GOV.UK
gov.ukAllows AI
Its robots.txt (the public rules file for bots) mentions Meta-ExternalAgent, but still lets all 12 major AI bots into the site.
Last checked 19 Oct 2026. Rules file read today. Everything we checked · History
Bots that train AI Allowed
0 of 9 blocked
Blocking these keeps the site out of future AI training. It does not stop AI from reading the site to answer questions.
Bots for AI search Allowed
0 of 8 blocked
Blocking these usually means the site stops showing up as a source in AI answers.
AI assistants Allowed
0 of 7 blocked
5 of these 7 companies say their assistants may ignore robots.txt.
Bot by bot
| AI bot | Allowed? | Where the rule is | Company says it follows robots.txt? |
|---|---|---|---|
| Bots that train AI · Allowed | |||
| GPTBotOpenAI | Allowed | Not named; the catch-all rule applies | Says yes |
| ClaudeBotAnthropic | Allowed | Not named; the catch-all rule applies | Says yes |
| Google-ExtendedGoogle · opt-out name, not a bot | Allowed | Not named; the catch-all rule applies | Says yes |
| Applebot-ExtendedApple · opt-out name, not a bot | Allowed | Not named; the catch-all rule applies | Says yes |
| CCBotCommon Crawl | Allowed | Not named; the catch-all rule applies | Says yes |
| Meta-ExternalAgentMeta | Allowed | Named, line 9 | Suggests yes |
| BytespiderByteDance | Allowed | Not named; the catch-all rule applies | No statement found |
| MistralAI-TrainingMistral AI | Allowed | Not named; the catch-all rule applies | Says yes |
| AI2BotAllen Institute for AI | Allowed | Not named; the catch-all rule applies | Does not say |
| anthropic-aiAnthropic (legacy) · old name | Allowed | Not named; the catch-all rule applies | Unknown |
| Bots for AI search · Allowed | |||
| OAI-SearchBotOpenAI | Allowed | Not named; the catch-all rule applies | Says yes |
| Claude-SearchBotAnthropic | Allowed | Not named; the catch-all rule applies | Says yes |
| PerplexityBotPerplexity | Allowed | Not named; the catch-all rule applies | Suggests yes |
| AmazonbotAmazon | Allowed | Not named; the catch-all rule applies | Says yes |
| Amzn-SearchBotAmazon | Allowed | Not named; the catch-all rule applies | Says yes |
| DuckAssistBotDuckDuckGo | Allowed | Not named; the catch-all rule applies | Says yes |
| MistralAI-IndexMistral AI | Allowed | Not named; the catch-all rule applies | Suggests yes |
| Meta-WebIndexerMeta | Allowed | Not named; the catch-all rule applies | Suggests yes |
| AI assistants · Allowed | |||
| ChatGPT-UserOpenAI | Allowed | Not named; the catch-all rule applies | Says not always |
| Claude-UserAnthropic | Allowed | Not named; the catch-all rule applies | Says yes |
| Perplexity-UserPerplexity | Allowed | Not named; the catch-all rule applies | Says it mostly ignores it |
| Google-AgentGoogle | Allowed | Not named; the catch-all rule applies | Says it mostly ignores it |
| Meta-ExternalFetcherMeta | Allowed | Not named; the catch-all rule applies | Says not always |
| Amzn-UserAmazon | Allowed | Not named; the catch-all rule applies | Says not always |
| MistralAI-UserMistral AI | Allowed | Not named; the catch-all rule applies | Suggests yes |
| For comparison | |||
| GooglebotRegular search engine, for comparison | Allowed | Not named; the catch-all rule applies | |
| BingbotRegular search engine, for comparison | Allowed | Not named; the catch-all rule applies | |
We test two pages: the homepage, and a made-up page that stands in for the rest of the site. "Blocked" means both are off limits. "Some pages blocked" means only one is. Rules for search results, carts, checkout and accounts don't change these labels, because almost every site keeps all bots out of those pages.
