it circulates on agency blogs: there is no OpenAI- or FTC-published, ChatGPT-specific per-violation penalty figure or “double disclosure” requirement for AI ads. That framing is an agency interpretation, not a rule OpenAI or the FTC has issued for ChatGPT. What is real and simpler: ordinary truth-in-advertising law — the FTC Act’s ban on deceptive and unfair advertising — applies to advertising everywhere, ChatGPT included, and OpenAI’s own policy already prohibits misleading or deceptive ads. State the general rule; do not import an invented penalty number.

Ad–Answer Integrity: OpenAI’s Five Stated Principles

This is the section that earns citations, and the one where honesty matters most. OpenAI frames the relationship between advertising and ChatGPT’s answers as five principles — and the accurate way to present them is as OpenAI’s stated, unaudited position, not as independently verified fact. The headline is answer independence. OpenAI states, in its launch post “Testing ads in ChatGPT” as quoted by reputable outlets, that “ads run on separate systems from our chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses.” The sponsored card is labeled and sits apart from the answer; ads are matched contextually and served on separate systems from the answer itself.

The other four principles round out the position: advertising supports the mission of expanding access; conversation privacy keeps chats private from advertisers with data never sold; choice and control lets users turn off personalization and clear their ads data; and long-term value is captured in OpenAI’s line that “we don’t optimize for time spent.” Read them in Table 2 with the confidence label made explicit. The load-bearing caveat: these are policies OpenAI states, not audited facts — and they are exactly what a 42-state attorney-general coalition began probing in June 2026, naming advertising practices and engagement directly. So the honest phrasing is “OpenAI states ads do not influence answers, and that claim is currently unaudited and under regulatory scrutiny” — never “it is proven that ads cannot.” The regulatory detail routes to the privacy pillar, and the broader trust debate to the future of conversational advertising.

What This Means for Advertisers and Agencies

Strip away the novelty and the practical playbook is familiar. Check the whitelist before you plan a campaign: most mainstream consumer brands are eligible, and regulated verticals are gated, not banned. Write ads that pass a manual per-ad review and land on a page that matches the promise — the same content-and-compliance discipline that keeps a Google or Meta campaign approved. And treat answer independence as OpenAI’s stated policy, not a guarantee, so your brand-safety monitoring does not lean on an unaudited claim.

For agency strategists, the useful move is to map ChatGPT’s rules onto what you already run. Advertisers who know Google’s ad policy and approval will find the manual-review discipline familiar, and the category structure rhymes with Meta’s policy, compliance, and privacy rules. Because this is a conversational surface fed by the same AI-answer shift reshaping organic search, it is worth reading alongside how AI is changing SEO — the discovery layer and the ad layer are moving together. Regulated advertisers should read the restricted tier closely; the kind of governance work behind our financial services PPC translates directly.

The last takeaway is to expect movement. This is a five-month-old product whose whitelist is set to expand and whose advice-context placement rule has already shifted once. MB Adv Agency runs the paid-media programs — Google Ads, Meta, and the broader mix — that already hold to strict brand-safety and policy standards, and a new channel like ChatGPT Ads slots into that same discipline rather than replacing it. If you want to pressure-test where it fits your paid-media mix, start with a Google Ads audit and talk to our team.

Frequently Asked Questions

Methodology & Sources

This pillar is an honest orientation to advertiser-facing policy for a five-month-old ad product, so its load-bearing method is confidence-labeling, not precision. Confirmed facts — the launch whitelist, the prohibited and restricted categories, the context exclusions, the minors exclusion, and the five principles as OpenAI states them — come from OpenAI’s Ad Policies page and Help Center (“Ads in ChatGPT,” article 20001047), confirmed via reputable outlets quoting those pages because the live pages return limited access to automated retrieval. Reported items are labeled as such in the text: the advertiser-eligibility and review mechanics, and the April-2026 placement refinement (Marketing Brew, Custom Legal Marketing, Paubox). The five principles are presented as OpenAI’s stated, unaudited position, with the June-2026 42-state attorney-general probe noted as active scrutiny.

Search-demand figures are from Ahrefs Keywords Explorer (US, 12-month average, retrieved 2026-07-01). Launch scope is corroborated by Search Engine Journal. No mbadv client metrics appear in this article — MB Adv Agency has no ChatGPT-Ads service page and no ChatGPT client dataset, so its perspective here is qualitative, and every figure traces to a named third party or is omitted. Reviewed by MB Adv Agency, July 2026.