Seer Interactive (2026) — Early observations on ChatGPT ads

Summary:

This article from Seer Interactive, a marketing agency that sells SEO, paid media, AI and consulting services and also runs ChatGPT ads for its own clients, looks at ChatGPT ads in their first weeks: how often they show, on which prompts, who is buying, and how well each ad fits the answer above it.

The data comes from Seer’s own AI tracker across 97 client workspaces, about two weeks of data, updated on July 28, 2026. Ad-to-answer fit was scored on Seer’s own 1-to-5 rubric. The article does not say how many prompts or responses were tracked, which markets they cover, or how the rubric was applied.

Key findings:
• Ads appeared in 77.1% of tracked responses, up 34 percentage points in two weeks from 38% on June 9. They showed in 96 of 97 client workspaces.
• Ads run across the whole funnel. Evaluation prompts took 38.7% of impressions, awareness 24.3%, advice 18.8% and comparison 16.5%.
• B2B software and SaaS buy the most, with 21.7% of impressions, followed by cybersecurity at 14.2% and comparison and review sites at 10.7%. Insurance and finance, education, retail, legal, healthcare and travel are also testing.
• Brands rarely own their branded prompts. A brand’s own ad showed on 7.3% of its prompt responses, up from 2.2% in the first pull. Seer saw one rival use the slot to steer users away from the brand.
• Ads often sit under answers that recommend someone else. The advertiser appeared in ChatGPT’s response 5.4% of the time and was cited as a source 3.3% of the time.
• Fewer ads are fully off-topic, but fewer are fully on-topic too. Ads with no meaningful connection fell from 22% to 4%, and on-topic ads fell from 50% to 38%, while category-adjacent ads rose from 28% to 58%. The average fit score stayed near 3.3 out of 5.
• Affiliates and aggregators made up 11.6% of ads, down from 16.2%. BestMoney (~4.5) and Top10 (~4.3) scored higher on fit than most direct brands next to them.

Margin notes:
This is a vendor report: Seer sells ChatGPT ads management, and the data comes only from its own clients’ tracked prompts, not a random sample of ChatGPT use. It covers about two weeks of a fast-changing pilot, and the first-pull ad share was revised from 7.1% to 38%, so the numbers are unstable. Fit scores rest on Seer’s own unpublished rubric, and the article calls alignment “improving” even though the share of on-topic ads fell.