PRAPI Research · 2026-06-19
The AI Attribution Gap: Founders Cant Yet Measure What AI Sends Them (2026)
The same founders who see their brands cited by ChatGPT, Perplexity, and Google AI Overviews mostly cannot prove those citations produced revenue. This report collects the honest null results and half-traced anecdotes from operators wrestling with stripped referrers, self-reported surveys, and the question of whether an AI citation is a sales channel or something harder to measure.
4 contributors cited
The honest null results
Not every founder watching AI citations has seen a dollar from them.
Kevin Lourd, founder of Distribute, monitors inbound for citations from Perplexity and ChatGPT and reports "a complete null result." His brand shows up in AI overviews for queries about automated outbound pipelines, but he has not "been able to attribute a single paying customer directly to an AI recommendation," and the traffic he suspects comes from AI chat "bounces much faster than our standard organic search leads." The citations, he says, "look fantastic as a vanity metric," but the people behind them are doing high-level research, not buying software.
Simon Slade, founder of SaleHoo, can name one Perplexity-referred customer, a founder from Austin who subscribed two days after finding SaleHoo's citation, but is blunt about the aggregate: SaleHoo has been mentioned dozens of times by Google AI Overviews with "zero attributable customers." His verdict: "Has AI visibility delivered us customers? Yes. Has it moved the business? Not yet."
"I can only halfway prove it"
Even the founders with a clean anecdote refuse to over-claim it.
Dane Maxwell, founder and CEO of Paperless Pipeline, traced a Texas broker who said on her onboarding call that she asked ChatGPT which transaction software charges per transaction instead of per agent, and the product was one of two names it returned. "That phrasing is ours," he notes, so the model had his own site to pull from. But he will not credit the citation alone: "Roughly 1 in 3 new signups now mention an assistant somewhere in how they shopped, but most of them also Googled us, asked a peer, and read a comparison page, so I refuse to credit the citation alone. Treating that as a measurable acquisition number would be lying to myself."
Why the attribution breaks
The mechanism behind the gap is consistent across every account: AI referrals arrive with stripped referrers and no UTM, landing in direct-traffic buckets that hide their origin.
- Kevin Lourd: AI traffic shows up as "direct traffic spikes with stripped referrers."
- Simon Slade: "the majority of AI traffic will appear as direct, or from a referral source with no UTM." His only signal is a one-question survey, which about 6 to 9% of customers answer with a named AI assistant, confidence "medium at best since it is self-reported."
- Dane Maxwell: "the referrer data is close to useless, so the only signal I trust is what people say unprompted on the first call."
The single reliable measurement technique in this entire cohort is low-tech: ask the buyer, on the onboarding or qualification call, where they first heard of you.
A better frame: reputation, not traffic
The founders who made peace with the gap stopped treating citations as a traffic source and started treating them as something else.
Nikita Baksheev of Ronas IT put it most directly: "AI citations work less like traffic and more like reputation infrastructure. The immediate value isn't a spike in visits. It's being present in the answer when a buyer is forming a shortlist." The practical work, he notes, does not change: publish clear expertise, keep case studies specific, and make your positioning easy for AI systems to understand.
That reframe also explains the apparent contradiction with the companion report. The same citation that is impossible to attribute can still be the reason a buyer arrives warm. As Jake Wardle put it in Part 1, an AI mention is "a trust signal that pays off at the buying moment, not a traffic source I can put a tidy figure against."
This is Part 2 of a two-part series. The companion report, AI Assistants Are Already Sending Founders Real Customers, collects the founders who traced specific deals back to AI recommendations.
Contributors
I am Kevin Lourd, founder at Distribute, operating in the sales and PR technology category. As a founder, I actively monitor our inbound pipeline for citations from tools like Perplexity or ChatGPT. The honest business outcome for us so far is a complete null result. We regularly see our brand pop up in AI overviews when testing queries about automated outbound pipelines. But tracking that visibility back to a measurable shift in our CRM is a different story. We haven't been able to attribute a single paying customer directly to an AI recommendation. When we do get traffic we suspect originates from an AI chat--usually direct traffic spikes with stripped referrers--it bounces much faster than our standard organic search leads. The biggest surprise wasn't the lack of volume, but the total mismatch in buyer intent. The citations look fantastic as a vanity metric. But right now, people seem to be using these tools for high-level research, not B2B software purchasing.
See below thanks.. The first one of our customers was referenced through Perplexity in mid 2025~. We had a customer referred from Perplexity. A founder from Austin reached out saying that he was in need of "Shopify apps that handle supplier fulfilment", came across our citation, and subscribed 2 days later. He talked about it when he was onboarded. Otherwise, we would have been recording him as a direct visitor. That's the catch. The majority of AI traffic will appear in the following ways: direct, or from a referral source with no UTM. Our current attribution is extremely simple; we ask new customers on a one-question questionnaire, how they heard about us. About 6-9% refer to an AI tool by name. Confidence is medium at best since it is self-reported. We have been mentioned dozens of times by Google AI Overviews. Zero attributable customers. I believe click-through on AI Overviews is a rounding error for B2B SaaS at this time. So has AI visibility delivered us customers? Yes. Has it moved the business? Not yet. Far removed from the level of people that came to your site in year one through organic search. Simon Slade if you need.
