OpenAI has opened ChatGPT advertising to health and finance advertisers, each subject to individual review, and has extended the announcement to five more countries. OpenAI updated its ChatGPT advertising policies on 10 August 2026 to add housing and job listings, then on 11 August 2026 extended the announcement to the United Kingdom, Mexico, Brazil, Japan and South Korea. Search Engine Journal reported the change on 11 August 2026.
Thailand is not on that list. Thai clinics, hospitals and insurers cannot buy these placements today, and nothing in the announcement says when or whether they will be able to.
What OpenAI changed on 10 and 11 August 2026
The 10 August update added housing and job listings to the categories advertisers can run. Health and finance advertisers became eligible as well, with the qualification that each one goes through individual review rather than being approved as a category. OpenAI named health insurance, dental services, supplements, vision products, medical testing, hospitals and urgent care as approvable.
Legal advice and court related services stay restricted, with an exception for general legal education. Political advertising remains banned outright. On 11 August the announcement was extended to the United Kingdom, Mexico, Brazil, Japan and South Korea.
Eligibility is not placement, and the difference decides whether your ad ever appears
This is the part most readers will get backwards, so it is worth stating flatly. OpenAI runs two separate sets of rules. Advertiser eligibility asks whether your business is allowed to buy ads at all. Placement asks whether an ad may appear in the particular conversation a user is having right now.
Ads stay blocked from chats OpenAI treats as unsuitable. That includes political conversations and what OpenAI calls sensitive user contexts, which cover discussions of mental and personal health. So a dental clinic can pass individual review, become an eligible advertiser, and still never appear in a conversation where someone is describing a health worry. Passing eligibility buys you the right to compete for placements, not the placements themselves.
For anyone modelling reach, that gap is the whole model. A health advertiser's addressable inventory is not "health conversations in ChatGPT". It is whatever subset of commercially framed conversations survives the placement filter, and OpenAI has not published a figure for that.
Four policy revisions in five months
Search Engine Journal counts four revisions to these rules in five months. Laid out in order, the pattern says more than any single change does.
| When | What changed |
|---|---|
| February 2026 | OpenAI announced the advertising pilot |
| April 2026 | Categorical blocks on medical, legal and financial placements removed |
| July 2026 | Health and finance sections reworked |
| 10 August 2026 | Housing and job listings added; health and finance eligible subject to individual review |
| 11 August 2026 | Announcement extended to the UK, Mexico, Brazil, Japan and South Korea |
Four rewrites in five months is the planning constraint. Any media plan written against the August rules is being written against a document that has changed roughly every six weeks since February.
The SE Ranking number, and why it is not a clean comparison
Search Engine Journal cited SE Ranking data finding ads on 28.69% of healthcare commercial prompts in ChatGPT against 2.64% in Google's AI Mode. The wider study covered 50,006 commercial prompts across 20 niches and put the overall ad rate at 25.94%.
SE Ranking states the caveat itself, and it should travel with the number every time the number is quoted. The two prompt sets were different and were collected 23 days apart. That is not a like for like comparison, and treating the ratio as "ChatGPT shows ten times more health ads than Google" reads far more into it than the method supports. The study also measured commercially framed prompt categories rather than the personal health conversations OpenAI's placement rules are written about, which is a second reason the healthcare figure does not describe what a worried user sees.
What the announcement did not say
The gaps are as useful as the contents, and several of them are the questions a media planner would ask first. Search Engine Journal's report does not answer them, and neither does this post.
- What individual review actually requires. No documentation list, no licence check, no turnaround time was given.
- Whether more countries are coming, or when. The 11 August list is a list, not a roadmap.
- How large the eligible inventory is once the placement filter has done its work.
- Whether advertisers already approved under the earlier rules have to do anything.
- What happens to an approved health advertiser whose ad would otherwise land in a sensitive user context. The rules say the ad is blocked there, not where it goes instead.
Where a source is silent, the honest answer is that it is silent. Anyone telling you what the review process asks for is filling a gap OpenAI left open.
