EZY Research · AI Traffic Index #1 · July 27, 2026
AI crawlers now hit small-business websites 1.8x as often as Googlebot
1.8x
AI crawlers vs Googlebot
763k AI vs 444k Googlebot hits
<1%
AI referral share
17,100 clicks vs 1.80M human visits
43
Median weekly OpenAI hits
Per connected site (63% of Googlebot)
Twelve weeks of server logs across 125 connected business websites, every bot counted at the server, where JavaScript analytics can't see. The full count:
Visitor class · 12-week hits
| Visitor class | 12-week hits |
|---|---|
| Humans | 8,550,000 |
| Generic/unverified bots | 6,360,000 (incl. 489k flagged malicious) |
| SEO tools (Ahrefs + Semrush) | 1,440,000 |
| All AI-ecosystem crawlers | 763,000 |
| Googlebot (+GoogleOther) | 444,000 |
| Applebot | 273,000 |
| Bingbot | 271,000 |
| OpenAI (GPTBot 39k · OAI-SearchBot 26k · ChatGPT-User 148k) | 213,000 |
| Bytespider (TikTok) | 188,000 |
| Meta-ExternalAgent | 150,000 |
| Amazonbot | 126,000 |
| Anthropic (ClaudeBot + Claude user fetches) | 76,000 |
| Perplexity | 7,400 |
EZY Research · 125 connected sites · 12-week totals

AI crawlers out-hit Googlebot 1.8 to 1
Identified crawler hits, all sites, measured at the server
1.8x
AI ecosystem = OpenAI, Anthropic, Perplexity, Meta-ExternalAgent, Bytespider, Amazonbot, smaller labs
AI crawlers collectively out-hit Googlebot 1.8 to 1. And for the typical site, not just the total:
Weekly visits, median connected site
| Visitor | Median weekly visits |
|---|---|
| Humans | 780 |
| Googlebot | 68 |
| OpenAI systems | 43 |
| Anthropic | 28.5 |
| Perplexity | 7 |
AI reads like a search engine. It refers like a rounding error.
The same logs hold both halves of the story. Machine side: OpenAI's systems touch the median site 43 times a week - 63% of Googlebot's rate - and most of those touches are ChatGPT-User fetches, a live request made because a real person just asked something. Human side: in the final four weeks of the window, AI assistants sent 17,100 identifiable referral clicks against 1.80 million human visits - under 1% of traffic, on the same sites AI systems read hundreds of thousands of times. Public data agrees on both halves. On the referral side, the ChatGPT-vs-Google tracker currently shows AI assistants at 0.39% of web traffic channels - ChatGPT's 3.5M monthly referrals running at roughly 1% of Google's 309M - and GA4-based studies land between 1% and 2%; our 0.95% sits squarely inside that range. On the volume side, DataDome's Q2 2026 network report counts 17.7 billion AI agent requests, up 45% in a quarter - consumption is exploding at the network level even where, on our fixed small-business cohort, per-site crawl intensity held steady. The reading is spreading to more sites faster than it is intensifying on any one of them.
That gap is the new economics of visibility: AI consumes your site at nearly Google's rate and answers with it, while sending back a fraction of the clicks. The value moved from the visit to the citation. AI crawling stopped being a novelty line in the log file; it is now the second-largest identified visitor class after humans - and the class least likely to show up in your analytics.
Read that gap correctly and it is the case for AI visibility, not against it. When an assistant answers with your name, the customer often skips the click entirely - they call, visit, or search your brand directly, and analytics files the sale under Direct. A 1% referral share doesn't mean AI influences 1% of decisions; it means most of AI's influence never passes through a link. That is why the metric that matters has shifted from clicks to citations: which businesses the AI names when a customer asks. AI already reads at Google's scale, seven in ten sites have made no decision about it, and each answer names only two or three businesses. The share is small. The leverage is not.
Two patterns from the same logs worth a chart. AI crawlers barely rest: Googlebot's volume drops 13.3% on weekends, OpenAI's drops 3.0%, and Bingbot actually crawls 7% more on Saturdays and Sundays. And Meta's AI crawler is a content specialist - 55% of its path-classified visits land on blog and article pages, against 4-6% for OpenAI, Anthropic and Google.
One number we are deliberately refusing to publish: "AI traffic grew 4.6x in 12 weeks." It's true in our raw panel and it's misleading, because the panel itself grew. On a fixed cohort of 42 sites present all quarter, AI crawl volume was flat. The explosive growth stats circulating in this industry are mostly measuring adoption of measurement, and we'd rather say so than borrow the headline.
The Index will publish monthly from the same panel, same method, with the aggregate tables downloadable. Connect a WordPress or Cloudflare site to EZY free and your traffic is counted - and you get to see your own AI visitor log, live, from the next crawl onward.
Methodology
Data: EZY.ai's 556-domain research panel - 125 connected business websites (WordPress and Cloudflare) with server-side logs, and 387 never-connected audited sites, 27 Apr-19 Jul 2026, UTC. Bots identified by user agent with signature verification where available. Panel skews toward small-business sites whose owners sought an AI-visibility tool. Aggregates only. Full tables available on request.
Aggregate CSVs for every table are available on request. The fixed-cohort check (42 domains present all quarter) is applied to all growth claims; raw-panel growth is reported only alongside it. Dataset license: CC BY 4.0.