Skip to main content
Mythos

Why Referral Traffic Is the Wrong Measure for AI Search

Summary

AI search sends very little referral traffic to most companies, yet AI now plays a part in almost every business purchase. When you measure AI search with the traffic and click metrics built for traditional SEO, the channel looks small, and budget most likely follows that reading. A more useful approach measures AI search with three questions: is your brand in the answer, does the answer describe you correctly, and does the answer recommend you.

What the traffic numbers show

Public companies have started to report AI referral traffic on earnings calls, and the figures are small.

  • Booking Holdings reported on its Q2 2026 earnings call that AI referral traffic, paid and unpaid, was less than 1% of total room nights, and management described the share as "not really moving so much." For a company of Booking's size, AI tools are not yet a meaningful booking source on their own.
  • Etsy reported on its Q2 2026 earnings call that traffic from agentic commerce experiences remains below 1% of total traffic. Etsy also said that this traffic shows higher purchase intent and higher average order value, which suggests the visits are few but further along in the decision.
  • LegalZoom reported on its Q2 2026 earnings call that AI accounted for about 3% of its LLC formation traffic in June 2026. That is the highest of the three figures, and it is still a small share of total traffic.

What the buying data shows

The buying data points the other way.

  • Forrester's State of Business Buying 2026, based on nearly 18,000 business buyers worldwide, found that 94% of buyers used AI during their buying process. AI is now part of almost every business purchase, whether or not the buyer ever clicks through to a vendor's site.
  • Semrush surveyed 2,338 US adults in July 2026 and found that 41% had made a purchase based explicitly on an AI chatbot recommendation. The recommendation itself drove the sale, and in many of those cases no referral visit would have been recorded.

Why the two sets of numbers disagree

A buyer reads the AI answer and acts on it later through a different path. They search your brand name, type your URL, ask a colleague, or request a demo. Each of those actions shows up in your analytics as branded search, direct traffic, or a form fill, and none of them is credited to the AI answer that started it. Referral traffic only counts the small group of people who click a link inside the answer.

In essence, AI search works more like brand exposure than a click channel. Your brand appears in front of a buyer at the moment they are making a decision, and the value of that exposure is mostly invisible to a referral report. This creates a budget problem. If AI search is judged on referral traffic alone, it will look like a rounding error, and teams will underfund the work that decides whether the buyer ever hears your name.

What to measure instead

AI search can be measured in three stages. Each stage answers a question that leadership already asks about any brand channel.

1. Presence: is your brand in the answer?

Presence tells you whether AI includes your brand when a buyer asks about your category. To measure it:

  • Build a list of the questions your buyers actually ask. Pull them from sales call notes, CRM notes, support tickets, and the longer queries in Google Search Console, because these show the wording and constraints real buyers use.
  • Write each question the way a buyer would type it, including their role, company size, budget, and must-have features. AI tailors its answer to these details, so a generic question gives you a generic result.
  • Run each question in the AI tools your buyers use, such as ChatGPT, Google AI Mode, Gemini, Perplexity, and Claude. Each tool draws on different sources, so your presence in one does not predict your presence in another.
  • For each answer, record whether your brand is named, which competitors are named, and where you appear: as the first option, as an alternative, or in a list of others. Position matters because the first option named is most likely the one a buyer looks into first.

2. Accuracy: does the answer describe you correctly?

Accuracy tells you whether AI gets your facts and your message right once it names you. To measure it:

  • Write down the facts buyers use to decide, such as pricing, plan limits, features, integrations, contract terms, and who the product is for. This becomes your answer key.
  • Write down your three to five key messages, the points you want every buyer to hear about your brand.
  • Compare each AI answer that mentions you against the answer key and the key messages. Mark each fact as correct, wrong, or missing, and note which key messages appear.
  • When a fact is wrong, check the sources the AI cited. Errors usually trace back to an outdated page on your own site or a third-party page such as a review profile, comparison article, or forum thread. Fix your own pages first, then update or request corrections on third-party pages, because AI can only repeat what its sources say.

3. Selection: does AI pick you, and does demand follow?

Selection tells you whether AI recommends your brand and whether that recommendation turns into buyer demand. To measure it:

  • For each answer, record whether AI makes a single pick and whether that pick is you. Being named and being recommended are different results, and the second is the one most likely to drive a decision.
  • Ask a follow-up question that names your strongest competitor, and record whether the pick holds. A pick that holds when a buyer pushes back is the one most likely to survive the buyer's own comparison.
  • Track three demand signals monthly: branded search from Google Search Console, direct traffic from your analytics, and demo or contact requests from your CRM. Buyers who act on an AI answer usually arrive through these paths rather than through a referral link.
  • Chart the demand signals on the same timeline as your AI metrics. If they rise together over several months, that is evidence AI search is contributing to pipeline, though it shows correlation rather than proof of cause.

Each stage has two metrics. One is for leadership and ties to market position or revenue. The other is for the team doing the work and shows what to fix. @How to Measure Presence and Accuracy in AI Search covers the first two stages. @How to Measure Selection and Pipeline Impact from AI Search covers selection and how to connect AI search to pipeline.

Related

Sources

About the author

@Ryan Edwards is the founder of @CAMINO5, an SEO and AI search advisory firm based in Burbank, California. He has worked in search for more than 20 years.

Contexts

MythOS is a product of Cooperating Citizens of One, Inc.

TermsPrivacyRefundsContact

© 2026 Cooperating Citizens of One, Inc. · MythOS since 2017 · Not affiliated with Anthropic.