AI Search: A 5-Layer Framework

Traditional attribution sees only part of the buyer's journey in AI search — the link click. Everything that happens inside the conversation with the AI slips past it. To close that blind spot, Search Engine Land proposes measuring performance across five layers — from whether bots even reach you, all the way to revenue in the CRM.

The point is to stop arguing over "the one metric that matters" and build the picture in layers: each level explains what became of the traffic and demand from the one below. The Search Engine Land piece walks through every layer in detail.

What to measure, layer by layer:

• Access — verified AI-bot visits, crawl frequency and depth: if LLMs can't retrieve your content, nothing else matters;
• Visibility — mention and citation rate, plus AI Overviews and AI Mode impressions in Search Console against a fixed prompt set;
• Assistants and referrals — GA4 traffic from ChatGPT, Gemini and Perplexity as one signal, not the final figure;
• The dark funnel — branded clicks and branded conversions from people who discovered you in AI but returned via another channel;
• Business outcome — pipeline, deals, revenue.

Build the report from the bottom up and check where the chain breaks: usually the failure isn't in revenue but at the access or visibility layer. AI doesn't change the business goals — it changes the evidence you use to prove progress.