ChatGPT Opens Advertising to All: Not a New Channel, a Rule Rewrite
On February 9, 2026, OpenAI officially launched advertising inside ChatGPT. Six weeks later, CNBC reported: ChatGPT ads had surpassed $100 million in
On February 9, 2026, OpenAI officially launched advertising inside ChatGPT. Six weeks later, CNBC reported: ChatGPT ads had surpassed $100 million in annualized revenue. Three months later, on May 5, OpenAI opened its self-serve ad platform to all U.S. businesses — from a closed beta at $250,000 to a zero-barrier public beta in under a hundred days. Many people assumed this was just a "new ad channel." It is not. This is the first true rewrite of the advertising industry's underlying rules. 1. A Hundred-Day Blitz: The Full ChatGPT Ads Timeline The pace of ChatGPT ads exceeded almost every industry analyst's expectations. January 2026, the ad pilot launched — U.S. only, $250,000 minimum, pure CPM at $60 per thousand impressions. A closed test that "only let the wealthy sit at the table." February 9, ads went live for Free and Go tier users, appearing below the conversation and strictly independent of the AI's answer. Almost no public data existed; OpenAI stayed silent on the system's details. March 26, CNBC and Reuters reported intensively — ChatGPT ads topped $100 million annualized in six weeks, with under 20% of eligible users seeing ads, implying a huge inventory expansion was coming. April 10, the self-serve Ads Manager Beta launched. The minimum dropped from $250,000 to $50,000, with a CPC bidding option added at a suggested $3–$5 per click. April 17, expanded to Australia, New Zealand, and Canada — the first multi-country coverage. Late April, the Conversions API and Pixel launched for conversion tracking — advertisers could finally track post-click behavior instead of guessing. May 5, Beta registration opened to all U.S. businesses at ads.openai.com. Digiday reported OpenAI had dropped the $50,000 minimum. Partners included the four agency groups Dentsu, Omnicom, Publicis, and WPP, plus Adobe, Criteo, and StackAdapt. In three months, a closed enterprise ad experiment became a self-serve platform for everyone — faster than any traditional ad platform. 2. From $60 to $3: The Market Truth Behind the Pricing ChatGPT's pricing evolution tells a complete market story. Phase one: $60 CPM, the illusion of scarcity. For context, Facebook's average CPM is around $12 and Google search around $20–$40. $60 meant advertisers paid a premium for "unknown conversion quality." But pricing here wasn't set by the market — it was set by scarcity, with only about 600 advertisers allowed in. Phase two: $60 fell to $25 within ten weeks. The Next Web reported that once true supply and demand emerged, CPM naturally settled. StackAdapt's DSP showed CPMs as low as $15 in less competitive scenarios — the market's real answer to "what is a ChatGPT ad worth": not $60, but a $15–$60 range depending on competition density. Phase three: CPC $3–$5, from awareness to performance. OpenAI did what every ad platform eventually does — moved from CPM to CPC, because performance advertising is where the spend is, and performance advertisers only pay for clicks. The current suggested CPC is $3–$5. For reference, Google search averages about $2–$3. ChatGPT's slightly higher click pricing reflects the premium of "high-intent" traffic — users who have already expressed clear purchase intent in the conversation. OpenAI also announced an upcoming CPA (pay-per-conversion) bidding mode — meaning ChatGPT Ads is converging toward the full capability matrix of Google Ads and Meta Ads. 3. How It Actually Works Placement: ads appear below the relevant answer in a user's ChatGPT conversation. The official phrasing is "appear below relevant conversations" — ads are context-matched to the conversation, not keyword- or audience-targeted. Importantly, ads are strictly independent of the AI's answer and clearly labeled as ads — unlike Google's approach of embedding ads in AI Overviews. ChatGPT physically separates ads from the AI answer. Who sees ads: only Free and Go tier users. Pro, Business, Enterprise, and Education users see no ads. OpenAI protects the paid experience while monetizing free users. As of May 2026, ChatGPT Ads Manager offers: – A Campaign / Ad Group / Ad three-level structure – Reach objectives (CPM buying) and Clicks objectives (CPC buying) – Conversions API and Pixel conversion tracking – Real-time impression and click reporting – Third-party measurement integration (Adobe, etc.) What's missing: audience targeting and custom audiences (not yet supported), A/B testing, and CPA bidding (coming soon). For now, only conversation-context matching is available — you cannot control keywords and audiences as precisely as in Google Ads. This means less "controllability" than traditional platforms, but higher "naturalness." 