crawlindex

Search the index. A domain that is not here can still be measured live on the check page.

Report . September 2026

The state of AI crawler access, September 2026

Drawn from 28 daily crawls of up to 3,695 Tranco-ranked domains, of which 3,675 could be measured.

Sealed record. This month closed and was written to a permanent file on , under rubric 2.0.0 and probe 3.0.0. Nothing on this page will change again, including the cross-tabs and the change list. Cite it freely. The sealed JSON.

Research and data by Fidget Labs BV, Breda, Netherlands. Published as CrawlIndex and free to reuse under CC BY 4.0 with attribution.

15.6%
Block at least one answer-surface crawler
575 sites
4.0%
Block every answer-surface crawler
147 sites
13.7%
Publish an llms.txt
505 sites
64
Mean readiness score
Out of 100

What moved

Between 2026-09-01 and 2026-09-30, the number of measured sites blocking at least one answer-surface AI crawler fell by 90 sites. Sites publishing an llms.txt rose by 36 sites, and sites publishing an agents.md fell by 5 sites.

The mean readiness score across scored sites rose 0.8 points to 64.34.

8 sites answered an agent with HTTP 402 Payment Required. That is a pay-per-crawl gateway rather than a block: access is being metered and sold rather than refused.

Sites that could not be measured are excluded rather than counted as failures, so these totals are over the 3,675 domains actually observed. Method: /methodology. Coverage and its limits: /coverage.

Across the month

Blocking an answer-surface crawler665 to 575 over 28 crawls
Publishing llms.txt469 to 505 over 28 crawls
Mean readiness score63.54 to 64.34 over 28 crawls

Who set the policy

Blocking rate grouped by edge network. Sites behind varnish blocked at 26.3% against 3.0% behind alibaba, which is a far wider spread than anything explained by what the sites themselves publish.

Blocking rate by edge network
Edge networkSitesBlocking AIProportion blockingMean score
Cloudflare1,130116 (10.3%)70.2
Amazon CloudFront42884 (19.6%)65.3
Akamai27243 (15.8%)67.5
Fastly21552 (24.2%)64.6
Google Cloud / GFE1547 (4.5%)57.2
Vercel523 (5.8%)74.3
Varnish3810 (26.3%)60.9
Alibaba Cloud CDN331 (3.0%)64.2
Azure Front Door281 (3.6%)60.3

By publishing platform

Blocking rate by publishing platform
PlatformSitesBlocking AIProportion blockingMean score
WordPress30742 (13.7%)70.4
Next.js30251 (16.9%)66.4
Adobe Experience Manager1346 (4.5%)65.8
Drupal1053 (2.9%)63.8
HubSpot CMS835 (6.0%)71.7
Contentful542 (3.7%)67.1
Nuxt542 (3.7%)62.4
Sanity324 (12.5%)75.3

Blocking by crawler

AI crawlers by number of sites blocking them in September 2026
CrawlerOperatorBlocked byShareOver period
CCBotCommon Crawl49413.4%-97
GPTBotOpenAI46712.7%-93
BytespiderByteDance46212.6%-96
ClaudeBotAnthropic43711.9%-97
meta-externalagentMeta40110.9%-101
Google-ExtendedGoogle38510.5%-103
DiffbotDiffbot37410.2%-3
cohere-aiCohere37010.1%no change
Applebot-ExtendedApple3659.9%-101
PerplexityBotPerplexity3499.5%-13
AmazonbotAmazon3499.5%-103
YouBotYou.com3268.9%+1
ChatGPT-UserOpenAI2998.1%-11
DuckAssistBotDuckDuckGo2617.1%-6
Perplexity-UserPerplexity2556.9%-6
Claude-UserAnthropic2476.7%-7
MistralAI-UserMistral2466.7%-5
OAI-SearchBotOpenAI2436.6%-10
Claude-SearchBotAnthropic2426.6%-6
FirecrawlAgentFirecrawl2055.6%+1
TavilyBotTavily1925.2%-3
Google-NotebookLMGoogle1825.0%-3
kagi-fetcherKagi1604.4%-4

Policy changes this month

Where the index stood

The top and bottom of the leaderboard at the moment this month was sealed.

Most agent-ready

Least agent-ready

Citing this report

Free to quote and republish under CC BY 4.0. Credit Fidget Labs BV and include the measurement date so the claim stays checkable as the index moves.

CrawlIndex by Fidget Labs BV. "The state of AI crawler access, September 2026." https://crawlindex.org (measured 2026-09-30). Licensed CC BY 4.0.

Download the underlying dataset