ChatGPT's ad platform keeps growing. OpenAI still says ads don't change the answer.

What shipped, and what is only reported

Two different levels of sourcing are in play here, and they are worth keeping apart. The product-manager updates — the overview tab, custom audiences, suggested ad drafts, the refreshed static card, the e-commerce format, and the Japan/South Korea activation — were announced by OpenAI in an email to advertisers, reported on July 9–10, 2026 by Search Engine Land and Search Engine Roundtable. We have not seen that email; we are reporting what two independent trade outlets say it contained.

The rules of the system come straight from OpenAI's own durable post, “Testing ads in ChatGPT”, first published February 9, 2026 and updated March 26 and May 7. That post states the mechanics plainly: the test runs for logged-in adult users on the Free and Go tiers only — “Plus, Pro, Business, Enterprise, and Education tiers will not have ads.” Ad selection works by “matching ads submitted by advertisers with the topic of your conversation, your past chats, and past interactions with ads.” And on the point that matters most: “Ads do not influence the answers ChatGPT gives you… When you see an ad, they are always clearly labeled as sponsored and visually separated from the organic answer.”

Why this is good news for the organic slot, not bad news

The instinct when ads arrive on a new surface is to assume the organic result is about to be crowded out — that is what happened to the ten blue links. Read the design here and it is close to the opposite. OpenAI has drawn the line in public and in writing: the answer is generated independently, and the ad is a separate, labeled unit next to it. A competitor with a budget can appear beside the answer. They cannot appear in it.

Which means the sentence a business actually fears — “For a med spa in Scottsdale, I'd recommend…” — is still decided entirely by what the model knows about your business from the open web. It is not an auction. There is no bid that puts your name in it, which also means there is no bid that keeps your name out of it. For a local business, that is one of the few remaining channels where the outcome is earned rather than bought — and it is the one nobody is watching. Across the 153 businesses and 1,472 AI answers in our own audits, 42% had one specific competitor named more often than the business itself — and none of those businesses knew it.

Two things to do with this

• Stop waiting for a paid shortcut. There is currently no way to buy the recommendation, in any of the four major engines, and OpenAI has publicly committed against it. The lever is the same as it was: be the business the model has clear, consistent, corroborated information about. Our guide to getting recommended by ChatGPT is the short version.
• Note who does and doesn't see the ads. Ads run on Free and Go. Plus, Pro, Business, Enterprise, and Education see none. The higher-income, higher-intent professional asking ChatGPT which firm to hire is disproportionately on a paid tier — and for that user, the only thing on the screen is the organic answer.

The takeaway

ChatGPT ads are becoming a serious business with real targeting tools and real geographic reach, and that is worth tracking. But the news underneath the news is that OpenAI keeps re-committing, in writing, to keeping the answer itself unbought. The competitive question for a dental practice, a med spa, a law firm or a home-services company is unchanged: when a customer asks, does the model name you? A free 60-second check tells you, across ChatGPT and Perplexity (the $79 audit adds Gemini and Claude, 25 questions sampled three times each). Agencies running this for clients can use the white-label 5-pack.