ChatGPT opened a self-serve ad channel with no minimum spend and no reliable benchmark, so treat it as a ring-fenced experiment measured by your own revenue, not the platform's clicks.
A new answer surface just became an ad channel with self-serve access. The pull to test it before it is understood is strong, and the pressure to scale it once it shows a pulse is stronger. Both instincts are traps until first-party outcome data exists, because the platform has not published the numbers a director would need to size a bet.
The Answer Box Grew a Storefront
The incumbent way judges a channel on the platform's own metrics: impressions, clicks, the CTR in the dashboard. That worked when the metrics were independently benchmarked. A brand-new surface with no cross-advertiser benchmark breaks that habit, because there is nothing to compare the dashboard number against.
ChatGPT is now that surface. The answer box grew a storefront, and self-serve access means any team can start spending before anyone can tell them what good looks like.
Who Sees These Ads, and Who Does Not
Wired reports OpenAI announced on January 16, 2026 that it would begin ad testing in ChatGPT, that ads will not affect ChatGPT's responses, appear in clearly-marked sections beneath replies, and show only to logged-in users of the free version and the new $8 per month Go tier, not Plus, Pro, or Enterprise (Wired). Wired is reputable secondary reporting behind a paywall, and OpenAI's own announcement page was not fetchable in this session, so re-verify the exact wording before relying on it (Wired).
Read the audience carefully. The ads reach free and low-tier logged-in users, not the paying Plus, Pro, or Enterprise base. For a large firm selling to senior buyers, that audience skew is a targeting question before it is a budget question.
The 0.91% Problem
A 2026 marketer guide states OpenAI opened the self-serve ChatGPT Ads Manager on May 5, 2026, using a relevance-weighted second-price auction with reported entry bids around $3 to $5 per click (Segwise). That is secondary vendor content, so treat the mechanics as directional (Segwise). Separately, advertiser-reported clickthrough rates as low as 0.91%, against a 6.4% Google Search benchmark, circulated via Adweek, with the caveat that OpenAI has no published cross-advertiser benchmarks (Digital Applied).
Attribute the 0.91% to advertisers, not to OpenAI, because it is advertiser-reported and directional only (Digital Applied). A low reported CTR on a conversational surface is not automatically a failure, but it is a warning against scaling on engagement metrics before downstream revenue is proven.
No Benchmark, No Scale
Two facts define the discipline. There is no minimum spend, so the door is open. And there is no reliable benchmark, so nobody can tell you what performance to expect. A channel that is easy to enter and impossible to benchmark rewards small, controlled tests and punishes anyone who scales on the platform's own numbers.
Because the surface is conversational, creative optimization has limited headroom today, so the edge is offer and audience fit, and the discipline is refusing to scale on a number OpenAI itself will not publish.
The Fork You Now Face
You have two ways to approach ChatGPT ads. Scale on the platform's CTR because there is no minimum stopping you, and bet budget on a number the vendor does not benchmark. Or run a capped pilot tied to server-side and first-party conversion tracking, judge it on downstream revenue such as booked calls and closed deals, and keep spend ring-fenced until independent performance data exists.
That is not a channel yet. That is an experiment with a storefront.
A new surface with no minimum is an open door. A metric with no benchmark is a number you cannot trust. A pilot with no downstream tracking is spending you cannot judge.
Test small, measure downstream. Ring-fence the budget, tie it to first-party revenue outcomes, and refuse to scale on OpenAI's clicks until someone independent can tell you what a good result looks like.
Work With Magnet
Magnet runs new AI ad surfaces as ring-fenced experiments wired to first-party revenue tracking, so a channel gets scaled on downstream outcomes rather than a platform's own CTR. See how Magnet approaches AI-native marketing at https://www.magnet.co.
Sources
- Wired, OpenAI testing ads in the US: https://www.wired.com/story/openai-testing-ads-us/
- Segwise, ChatGPT ads 2026 guide: https://segwise.ai/blog/chatgpt-ads-2026-guide
- Digital Applied, OpenAI ChatGPT ads audience and creative tools 2026 guide: https://www.digitalapplied.com/blog/openai-chatgpt-ads-audience-creative-tools-2026-guide
- Magnet: https://www.magnet.co
