By OpenAI just named Smartly as its first creative adtech partner, and the 13-year-old ad optimization firm has a mandate that goes well beyond tweaking banner placements. Smartly's long-term goal, according to CEO Laura Desmond, is to build interactive ad formats that mimic ChatGPT's own conversational interface, letting brands talk back to users inside the chatbot.
The partnership, reported exclusively by Business Insider on April 1, arrives at a specific inflection point: OpenAI's U.S. ad pilot crossed $100 million in annualized recurring revenue within six weeks of launch, according to CNBC, with more than 600 advertisers now running campaigns inside ChatGPT.
From static placements to conversational ads
The current ChatGPT ad experience is deliberately plain. A user researching smartphones might see a Best Buy ad at the bottom of a response, clearly labeled as sponsored and visually separated from the organic answer. Sensor Tower found that more than 100 brands advertised in the pilot's first weeks, with 44% being retail companies, according to Business Insider.
Smartly wants to change the format entirely. Desmond pointed to a campaign the company built for the UK retailer Boots on Meta platforms as the template. In that prototype, clicking an ad launched a chatbot that served personalized gift recommendations through a series of questions. Smartly reported the conversational format was nearly five times as effective at driving sales as Meta's standard ad units.
"The opportunity with conversational advertising is you can do more follow-ups, and you can ask again," Desmond told Business Insider. "The experience for people will get way more relevant, way more personal, and hopefully be seen as a much better value exchange."
Translating that to ChatGPT is the next step. Initially, Smartly will help entertainment, retail, and sports clients optimize their existing ChatGPT ads in real time, the same kind of performance tuning it already does for clients like Spotify and Uber on other platforms. The conversational ad formats are the longer-term play.
The revenue math and its constraints
The $100M ARR figure is striking velocity for a pilot that only began in February. OpenAI's own blog post on the ad test noted that roughly 85% of free and Go-tier users in the U.S. are eligible to see ads, but fewer than 20% are shown them on any given day. That restraint is intentional. Ads only appear on the free tier and the $8/month Go plan; Plus, Pro, Business, Enterprise, and Education subscribers see none.
Analyst projections vary widely. Evercore ISI's Mark Mahaney projected OpenAI could reach $25 billion in ad revenue by 2030, according to Business Insider. Internal documents leaked earlier this year reportedly projected $1 billion in ad revenue for 2026. But MoffettNathanson flagged a structural constraint in a note this week: ChatGPT has limited ad real estate because the company has committed to keeping ads separate from organic answers. That is a fundamentally different architecture than Google Search, where ads sit atop results, or Meta's feed, where sponsored content blends with organic posts.
OpenAI has also partnered with Criteo for ad placement logistics, and the company has told advertisers that ads will not appear near sensitive topics including politics and health. Users under 18 are excluded entirely.
The trust fault line
This is where the story gets more interesting than a revenue update. When choosing between Claude, ChatGPT, and Gemini, the advertising question now factors into the decision in a way it didn't six months ago.
Anthropic has staked a public position. "Ads in conversations with Claude would be incompatible with what we want Claude to be: a genuinely helpful assistant for work and for deep thinking," the company said in a statement ahead of its Super Bowl campaign in February. That campaign, which featured the tagline "Ads are coming to AI. But not to Claude," drew a sharp response from Sam Altman, who called it "deceptive" on X. Anthropic reportedly saw an 11% user boost from the spots.
Google occupies a middle position: it shows ads in AI Overviews within Search but not inside the Gemini chatbot itself. That distinction matters because it preserves the chatbot as an ad-free space while monetizing the search surface.
OpenAI's approach is to let users control the tradeoff. Pay $20/month for Plus, and you see no ads. Use the free tier, and you subsidize access by seeing sponsored content. The company reports no impact on privacy-related trust metrics so far, and dismissal rates on ads have been low. But "so far" does six weeks of work on a deliberately conservative pilot, where less than a fifth of eligible users see ads daily.
Conversational ads change the calculus. A static banner at the bottom of a response is easy to ignore. An interactive ad that mimics the chatbot's own conversational interface blurs the line between the assistant helping you and the assistant selling to you. OpenAI's stated principle is that "ChatGPT's answers remain independent and unbiased, conversations stay private, and people keep meaningful control over their experience." How that principle holds up when the ad itself is a conversation happening inside the same interface is an open question.
What we don't know yet
- Whether conversational ads will be clearly distinguishable from organic ChatGPT responses in practice, not just in policy language. The Boots prototype ran on Meta's platform, which has different UX conventions than ChatGPT.
- How OpenAI will handle the targeting mechanics for conversational ads. The current system matches ads to conversation topics and chat history, but an interactive ad format could generate new user data mid-conversation that raises fresh privacy questions.
- Whether the $100M ARR figure holds up as the pilot scales internationally to Canada, Australia, and New Zealand, or whether early advertiser enthusiasm was driven by novelty and limited inventory.
- What regulatory bodies in the EU and elsewhere will make of conversational ads in AI assistants, given that the AI Act's transparency requirements for AI-generated content could intersect with ads that behave like chatbot responses.
What comes next
OpenAI is expanding its ad pilot to Canada, Australia, and New Zealand in coming weeks. Self-serve ad buying is reportedly on the roadmap. And with Smartly now building the creative optimization layer, the ad experience inside ChatGPT is going to evolve faster than the cautious pilot phase suggested.
The deeper shift is structural. The AI assistant market is splitting into two revenue models: ad-supported (OpenAI, and eventually others) and subscription-only (Anthropic, at least for now). That split mirrors the one the streaming industry went through, and it will shape not just pricing but the fundamental relationship between user and assistant. An AI that makes money when you click an ad has different incentives than one that makes money when you renew your subscription.
Desmond acknowledged the tension. "As we gain confidence about what people are comfortable with," the companies will adapt the ads, she told Business Insider. That framing, comfort as a moving target, is honest. It is also the kind of language that should make users pay close attention to what "comfortable" means in practice as conversational ads roll out.
Paul Menon covers AI for The Daily Vibe.
