Cited, not clicked

One click for every 322 times we were quoted

Each square below is one of 322 citations. The single green one is the click. That is the trade as it stands: the AI reads our research, answers the buyer inside the chat, and the buyer almost never arrives on the page the answer came from.

Where the few clicks come from

When an AI tool does send someone through, it is mostly ChatGPT, which sent 225 of the 336 clicks. Microsoft Copilot sent 34, even though Copilot is the engine we can see citing us the most. The tools that quote us the hardest are not the ones sending the readers.

What the machines are mining

This is not a few stray mentions. On the buyer queries where we appear at all, our pages are about 22% of every source Copilot cited, so we are a heavily used source, not a marginal one. And the pages getting mined hardest are the practical, research-heavy ones: detailed buying guides and plain-English explainers. Exactly the work that takes the most effort to produce.

This is not just us

One site is one site, so we went looking for whether the wider web sees the same thing. It does, and the numbers from the big measurement firms are if anything starker than ours.

Two honest caveats on the outside data. Cloudflare measures how often AI crawlers take content versus send a visit back, which is a close cousin of our citation-to-click gap rather than the exact same metric. And Google disputes the Pew study as unrepresentative, so we lean hardest on the Ahrefs and Similarweb figures and include Pew only because all four, measured very differently, point the same way. Nothing here rests on one source.

Why it matters

For twenty years the open web ran on a simple bargain. You write something genuinely useful, a search engine indexes it, and in exchange it sends you readers. Those readers are how independent sites survive, through subscriptions, ads, or in our case helping people find a broker.

Generative AI keeps the first half of that deal and drops the second. It still needs good source material, in fact it needs more of it than ever, but it now answers the question inside the chat window. The reader gets the answer and the site that did the work gets a citation, not a visit. Scale that across every research-led publisher and the model that funds good information starts to wobble. That is the quiet story under these numbers, and it is the same one Cloudflare, Pew and Similarweb are telling at internet scale.

We obviously have skin in the game here, clicks are part of how we keep the lights on, so weigh our framing with that in mind and judge the raw numbers for yourself. We are not crying poor about it either. We would rather be the source AI reaches for than not, and being cited at all is a signal the work is trusted. But it is worth naming plainly, because a lot of sites are watching their AI citations climb and their traffic flatten and have not connected the two.

Method, and where it is rough

Both figures come from our own Microsoft Clarity dashboard, the same one we use to run the site. We have laid out the limits honestly because they all push the same way: the true gap is wider than the headline, not narrower.

• Citations (108,214) are Microsoft’s own count, from Clarity’s AI Visibility report, of the times a Copilot answer referenced a NestPath page as a source. It covers the Copilot and Bing AI ecosystem only. ChatGPT, Claude and Gemini cite us too but do not publish those counts, so this is just the slice we can see, and the real total is higher.
• About the 322 to 1. Inside the Copilot ecosystem, where Clarity measures both sides, it was 108,214 citations against 34 click-throughs, a gap of about 3,200 to 1. The 322 to 1 headline is deliberately more conservative: it divides those Copilot citations by clicks from every AI tool combined, not a like-for-like ratio, so nobody can say we picked the friendliest denominator.
• Clicks (336) come from Clarity’s referrer report across every AI tool. AI products increasingly strip the referrer, so the true click count is genuinely uncertain and most likely undercounted. That is why we stress-test it: even at five times the measured clicks, the gap is still about 64 to 1.
• The two windows are close but not identical: citations to 23 Sept 2026, clicks to 22 Sept 2026, each a rolling 30 days.
• It is one site, in one niche, in one country. We have set it beside the named industry data above precisely so it does not rest on our data alone.

Sources

• NestPath first-party data: Microsoft Clarity AI Visibility and referrer reports, 2026.
• Cloudflare
• Pew Research Center
• Ahrefs
• Similarweb