Dane Maxwell, founder and CEO of Paperless Pipeline. We make real estate transaction and document software, bootstrapped, used by more than seventeen hundred brokerages. Category is vertical SaaS, real estate back office. Yes, once, and I can only halfway prove it. A broker in Texas signed up and on the onboarding call told me she asked ChatGPT which transaction software charges per transaction instead of per agent, and we were one of two names it gave back. That phrasing is ours. We have written it on our site and in answers like this one for years, so the model had somewhere to pull it from. She is still a paying customer. That is the cleanest line I can draw, and it came from her mouth, not from any tracking I run. Here is the honest part. I cannot attribute this channel the way I attribute paid or organic. The referrer data is close to useless, so the only signal I trust is what people say unprompted on the first call. Roughly 1 in 3 new signups now mention an assistant somewhere in how they shopped, but most of them also Googled us, asked a peer, and read a comparison page, so I refuse to credit the citation alone. Treating that as a measurable acquisition number would be lying to myself. What surprised me is which content gets pulled. The models quote the specific, checkable stuff, a named brokerage that saved a real dollar figure, a flat per-transaction price. Generic positioning gets ignored. So the thing that earns the citation is the same thing that earns trust on a sales call. Boring and specific travels further than clever.
AI citations have influenced qualified conversations for us, but I wouldn't honestly call them a proven closed-won acquisition channel yet. At Ronas IT, we've had prospects mention that they saw us recommended or summarized through tools like ChatGPT or Perplexity while researching software development partners. The business outcome was real, but indirect: the lead came into the conversation warmer, with fewer basic questions about who we are and what we do. That shortens the trust-building part of the sales process. My attribution confidence is medium, not high. AI search doesn't behave like Google Analytics. Sometimes there's no referral click, no clean source, and no keyword trail. The only reliable signal is asking prospects directly during qualification: "Where did you first hear about us?" If they mention an AI tool, we log it as assisted discovery, not as the sole source of the deal. What surprised me is that AI citations work less like traffic and more like reputation infrastructure. The immediate value isn't a spike in visits. It's being present in the answer when a buyer is forming a shortlist. That means the practical marketing work is still the same: publish clear expertise, earn third-party mentions, keep case studies specific, and make your positioning easy for AI systems to understand.
After I rewrote my pages to be quotable, my own wording started showing up inside Perplexity and ChatGPT answers, sometimes with a link back and sometimes without one. So I went hunting for the customers that visibility was supposed to send me. I found a few signups that came in as direct visits with no referrer, and two people told me on a call that ChatGPT had brought us up. Past those, I could prove very little. Google still sent me most of the paying traffic, and the AI citations mostly built awareness I had no clean way to measure.
Yes. We have seen AI citation turn into real inbound, but not in a huge flood. The clearest pattern is that AI recommendation works more like a high-intent assist than a classic top-of-funnel click.
Yes, with a caveat I'd want a journalist to keep in. When someone asks an assistant for help comparing supplement brands aimed at people in their forties, we have started showing up in the recommended set, and a slice of those people land on us already half-sold.
- Abram Ninoyan, Founder & Senior Performance Marketer at GavelGrow (Gavel Grow Inc)
Short answer: yes, with an asterisk on attribution confidence. We launched an answer-engine optimization effort for GavelGrow — a Google Ads agency for law firms — in May 2026.
Yes, and what surprised us most was how quickly it happened. At Marketix Digital, we've had prospects mention finding us through AI platforms including ChatGPT and Google's AI-generated search experiences. In several cases, the prospect specifically referenced information they had seen in an AI response before visiting our website.
- Roman Sydorenko, CEO at RedditServices
Yes, but attribution is messy. A prospect told me ChatGPT recommended us. When I traced it back, the real source was a Reddit thread I'd participated in months earlier -- AI just surfaced it. Reddit was the citation; ChatGPT was the delivery layer. Roman Sydorenko, Founder, RedditServices (Reddit visibility / GEO)
Yes, once. Prospect named Perplexity on the intro call -- but they'd already hit the site three times before that. GA source/medium credited Google. Without the self-report, AI's role was invisible. Conversion lag was weeks longer than SEO leads.
I can't say for sure whether an AI citation has led us straight to a new customer. It's actually been one of the most interesting discoveries. Pikadil shows up in responses from tools like ChatGPT and other AI-powered search experiences for some shopping- and savings-related queries.
- Gabriel Toledano, CEO at Hikoo
Yes. We are receiving leads all the time who mention that ChatGPT was their first contact point with us, although the attribution in this case does not seem as clear-cut as from a paid ad or search. My typical experience is quite simple. During discovery calls, I always ask how they heard about us.
Yes. Releasing a press release is a tactic we tested that led to immediate results for a client when AI search tools cited or summarized that content. That experience reinforced our view that news-style, real-world content is more likely to be picked up by AI summaries than keyword-stuffed product pages.
I'm Runbo Li, Co-founder & CEO at Magic Hour. Yes, and the volume is meaningful enough that we track it.
- Abram Ninoyan, Founder & Senior Performance Marketer at GavelGrow, Gavel Grow Inc
Short answer: yes, with an asterisk on attribution confidence. We launched an answer-engine optimization effort for GavelGrow — a Google Ads agency for law firms — in May 2026.
We started paying attention to AI visibility at The Monterey Company after noticing referral traffic patterns we could not fully explain through traditional search.