Skip to content
TodayTuesday 27 October

How AI discovers, crawls, and recommends the web.

LiveData through 27 Oct · updated nightly
Pageviews from AI a day · 7-day average
+11.2%on the 30 days before, across 6,600 sites
100K
200K
Pageviews from AI a day as a 7-day average of the feed's smoothed daily figures, 1 Jun 2025 to 27 Oct 2026, split by AI. The highest day in view was 27 Oct 2026, with 270,847.
Pageviews from AI a day
DateTotal
28 Sep 2026225,622
29 Sep 2026231,968
30 Sep 2026237,297
1 Oct 2026240,817
2 Oct 2026242,472
3 Oct 2026242,574
4 Oct 2026241,173
5 Oct 2026238,230
6 Oct 2026235,163
7 Oct 2026232,461
8 Oct 2026228,822
9 Oct 2026224,025
10 Oct 2026218,730
11 Oct 2026211,696
12 Oct 2026202,832
13 Oct 2026193,236
14 Oct 2026184,055
15 Oct 2026177,492
16 Oct 2026175,715
17 Oct 2026177,298
18 Oct 2026182,626
19 Oct 2026191,456
20 Oct 2026203,539
21 Oct 2026216,999
22 Oct 2026231,329
23 Oct 2026243,515
24 Oct 2026253,546
25 Oct 2026261,788
26 Oct 2026268,349
27 Oct 2026270,847
ChatGPT
79.4%−4.5 pts
Claude
16.3%+5.3 pts
Gemini
3.1%−0.8 pts
Seven others
1.3%0.0 pts

Share of pageviews from AI over the last 30 days, change on the 30 days before

Across the datasets

One finding from each dataset, with its date
Open any to explore
Brands340of the top 500 changed score overnight. 159 rose, 181 fell. 28 Oct
Falls on the left, rises on the right.
Sources74%of cited links appear in one answer, then never again. 30 Mar 2026
Models4%of questions got the same top brand from every AI we compared. 11 Mar 2026
Searches1 in 182AI searches use the exact words you typed. Sep to Oct 2026
Crawlers67% vs 5%of best-for pages GPTBot read as HTML, against Markdown. 5 Aug 2026
HTML
Markdown
HTML 67%, Markdown 5%
Web readiness6%of the 50 biggest sites publish llms.txt. 16% of the top 5,000 do. 14 Mar 2026
Fake crawlers44%of requests signed GrokBot were fake, where we could check. 16 Oct 2026
Library79studies to read, from us and other researchers.Open the library

Common questions

What is Trakkr Data?

Trakkr Data is a free set of open datasets, several of them updated every day, showing how AI search engines like ChatGPT, Gemini, Perplexity and Claude discover, crawl, cite and recommend the web. It covers brand rankings, citation sources, AI traffic, crawler behaviour, model agreement, query fan-out and web adoption.

Is Trakkr Data free to use?

Yes. Every dataset is free to explore and reuse under a CC BY 4.0 licence, with attribution to Trakkr. The rankings are also published as the open AI 500 dataset on GitHub.

How often is the data updated?

Brand rankings are scored every morning, starting at about 08:30 UTC. AI traffic and the citation index update nightly, and crawler reach updates weekly. The studies behind pages like models, searches and web adoption are dated snapshots, and each one shows its date.

Where does the data come from?

Every day Trakkr puts thousands of buyer questions to AI and records which brands the answers recommend and which pages they cite. AI traffic comes from Google Analytics properties connected to Trakkr, and crawler data from the server logs of connected sites. The datasets describe observed behaviour, not editorial opinion.