The exchange rate
Every crawler that hits your site is proposing a trade: content in, visitors out. For twenty years the Google exchange rate was good enough that nobody did the math. The 2026 numbers say it is time to start.
According to SearchSignal's 2026 AI referrals and citations benchmark, which aggregates studies covering between 13,770 and 82,000 domains, Google's classic search sends one referral for every 5.4 pages it crawls. ChatGPT crawls 1,091 pages per referral. Claude crawls 38,066. These engines are heavy readers and light senders, and no amount of optimization changes that structure. What you can change is whether all that reading turns into your brand being named in the answer.
| ENGINE | PAGES CRAWLED PER REFERRAL |
|---|---|
| Google (classic search) | 5.4 |
| ChatGPT | 1,091 |
| Claude | 38,066 |
Where AI referrals actually come from
When AI engines do send traffic, the distribution is lopsided. ChatGPT drives 78% of all AI referral traffic in the benchmark; Perplexity holds about 15%; everything else splits the remainder, with Gemini growing fastest at 388% year over year. Overall AI referral share is still modest — between 0.1% and 1.08% of total site traffic in most studies, with IT and technology sites at the high end around 2.8% — but the 357% annual growth rate is the number strategy should be built on.
Share of AI referral traffic by engine (SearchSignal, 2026)
The visitors are better
The volume argument against GEO falls apart on conversion data. The same benchmark tracks AI-referred visitors moving from converting 49% worse than other traffic in January 2025 to 31% better by the holiday season. That trajectory matches what we see in client analytics: a buyer who arrives from an AI answer has already been pitched, compared, and pre-sold by the engine. Fewer visitors, further down the funnel.
“AI referrals are not a traffic channel. They are a trust channel with a click attached.”
What to do with this
First, let the crawlers in — blocking GPTBot or ClaudeBot to punish the exchange rate just removes you from the answers your buyers read, and shipping an llms.txt file makes the reading count. Second, measure mention rate and citation rank weekly, the way you measure rankings; a 38,000-to-1 crawl ratio means the value shows up as brand presence in answers long before it shows up in referral reports. Third, weight your GEO roadmap toward ChatGPT visibility, because 78% of the referral upside lives there — our Perplexity teardown shows how fast an engine-specific push can move. The engines have decided the terms of the trade. Cited brands collect on it; everyone else just feeds the machine.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.