Zero-click data reveals who LLMs cite and who disappears
Zero-click erodes traffic without erasing influence — but only for brands that understand the difference between being cited and being visited.
Key takeaways
- Nearly half of Google searches end without a click, but the impact varies sharply by content type.
- Google absorbs roughly a third of Wikipedia's referral traffic; Reddit's conversational format largely protects it.
- ChatGPT now drives more outbound clicks from paid placements than from any other single destination category.
- Declarative, self-contained content faces the highest extraction risk from AI Overviews and LLM responses.
- Brands must track AI citation as a separate KPI — analytics dashboards no longer capture actual reach.
Nearly half of all Google searches now end without a click. That figure, drawn from iPullRank's analysis of 13 billion searches conducted in Q3 2026, has circulated widely enough to become a planning assumption for most senior marketers. What has received far less attention is what the data says underneath that headline rate: zero-click is not one phenomenon. It is several, and they fall very differently depending on who you are.
The distinction matters enormously for any brand whose authority depends on being cited rather than visited.
The encyclopedia problem and the forum exception
Google absorbs roughly a third of Wikipedia's would-be traffic through featured snippets and AI Overviews that reproduce encyclopaedic content without sending users anywhere. For a reference institution, that is a structural loss: the content is used; the visit never arrives. The pattern will be familiar to multilateral organisations and UN agencies whose policy definitions, data tables, and explainer content function as reference material. If Google can answer "what is the Sendai Framework?" from a UNDRR page without a click, it will. The page is cited in the model's response; the session never registers.
Reddit, by contrast, barely feels the effect. iPullRank's data shows Google makes only a marginal dent in Reddit's referral traffic despite the platform's outsized presence in search results. The reason is structural: Reddit answers are inherently conversational, contested, and contextual. A snippet cannot do what a thread does. Users click through because the value is in the exchange, not the answer.
That asymmetry is the first signal brands should internalise. Content that is declarative and self-contained (definitions, statistics, policy summaries) is most exposed to zero-click extraction. Content that is relational, argued, and community-situated is far less so. A financial services firm publishing standalone explainers on Basel III capital requirements faces a different exposure than one whose content drives community discussion or hosts primary data that cannot be reproduced in 40 words.
Where ChatGPT's clicks actually go
The second finding deserves more scrutiny than it has received. ChatGPT now draws a higher share of its outbound clicks from paid placements than from any other single destination category in this dataset. That is not how most marketing teams have been thinking about AI search. The default assumption has been that LLM citations reward editorial authority: well-sourced, long-form, organically credible content. The iPullRank data suggests the traffic mechanics are already more complicated.
For B2B brands in financial services or industrial sectors, where paid search has historically been a supporting channel rather than a primary authority signal, this is a recalibration. If a proportion of ChatGPT's outbound traffic flows through ad placements, then brands treating AI visibility purely as an organic content problem are modelling only part of the equation. The channel mix in AI-referred traffic is not the channel mix in classic organic search.