The loudest ad launch of the year got louder on Tuesday: five more countries, nine markets, “continuing to expand.” Seventeen words in it have not changed since January — the ones that lock every advertiser out of the conversation where a patient decides.

At some point tonight, in your client’s market, a person who was told this week that they need imaging is going to open ChatGPT and ask about it. What an M R I costs without insurance. Whether the contrast is dangerous. Whether the shadow on the report is the bad kind. That conversation is the most commercially valuable moment in your client’s funnel — a person deciding, alone, at night, with money on the table — and as of this week it is one of the only places on the consumer internet where no advertiser is allowed to follow them in.

Not because nobody wants to. Because the platform wrote it down.

On Tuesday OpenAI updated the launch page for ChatGPT Ads: now live in the United Kingdom, Mexico, Brazil, Japan and South Korea, “continuing to expand to more markets this year.” That makes nine markets since the U.S. test opened in February. The trade press covered it the way it covers every new ad surface — formats, markets, what it means for budgets. All real. All beside the point for anyone whose clients wear white coats.

The point is seventeen words that have been sitting in the same announcement since winter.

markets with ChatGPT ads when the U.S. test opened in February → after the March expansion into Canada, Australia and New Zealand → as of Tuesday’s launch in five more countries

The program only grows. The February 9 post opened the U.S. test for logged-in adults on the Free and Go tiers; the March 26 update reported “no impact on consumer trust metrics, low dismissal rates” and announced the next three countries; May 7 named five more; Tuesday declared them launched. Through all of it, the sentence above did not change by a word — I pulled both pages this morning and compared them. Seventeen words on the health clause, twice, identical.

Read the sentence the way a lawyer would before you repeat it in a client meeting, because it is narrower and stranger than the summary version going around. It does not say healthcare companies may not advertise. It says ads are not eligible to appear near health topics — the exclusion is drawn around the conversation, not the advertiser. A meal-kit ad cannot show up in a mammogram conversation any more than a hospital’s can.

But run either reading through an imaging center and you land in the same place. The conversation where the patient decides — cost, safety, where to go — is inventory that does not exist. Nobody outbids your client there. Your client also cannot get in with money. Presence in that room is decided entirely by what the model has to work with when it answers.

Two weeks ago I sent you the government’s theory of what a click near a health page reveals — the Hims complaint. This is the same collision from the other direction: the advertising industry keeps discovering that health is the category it is not allowed to touch the normal way. OpenAI, to its credit, put the fence up in the launch documents instead of the settlement.

Kept honest. Everything quoted here is OpenAI describing its own product — there is no regulator and no third party in this edition, and no independent verification of where ads do and do not actually appear. The boundary of “near health” is drawn by a classifier nobody outside the company can see, and its edges are not published. The March claim of “no impact on consumer trust metrics” is the company grading its own homework; I treated it as evidence the program will keep growing, not as truth about trust. The inference that the organic answer is the only commercial surface inside a health conversation is mine, not OpenAI’s — the company says nothing about how providers should respond to any of this. And the sentence this edition is built on says “during our test” in its own first clause — if you repeat it to a client, repeat that part too. It costs nothing today and it protects you the week the policy changes.

Source, primary, both fetched from openai.com this morning, August 12, 2026, with a browser user-agent: OpenAI, “Testing ads in ChatGPT,” originally published February 9, 2026, carrying dated updates of March 26, May 7 and August 11, 2026 — source of every launch fact above: the U.S. test opening for logged-in adult users on the Free and Go tiers with Plus, Pro, Business, Enterprise and Education excluded from ads; the ad-matching description (“the topic of your conversation, your past chats, and past interactions with ads”); the March 26 language on “no impact on consumer trust metrics, low dismissal rates of ads” and the expansion “starting with pilots in Canada, Australia, and New Zealand”; the May 7 naming of the United Kingdom, Mexico, Brazil, Japan and South Korea; the August 11 update declaring “ChatGPT Ads has now launched” in those five and “continuing to expand to more markets this year”; the sentence quoted verbatim in full (“During our test… health, mental health or politics”); and the safeguard language “We’ll expand responsibly as safeguards mature.” And OpenAI, “Our approach to advertising and expanding access to ChatGPT,” January 16, 2026 — source of the five ads principles; the identical seventeen-word health clause, confirmed word-for-word against the February post this morning; “We keep your conversations with ChatGPT private from advertisers, and we never sell your data to advertisers”; “We do not optimize for time spent in ChatGPT”; ads placed “at the bottom of answers”; and ChatGPT Go at $8 a month in 171 countries as the access program the ads fund. The nine-market count is the two posts’ own arithmetic: the U.S., then Canada, Australia and New Zealand, then the five launched Tuesday. Trade coverage (Search Engine Land’s feed) NOMINATED this topic by surfacing the August 11 update; no fact in this edition comes from coverage. The Hims reference is to this desk’s own August 5 edition and its underlying complaint; no new claim about that case is made here.

Sometime tonight, a person in your market who was just told they need imaging will open ChatGPT and ask about it. What an MRI costs without insurance. Whether contrast is safe. Whether the shadow is the bad kind.

On Tuesday, OpenAI announced ChatGPT ads just launched in five more countries. Nine markets now since the US test opened in February. The marketing press covered it the way it covers every new ad surface — formats, budgets, forecasts. The next great inventory.

All real. And all beside the point if you're in healthcare.

The point is seventeen words that have been sitting in the same announcement since winter, unchanged through every expansion:

"...ads are not eligible to appear near sensitive or regulated topics like health, mental health or politics."

I pulled the January principles document and the launch page this morning and compared them. Word for word identical. The program went from one market to nine. The sentence never moved.

Read it carefully, because it's stranger than the summary version. It doesn't say healthcare companies can't advertise. It says no ad — anyone's ad — appears near a health conversation. The exclusion is drawn around the room, not the advertiser.

So the moment your future patient is deciding — alone, at night, money on the table — is a room no advertiser can enter. Nobody can outbid you there.

You can't buy your way in either.

The room isn't empty, though. Something answers. And what it says about your center depends entirely on what exists about you where a machine can read it: your prices, your prep instructions, your locations, your accreditation, the actual questions patients ask — on pages that load without a login.

That work has quietly become the only marketing that operates inside the room.

Two things I'd do this month.

First, the afternoon baseline. Ask ChatGPT the ten questions your patients ask before booking. What an MRI costs without insurance in your city. Where to get a low-dose CT. Whether they need a referral. Write down what it says about you and about your competitors, and date the file. In six months, that's the before picture.

Second, be ready for the pitch. Someone will offer you "ChatGPT advertising" this quarter — the advertiser signup page exists, and the sales motion will find healthcare eventually. The seventeen words are your reply. Placement inside a health conversation isn't restricted inventory. Per the platform's own launch documents, it's inventory that does not exist.

One caveat, because it's the honest one: the sentence begins "During our test." This is test-phase policy, not a constitution. Regulated categories have a long history of coming online later — with paperwork and an auction attached. If that door opens, the change will land on the same page, and whoever reads it first tells their clients first.

Which is exactly why I wouldn't wait for it. The paid door opens with restrictions and a bidding war. The organic footprint is free, it compounds starting now, and it's the same work that shows up in every AI surface at once.

The loudest ad launch of the year — and the most important sentence in it is about where ads will never run.

FIRST SOURCE · one verified original-source finding, composed for one reader · this edition: marketing — for the agency that will be asked about ChatGPT ads before the year is out, and would rather answer from the document than the coverage