Business tests of ChatGPT Ads are producing traffic without sales, and Peter Jaffray said his June campaign spent $415 for 60 clicks and no interested buyers. He ran three campaigns targeting Ontario and Alberta, then found the ads drew attention but not action.
Choice OMG and the 0.6% rate
60 clicks came from nearly nine-thousands impressions over two weeks, according to Jaffray, who manages Choice OMG as its managing director and privacy officer. He said he expected about four-thousand targeted impressions, but the campaign reached more people and still finished with a clickthrough rate of about 0.6%.
2% was the benchmark he said he would normally see when he compared the campaign with search ads rather than display, which is why the result looked weak before the money was even fully spent. “Ads only reach free ChatGPT users,” he said, and that audience mix helped explain why the campaign looked more like a display channel than an engaged channel.
August 2026 tools arrived later
August 2026 brought geotargeting, custom audience ad options, conversion optimization, OpenAI Pixel Conversions API, and third-party measurements to ChatGPT Ads, but those tools were not in place for Jaffray’s June run. That timeline matters for advertisers weighing whether the platform’s newer controls can do better than broad traffic buying; the first test here was built before the measurement stack expanded.
1 campaign from Nicholas Verity, CEO at Cleverly, ran into the same problem recently: he had to pause it quickly after decent CTRs still produced zero conversions. He said OpenAI reported 57 clicks while Google Analytics reported less than 20 visits from the same campaign, after he added UTMs to trace the traffic.
Google Analytics and mismatched counts
57 clicks versus less than 20 visits leaves a gap that can only come from measurement, attribution, or duplicate counting inside the platform pipeline. Verity said the numbers did not match reality, and added that if the platform’s own numbers do not match reality, either the metrics are unreliable or it is double-counting clicks from the same user. He said he could not justify testing more spend on top of it.
1 more test from Daniel Johnson, founder of We Scale Startups, pointed the same way: he said the click count never quite lined up with Google Analytics, and not always in the same direction. If OpenAI’s newer geotargeting and measurement tools improve over the next 6 months, advertisers may get a cleaner read on whether ChatGPT Ads can move beyond clicks and into purchases; for now, the gap between traffic and conversion is the number that matters.