4. The New Business for Ad Agencies StackAdapt, the first DSP partner, has already sketched a clear business model: a $15–$60 "proto-auction," plus a platform fee and management fee. It positions ChatGPT ads as a "discovery layer" — capturing user attention in the middle of the decision journey. All four major agency groups are in. But the value of traditional agencies is being redefined. In the ChatGPT ad system, the classic model of "advance payment + rebate + creative tweaking + audience targeting" loses much of its value, because: – There are no custom audiences to build — pure contextual targeting – Creatives matter less — ads are triggered by conversation context – Bidding optimization is limited — only CPC and CPM for now So where is the agency's new value? First, corpus engineering. Helping clients translate product information into structured information the AI can understand — industry question banks, selling-point vectors, user-scenario libraries. This is the core of GEO (Generative Engine Optimization). Second, intent mapping. Breaking down at which conversation nodes and question phrasings your ad should appear. A CRM vendor doesn't just bid on "CRM software" — it covers intent scenarios like "what to do when sales double-books a client" or "should a 10-person team adopt a CRM." Third, scenario design. Reconstructing the user's decision journey so the ad appears at the exact second the user is most persuadable. Agencies that can do these three things will be very valuable over the next three years. Those who only know how to tweak accounts will get cheaper. 5. Can Advertisers Make Money? This is what everyone cares about most. Early positive signals: $100 million annualized revenue in six weeks means more than 600 advertisers kept spending. High-ticket categories (SaaS, insurance, travel, B2B tools) responded most positively. Truist analysts called 2026 the "turning year" for AI-driven advertising. The core challenge is attribution. This is the biggest uncertainty. A user might see an ad in ChatGPT and click, or might have been "educated" in the AI conversation and then search for the brand directly to buy — the latter can't be attributed to the ChatGPT ad. OpenAI currently only provides impression and click data; the new Conversions API and Pixel are closing the gap. Until multi-touch attribution matures, most advertisers can only evaluate ROI with UTM tracking and incrementality testing. Which advertisers are more likely to profit? High-ticket, heavy-decision categories where users already compare for a while — insurance, home improvement, SaaS, B2B tools — may turn a profit, because ChatGPT captures high-intent users already asking for solutions, far more efficient than the "passive exposure" of regular feed ads. Low-ticket, impulse purchases — $9.99 links, snacks, phone cases — this isn't for you yet. A $3–$5 click cost can't be sustained by a $9.99 order value. 6. The Deeper Impact on the Ad Industry Short term (2026–2027): Google search ads face their first real "intent interception" competitor. A user asking "which robot vacuum is best for a $5,000 budget" used to go to Google search; now they ask ChatGPT directly. The AI answers, and an ad waits below. From "being seen" to "being trusted," two or three steps disappear. The optimizer skill model is being rebuilt. Feed optimizers spend 70% of their time adjusting bids, adding budget, pausing campaigns, and changing creatives — this craft is mostly useless in the ChatGPT ad system. Medium term (2028–2029): eMarketer forecasts AI-driven search ads at $2.08 billion in 2026 (1.3% of U.S. search spend) and $26 billion in 2029 (13.6%). OpenAI is more aggressive: $2.5 billion in 2026, $53 billion in 2029, $100 billion in 2030. Whether or not OpenAI's forecast holds, one thing is certain: domestic large models will inevitably follow the "end-of-conversation" ad model. In biology this is called "niche displacement" — the environment hasn't changed, but the spot you used to occupy is taken by a new species. You won't go extinct immediately, but you'll slowly run out of food. 7. The Far-Reaching Impact on Business This isn't just about the ad industry. Brand-building logic is being reshaped. Brands no longer win by "being seen" but by "being found by AI." Your structured information on the open web — official site, reviews, wiki, user-generated content — forms the new brand asset. If your brand is "not found" in the information AI retrieves, you won't even survive the buffer period. An overtaking opportunity for SMEs. Local service providers with enough structured, truthful information in the AI corpus can precisely reach users in high-intent scenarios. Remaking the content business. AI-generated content plus AI-placed ads forms a closed loop. Content creators must ask: is your content helping the AI understand your brand? The ongoing privacy and compliance tug-of-war. Conversation data is used for ad matching, but OpenAI promises "strict separation of conversation and ads." This balance will remain under regulatory scrutiny. Conclusion ChatGPT ads are not a "new channel" — they are the first large-scale practice of the ad industry shifting from "interrupt-driven" to "intent-driven." In biology there's a term — "niche displacement": the environment hasn't changed, but the spot you used to occupy is taken by a new species. You can stay where you are; you won't go extinct immediately, you'll just slowly run out of food. One last question, for every brand owner to ask themselves: Does your brand have enough structured, truthful information on the open web for AI to recognize? Does your team have anyone doing or learning corpus engineering, intent mapping, and scenario design? If not, now is the time to start. A changing world demands deep insight. See you next time.
