ChatGPT Ads go self-serve with CPC bidding as OpenAI expands beyond the US

OpenAI has moved ChatGPT advertising out of its early pilot phase and into a broadly available self-serve platform. According to a dev.to article, the company opened Ads Manager to all US advertisers on May 5, 2026 and introduced cost-per-click bidding alongside the cost-per-thousand-impressions model already in place. The shift gives advertisers a more familiar way to control spend, but the product remains structurally different from search advertising: relevance is determined by conversation context and advertiser-supplied hints rather than exact-match keywords or demographic filters.

From pilot to self-serve

The article traces the program back to a US pilot that began in February 2026, with HelloFresh named among the first participants in a February 9 press release. The original format set the template still in use: a sponsored placement shown below an eligible conversation, labeled and kept apart from the assistant's reply, rather than an ad woven into or shaping the answer itself.

The later move to self-serve buying is described as the more consequential step, because it lets advertisers run campaigns without direct negotiation and choose between paying for exposure through CPM or paying for engagement through CPC.

Geographic reach has also widened. The article reports that UK advertisers gained self-serve access on June 6, 2026, and that announcements later the same month covered Japan and South Korea, plus further non-US markets. Because rollout timing varies, it advises checking availability country by country rather than assuming parity with the US launch.

Context matching replaces keywords

ChatGPT Ads is not a keyword product. There is no exact-match keyword targeting and no demographic targeting; campaigns can be narrowed geographically by country or region. Instead, advertisers supply context hints at the ad-group level, and OpenAI's matching relies on those hints together with what is happening in the conversation.

That model demands a different planning discipline, the article argues. Advertisers have to spell out the situations, needs and adjacent topics where their offer is genuinely useful, and make that relevance legible in the creative itself, since there is no keyword list to lean on. They also need to test whether a click in a conversational context carries the same commercial intent as a click from a search result.

The platform has gained performance-oriented features as it has matured, including custom audiences and conversion-optimized campaigns. Even so, the article characterizes the ecosystem as actively evolving rather than a finished substitute for established search or social ad platforms.

Early pricing and measurement signals

Two provisional pricing data points appear in the write-up. Third-party reporting observed CPCs in a range of roughly $2 to $5, while OpenAI's Help Center recommends starting bids between $3 and $5 per click. Both figures should be read as early guidance rather than a forecast, since auction dynamics will shift as inventory grows and more advertisers participate. Reported CPMs are said to have declined from launch levels as supply expanded, which is a further reason not to treat initial results as representative.

On measurement, Ads Manager provides delivery metrics such as impressions, clicks, spend, click-through rate, average CPC and average CPM, and conversions can be tracked through a pixel or the Conversions API. The article stresses that having measurement plumbing is not the same as having mature benchmarks, and recommends that advertisers build their own baselines with controlled test budgets before scaling spend.

Ads cannot shape the answer

The governance boundary receives particular emphasis. OpenAI states that advertising cannot alter what ChatGPT tells a user: ads are labeled, appear separately at the bottom of conversations, and cannot be used to change the assistant's response. That restriction closes off a familiar form of promotional leverage, but it also gives brands a clean separation, where the commercial message is visibly commercial and the answer is meant to stay independent.

Why it matters

The expansion marks OpenAI's most serious step yet into advertising as a revenue line alongside subscriptions and API access, and it is one of the first large-scale tests of paid placement inside a conversational AI product. For advertisers, it opens a new discovery surface that cannot be worked with the search playbook: no keywords, no demographics, and pricing norms that are still settling. For the wider industry, how OpenAI separates ads from answers, and whether users and brands accept that separation, will likely shape the template other AI chat platforms follow as they weigh their own commercial models.