How do I measure AEO traffic when AI engines don't send referrers?

AI engines strip or anonymize referrers more aggressively than traditional search, which breaks naive GA4 attribution. ChatGPT Search sends a chat.openai.com or chatgpt.com referrer in some configurations but often sends none; Claude sends no referrer; Perplexity sends perplexity.ai reliably; Google AI Overviews sends a google.com referrer indistinguishable from organic. The result: AI-driven traffic shows up partly as Direct, partly as Organic, and only partly as a clean AI-engine referrer.

A working measurement stack: (1) Filter GA4 source/medium reports for known AI referrers (perplexity.ai, chat.openai.com, chatgpt.com, copilot.microsoft.com, bing.com/chat, you.com, phind.com). This catches the ~30–40% that send identifiable referrers. (2) Watch the Direct/Unknown spike — most AI-driven sessions land here; correlate weekly Direct traffic with citation-share movement to estimate AI contribution. (3) Add a UTM-tagged "smart link" — generate a unique short URL that appears in your AI-friendly content and is included in citation-bait paragraphs; the UTM survives most referrer-stripping. (4) Survey on conversion — a simple "How did you hear about us?" field on demo booking forms captures self-attributed AI discovery. (5) Pair with citation-share data — when share rises 10pp on a tracked prompt set, expect a 15–30% lift in brand-query GSC impressions and a measurable Direct-traffic bump within 4–8 weeks.

The honest constraint: AEO attribution today is closer to PR attribution than to performance-marketing attribution. Stop chasing last-click and accept share-of-voice + branded-search lift as the operational KPI. The cleanest dashboards combine citation share (leading), branded search volume (mid), and self-reported attribution at conversion (lagging).