The rules file (robots.txt)
User-agent: *Disallow: /*/print$# Don't allow indexing of site searchDisallow: /search/all*Sitemap: https://www.gov.uk/sitemap.xml # The Meta-ExternalAgent crawler crawls the web for use cases such as training foundation AI models.# It results in timeouts from Vertex that back up requests from users making genuine searchesUser-agent: meta-externalagentDisallow: /search/all* # https://ahrefs.com/robot/ crawls the site frequentlyUser-agent: AhrefsBotCrawl-delay: 10 # https://www.deepcrawl.com/bot/ makes lots of requests. Ideally we'd slow it# down rather than blocking it but it doesn't mention whether or not it# supports crawl-delay.User-agent: deepcrawlDisallow: / # Complaints of 429 'Too many requests' seem to be coming from SharePoint servers# (https://social.msdn.microsoft.com/Forums/en-US/3ea268ed-58a6-4166-ab40-d3f4fc55fef4)# The robot doesn't recognise its User-Agent string, see the MS support article:# https://support.microsoft.com/en-us/help/3019711/the-sharepoint-server-crawler-ignores-directives-in-robots-txtUser-agent: MS Search 6.0 RobotDisallow: /
Notes about AI written in the file: "The Meta-ExternalAgent crawler crawls the web for use cases such as training foundation AI models." · "The robot doesn't recognise its User-Agent string, see the MS support article:"
Everything else we checked
| Front door | Let in The homepage loaded normally for our bot, which says who it is (HTTP 200).GET https://www.gov.uk/ as WalledWebBot, from a residential connection in the United States |
|---|---|
| Anti-bot software | Served through Fastly.From response headers and cookie names only. A cookie shows the product is there, not how it is set up. |
| Readable by AI | No schema.org structured data on the homepage. Readable text is present in the HTML.Homepage only. Useful for site owners; not part of the result. |
| Terms of Service | www.gov.uk/help/terms-conditions (found on the homepage) Sentences flagged by keyword. These are machine-selected excerpts, not a legal reading; open the source for context.
|
| Not found | No llms.txt, ai.txt, tdmrep.json, Content-Signal or Content-Usage lines, licensing signals (RSL License, HTTP 402) or robots meta tags. |
History
47 versions on record since 1 Jan 2023
One square per month, from Internet Archive copies and our own daily checks. Dark: blocked. Gray: some blocked. Light: allowed. Striped: no copy that month.
- May 2026Signals Named Meta-ExternalAgent explicitly, without changing what it may reach.Found in the Internet Archive's copies: changed between 1 Apr 2026 and 1 May 2026. Evidence: before · after
Show exactly what changed
Disallow: /*/print$- # Don't allow indexing of user needs pages- Disallow: /info/*+ # Don't allow indexing of site search+ Disallow: /search/all*Sitemap: https://www.gov.uk/sitemap.xml+ # The Meta-ExternalAgent crawler crawls the web for use cases such as training foundation AI models.+ # It results in timeouts from Vertex that back up requests from users making genuine searches+ User-agent: meta-externalagent+ Disallow: /search/all*+# https://ahrefs.com/robot/ crawls the site frequently…- # https://www.deepcrawl.com/bot/ makes lots of requests. Ideally- # we'd slow it down rather than blocking it but it doesn't mention- # whether or not it supports crawl-delay.+ # https://www.deepcrawl.com/bot/ makes lots of requests. Ideally we'd slow it+ # down rather than blocking it but it doesn't mention whether or not it+ # supports crawl-delay.User-agent: deepcrawl…Disallow: /-- # Google's crawler was sending requests for each variation of query param for the sectors page of licence-finder- # resulting in millions of requests a day.- User-agent: Googlebot- Disallow: /licence-finder/*-
Reports from readers
No reports yet. Tell us what happened when your AI tool tried this site, even if it worked fine.