The first comprehensive look at how ChatGPT, Claude, Perplexity, and Gemini send visitors to websites. Based on anonymized data from 1,200+ sites tracked on PerchLens during Q1 2026 (January – March).
Share of total AI referral sessions by engine. Q1 2026 (January – March).
Developer tools and personal blogs see the highest share — content depth + technical specificity match what AI engines cite.
Across the 1,200+ sites in our sample, four content patterns predict whether a page will appear in AI search responses:
Pages with at least one statistic the LLM hadn't seen elsewhere were cited 4.2× more often. AI engines treat numerical claims as evidence.
Comprehensive guides outranked thin listicles 3.8× in citation frequency. AI engines prefer single sources to fragmented ones.
Bylined content with linked LinkedIn or academic profiles was cited 2.6× more than anonymous posts. Trust signals work.
Pages with structured FAQ blocks were cited at 2.1× the rate of similar pages without them. ChatGPT in particular uses schema as ranking input.
Data aggregated from sites tracked on PerchLens during Q1 2026 (January – March). AI referrers identified by HTTP referer header matching against 6 known engines (ChatGPT, Perplexity, Gemini, Claude, Copilot, Phind). Sites with fewer than 100 monthly pageviews excluded for k-anonymity. Industry tags self-reported by site owners. Content pattern analysis used a stratified sample of 100 sites with ≥1k AI sessions.
Numbers reflect early-access data and will be updated quarterly. For methodology questions or data partnerships, email research@perchlens.com.
See exactly which AI engines send you visitors, which pages they cite, and what to write to grow your share.