What advertisers in the named countries should check
If you are buying in one of the five named markets, the practical work sits before the campaign rather than inside it. Confirm which of your business lines maps to a named approvable subcategory, because the list is specific: health insurance, dental services, supplements, vision products, medical testing, hospitals and urgent care are named, and anything sitting next to those without being one of them is an open question that only review will settle.
Then separate your forecast into two numbers. One is the share of your category that OpenAI will let you buy against. The other is the share of that which survives placement rules you have no control over. A plan built on the first number alone will overstate reach, and the second number is not published.
The legal side is worth a look too. Legal advice and court related services remain restricted with general legal education excepted, which is a narrow carve out and one that a firm publishing explanatory content might read more generously than the policy intends.
Where this sits in the wider ChatGPT ads thread
This is one item in a run of them. Recent changes have included oCPC bidding and advanced matching, a shift to seven day budget averaging, and SE Ranking's earlier finding that ChatGPT was showing ads on roughly 26% of commercial prompts. Those are bidding and delivery mechanics. This change is about who is allowed to buy and where, which is a different layer, and the two should not be read as one story about ChatGPT ads maturing.
The connection to search visibility work is real though. If commercial answers in ChatGPT increasingly carry ads, the organic side of that surface becomes a distinct discipline from ranking in Google, which is the territory covered by generative engine optimisation and by day to day ChatGPT SEO work.
What this means for Thai marketers
Labelled analysis. Thailand is not among the named countries, so this is not a buying opportunity for Thai advertisers right now. The honest position is that a Thai clinic or insurer has nothing to action here beyond reading it.
Three points are worth carrying forward. First, the eligibility architecture is what will arrive, whenever it arrives: individual review for health and finance, a named list of approvable subcategories, and a separate placement filter you do not control. Preparing means having the documentation an individual review would ask for, not building campaigns.
Second, four policy rewrites in five months makes any ChatGPT Ads media plan short lived. Budget written a quarter ahead against these rules is budget written against a moving document.
Third, and this one matters more than the OpenAI news itself: Thai advertising rules for medical and health claims are a separate constraint and are completely unaffected by an OpenAI policy change. Thai regulation of what a clinic, hospital or supplement seller may claim applies to Thai advertisers regardless of which platform accepts the money. An OpenAI approval is not a Thai approval. Anyone in AI search planning for a healthcare brand should treat the local rules as the binding constraint and the platform rules as the looser one.
FAQ: ChatGPT ads for health and finance advertisers
Can Thai advertisers buy ChatGPT ads now?
No. The countries named in the 11 August 2026 announcement are the United Kingdom, Mexico, Brazil, Japan and South Korea, and Thailand is not among them. The source did not say whether Thailand is planned or when the list might expand.
Is my clinic automatically eligible now?
No. Health and finance advertisers are eligible subject to individual review, which means approval is granted advertiser by advertiser rather than to the category. OpenAI named hospitals, dental services, urgent care, medical testing, health insurance, supplements and vision products as approvable, but eligibility still runs through review.
Will my ad show when someone asks ChatGPT about a health worry?
Not if OpenAI classes that conversation as a sensitive user context. Placement rules block ads from conversations covering mental and personal health, and they apply on top of advertiser eligibility, so an approved health advertiser can still be kept out of the conversations that look most relevant.
Does the 28.69% figure mean ChatGPT shows ten times more health ads than Google?
No, and SE Ranking says so itself. The 28.69% ChatGPT figure and the 2.64% Google AI Mode figure came from different prompt sets collected 23 days apart, which makes the ratio unreliable as a direct comparison.
Do I have to do anything today?
If you advertise in Thailand, no. There is no account to open, no setting to change and no budget to move, because the placements are not available here. Keeping an eye on the country list is enough.
The full report is at Search Engine Journal.
If you want a view on how your brand currently appears inside AI answers, before any of this becomes buyable in Thailand, Relevant Audience can take a look with you.
