ChatGPT Ads Go Self-Serve With CPC Bidding

OpenAI, the company behind ChatGPT, published two articles four months apart that together describe one of the fastest ad platform buildouts in recent memory. In January, Fidji Simo authored “Our approach to advertising and expanding access to ChatGPT,” a principles document outlining five rules the company would follow as it introduced advertising to its AI assistant. On May 5, OpenAI followed up with “New ways to buy ChatGPT ads,” announcing a self-serve Ads Manager, cost-per-click bidding, a Conversions API with pixel-based measurement, and partnerships with four of the largest global agency networks. Going from principles to a self-serve platform with CPC bidding and conversion tracking in four months signals how seriously OpenAI is treating advertising as a revenue channel, and how quickly ChatGPT is evolving from something brands monitor into something brands buy media on.

The five principles OpenAI committed to in January

The January article laid out five principles for advertising inside ChatGPT, and three of them have direct implications for how the paid channel interacts with organic AI visibility.

Answer independence came first, and it appears to be the most consequential. OpenAI committed to keeping ads completely separate from the answers ChatGPT generates. Ads do not influence responses, responses are optimized based on helpfulness alone, and paying for ChatGPT ads does not make ChatGPT more likely to mention a brand in its organic answers. In practical terms, a brand cannot buy its way into a ChatGPT response.

Conversational privacy followed. Conversations with ChatGPT remain private from advertisers, and OpenAI committed to never selling user data to advertisers. Brands running ChatGPT ads receive aggregate campaign performance data but cannot see what users asked or how ChatGPT responded.

Choice and control rounds out the three that matter most here. Users control how their data gets used, can turn off personalization, and can clear their data at any time. Ads will not appear on accounts where the user is under 18 or has indicated they are a minor, and ads will not appear near sensitive or regulated topics including health, mental health, and politics.

The January article also previewed the first ad format: a sponsored product or service appearing at the bottom of ChatGPT’s answers, clearly labeled and visually separated from the organic response.

From principles to infrastructure in four months

The May article turned those commitments into a working platform. Three additions stand out.

The self-serve Ads Manager, currently rolling out in beta to US advertisers, lets businesses register, add payment information and budgets, upload ad creative, launch and manage campaigns, and view performance in a portal. OpenAI describes it as accessible to companies of all sizes, from small businesses and startups to global brands. Any business willing to create an account can now access ChatGPT advertising without needing a sales relationship or an agency partner.

OpenAI also expanded its partner ecosystem, with agency partnerships now including Dentsu, Omnicom, Publicis, and WPP (four of the largest global advertising holding companies) and technology partnerships including Adobe, Criteo, Kargo, Plexus, and BlackAdapt. Advertisers working through these partners can access ChatGPT ads through the buying tools they already use.

CPC bidding completes the picture. The initial ChatGPT ads pilot ran on CPM (cost per thousand impressions), where advertisers paid based on how many times their ad appeared. CPC (cost per click) means advertisers pay only when someone clicks. OpenAI’s framing of why this pricing model matters in ChatGPT specifically is interesting: conversations are “active and decision-oriented,” with people “often learning about a company, comparing options, or deciding what to buy.” A click from someone mid-conversation who has already described their problem and received an answer carries different weight than a click from a search results page or a social media scroll.

Two channels inside one product

The combination of the January principles and the May infrastructure creates something that looks structurally familiar. ChatGPT now operates as a dual-channel platform, similar to how Google Search has worked for two decades: organic results determined by relevance and authority, paid placements determined by advertiser bidding, and the two systems running independently.

On the organic side, when a user asks ChatGPT a question, the model retrieves information from the web, evaluates sources based on authority and relevance, and cites the pages it considers most trustworthy. The signals that drive organic citations are the same ones that drive traditional search ranking: backlinks from authoritative sources, third-party editorial coverage, structured content, entity recognition, and citation presence across credible publications. No amount of ad spend changes which pages get cited, because OpenAI’s answer independence principle keeps the two systems separate.

On the paid side, brands can now appear at the bottom of ChatGPT responses through the Ads Manager, bid on CPC or CPM, and track conversions. The paid channel provides visibility inside ChatGPT conversations regardless of whether the brand has organic citation presence.

The two sides reward different work. A brand with strong organic citation presence influences what ChatGPT actually says when users ask questions, which no ad buy can replicate. A brand with a paid strategy appears in conversations where it might not otherwise be mentioned, which no amount of organic citation guarantees. The brands reaching users on both sides of the ChatGPT response are the ones investing in both.

CPC clicks from a conversational context

The user context around a ChatGPT ad click is qualitatively different from other CPC environments, and that difference is probably the most underappreciated part of the launch.

On a search engine, a user sees ads alongside results for a query they typed. The intent is clear but the user may still be browsing. On social media, a user scrolls past content and occasionally engages with an ad, usually with low intent. On ChatGPT, the user has already described their problem in natural language, received an answer, and may be actively comparing options or evaluating recommendations. A click from that context comes from someone further along in their decision process than a typical search or social click.

The Similarweb data we covered in earlier posts supports this. Users referred from ChatGPT spend an average of 15 minutes on site versus 8 minutes from Google referrals, generate 12 pageviews per visit versus 9, and convert to transactional sites at a 7% rate versus 5% from Google. Those numbers came from organic referrals. Paid clicks from users who were mid-decision-conversation could perform comparably or better, though it is too early to confirm with data from the new ad platform.

Measurement closes on the paid side, stays open on the organic side

The Conversions API and pixel-based measurement in ChatGPT’s Ads Manager start to address one of the persistent measurement problems with AI visibility. Advertisers can now track whether a ChatGPT ad click led to a purchase, a lead form submission, a sign-up, or another meaningful action. The tracking is aggregate (no individual conversation data), but it gives advertisers the performance data they need to optimize spend and justify budget.

The organic citation side does not yet have equivalent infrastructure. A brand mentioned in ChatGPT’s organic response still cannot track whether that mention led to a site visit, a branded search, or a conversion. The attribution gap between organic AI visibility and business outcomes remains open, even as the paid side gets proper tracking. For now, brand mention share (the percentage of relevant AI responses that include the brand) remains the best proxy on the organic side.

Organic visibility and paid ads reward different investments

The ChatGPT ads launch does not reduce the value of organic AI visibility work. The answer independence principle means organic citations cannot be replaced by ad spend, which actually makes the organic side more valuable now that a paid alternative exists. Brands that appear in ChatGPT’s organic answers because of strong citation presence, entity recognition, and authoritative third-party coverage will continue to appear there regardless of whether competitors are running paid ads alongside those answers.

Link building and digital PR feed the organic side of the dual-channel model. Every editorial mention in a credible publication, every backlink from an authoritative domain, every guest post on a site with editorial standards contributes to the citation pool that ChatGPT draws from when assembling its organic answers. That work compounds in an environment where OpenAI has committed to keeping the organic and paid channels independent, which means the citation presence built through these tactics cannot be outspent by a competitor with a bigger ad budget.

Link insertions into existing authoritative content put a brand inside pages that ChatGPT’s retrieval system already trusts, strengthening the organic citation presence on a timeline that new content alone cannot match.

The dual-channel structure that emerged between January and May rewards brands that have built authority and content quality on the organic side while also being willing to test the paid channel as it matures. The January principles said advertising would support broader access to AI without compromising the product. The May launch delivered the infrastructure to make that real, and the two channels are now live, separate, and rewarding different kinds of work.