Zero-Click Attribution

Zero-click attribution is the practice of estimating the business value of search and AI answers that influence people without sending them to your website at that moment. A buyer reads a ChatGPT recommendation or an AI Overview that names your brand, and days later arrives through a branded search, a direct visit or a sales call. Last-click analytics credits that later channel; zero-click attribution tries to credit the answer that started it.

The need comes from how rarely AI answers produce clicks. SparkToro and Similarweb measured 68.01% of US Google searches ending without a click in January to April 2026. Semrush found 92 to 94% of Google AI Mode sessions produced no external visit in mid-2025, and Pew Research saw users click a link inside an AI summary in just 1% of visits. Influence that never becomes a click is invisible in referral reports, which is why AI search tends to look smaller in analytics than its effect on demand. This is the measurement side of zero-click search.

It helps to separate what can be measured directly from what has to be inferred. The observable chain runs from crawl to citation to click: a bot fetches the page, an AI answer cites it, a reader clicks through. Referral sessions from chatgpt.com, perplexity.ai and other AI domains (AI-referred traffic) can be joined to conversions and revenue in your own analytics. That number is a floor. Some AI visits arrive without a referrer and are counted as direct (AI dark traffic), and many buyers never click at all. Zero-click attribution estimates what sits above the floor.

Useful inputs, from most to least direct:

• Exposure: mention rate, AI share of voice and visibility for the prompts that matter, tracked over time and per engine.
• Google's own impression data: Search Console's generative AI performance reports, available to all sites since August 31, 2026, count impressions in AI Overviews and AI Mode (AI Overview impressions).
• Demand signals: branded search volume, direct traffic to key landing pages, and branded paid search clicks.
• Self-reported attribution: a 'how did you hear about us' field with ChatGPT, Perplexity and Google AI as explicit options, plus notes from sales calls.
• Experiments: comparing markets, product lines or time periods where AI visibility changed and where it did not.

None of these proves cause on its own. The aim is a defensible estimate with stated assumptions, not a precise number. Compare trends rather than totals, avoid adding overlapping signals together, and write down how each estimate was built so it can be repeated next quarter. A typical model correlates changes in AI visibility for a product category with later changes in branded search and pipeline for that category, using self-reported attribution as a check. Promptwatch measures the observable part of the chain per page, with crawler visits in agent analytics next to citations and AI-referred visits, which gives a baseline to build the estimate on. For the business case, see why teams invest in AI visibility even when referral clicks look small. Related concepts include share of model and ghost citations, where an answer names your brand without linking to your site.