Joe Spisak has traced approximately $180,000 in revenue and 11 customers to AI search citations. He knows this because he asked. Not with analytics software, not with a conversion tracking pixel — with a form field at sign-up that reads "how did you hear about us?"
The story around AI-driven customer acquisition has moved fast in 2026. Businesses are told to optimize for AI search visibility, to appear in ChatGPT and Perplexity results, to build authority so that AI systems cite them. What the story has left out is that the economics of this attention remain almost entirely unmeasured. A detailed report published by MediaPost on June 25 surveyed founders of commerce businesses about their ability to trace AI citations to actual revenue, and the results are uncomfortable. According to Tom Pinder, founder of Prapi and IT specialist at Startvest, the problem is structural: "AI referrals are nearly invisible in analytics. They arrive with stripped referrers and no UTM and land in the 'direct' bucket, so conventional tracking shows almost nothing for anyone."
Pinder's own brand appears consistently in AI overviews. It has not generated a single traceable paying customer from those citations. He doesn't conclude that the citations produced no business — he concludes that he cannot tell either way. The null result, he says, is mostly "I can't measure it," not "AI sent no one." Traffic may be arriving and landing in the analytics bucket labelled "direct," indistinguishable from someone typing a URL directly into a browser.
Narayan Prasath, founder at Metaflow, went further — segmenting raw event data by chatgpt.com referrer data to isolate what AI was actually sending. The findings suggest a pattern: AI citations tend to drive informational queries ("what is X") rather than commercial ones ("what's the best X for my situation"). The former produces traffic; the latter produces customers. The distinction matters enormously to anyone building a marketing strategy around AI visibility.
Three practical observations:
The attribution gap is solvable with low technology. Spisak traced his $180,000 the same way businesses traced radio ads in 1962: by asking. A post-signup survey, a brief onboarding question, or a simple form field captures what analytics cannot. If you are investing in AI search optimization without this in place, you are operating blind.
"Direct" traffic in your analytics now means something different. If your direct traffic has risen in the past 12 months without an obvious explanation, AI citations may be a significant contributor. The signal is there; the standard labelling obscures it.
AI visibility and AI conversion are different metrics. Appearing in a ChatGPT answer to "what is project management software" is not the same as appearing when someone asks "which tool is best for a ten-person agency." The first is awareness. The second is acquisition. Most AI search optimization advice does not make this distinction.
The broader marketing industry is still calibrating to a landscape where a significant share of customer discovery happens inside AI interfaces that do not pass referrer data in the conventional sense.
Asking customers how they found you has always been good practice. In 2026, it may be the most important marketing question you ask.