ChatGPT's First Advertisers Can't Prove Their Ads Worked

OpenAI launched its first advertising pilot with agencies including Omnicom, WPP, and Dentsu — and the brands who paid a minimum of $200,000 to participate cannot see whether their ads worked.

A technical glitch in OpenAI's Ad Manager tool is preventing advertisers from accessing their own campaign data. Combined with click-through rates running at 0.91% — roughly seven times below Google Search benchmarks — the pilot has produced early results that make ROI calculation nearly impossible for most participants.

The reporting failure

A glitch in Ad Manager blocks advertisers from viewing their own campaign performance data. One enterprise advertiser spent just 3% of a $250,000 budget over several weeks with no way to assess whether any of those impressions drove meaningful action. OpenAI has said that it is "in the early testing phase of ads in ChatGPT" and that "the goal right now is to learn and refine the experience for consumers before expanding it more broadly." Recently, OpenAI COO said ads in the chatbot will be “an iterative process.”

Adthena research puts ChatGPT's CTR at 0.91%, against 6.4% on Google Search. The gap reflects more than a platform finding its footing. ChatGPT users are in a task-completion mindset — they are using the tool to accomplish something specific, not to browse. That behavioral difference makes traditional sponsored-result formats a structurally poor fit for most campaign objectives.

The structural audience ceiling

Ads in ChatGPT reach only users on the free tier and ChatGPT Go, OpenAI's $8/month plan available in 171 countries. All paid subscribers — Plus, Pro, Business, and Enterprise — are permanently excluded from seeing ads. Ad-free access is positioned by OpenAI as a feature of paid tiers, not a temporary restriction.

The practical consequence is that advertisers in the pilot are buying reach among OpenAI's least-committed users. The product's most active and loyal audience — subscribers paying between $20 and $200 per month — is entirely beyond their reach, a ceiling that competitors like Meta and Google do not face.

Impression volume has grown. Sensor Tower data shows impressions jumped approximately 600% from early to mid-March 2026. Dentsu reports that delivery is "quickly building momentum, with volume increasing week-over-week." Omnicom has secured placements for more than 30 clients. But volume is not measurement, and measurement remains the pilot's central unsolved problem.

What the pilot reveals about AI advertising

Google has delayed ads inside Gemini's conversational interface despite significant investment. Meta's AI tools serve ads by default. Amazon's Rufus integrates sponsored product listings natively within its AI shopping responses. ChatGPT enters this competitive landscape with two compounding challenges: a broken reporting layer and a structural audience ceiling. Both are fixable in principle, but neither has a public resolution timeline.

For brands committed to the pilot, the $200,000 minimum entry point is non-recoverable. For agencies evaluating AI advertising more broadly, the early data reinforces an open question: whether conversational AI interfaces are better suited to brand awareness than to direct response, and whether CPM pricing that applies to awareness campaigns can justify investment in the absence of measurable conversion data.