Post

Has anyone written/built an `llms.txt` that they’re really happy with? And have you noticed it working/helping? I recently updated my research agent to check all websites for `/llms.txt` before diving straight into parsing pages of HTML and I’m looking forward to seeing if/how effective it is.

15/7/26, 3:15 pm0 reposts6 likes

AlexHarford-TechSEO

@alexharfordseo.bsky.social

15 July

Working in what way? Ahrefs published this recently: ahrefs.com/blog/llmstxt... And even if the llms.txt file is hit, it doesn't mean it's used.

ahrefs.comWe Analyzed 137K Sites: 97% of llms.txt Files Never Get ReadWe analyzed 137K sites & found 97% of llms.txt files are never read. See why bots ignore these files & what this means for your AI visibility

Harry Roberts

@csswizardry.com

15 July

Hmm. Sounds like a bit of a failed standard then 😞

AlexHarford-TechSEO

@alexharfordseo.bsky.social

15 July

Mark is right, it's a proposal rather than a standard.

Harry Roberts

@csswizardry.com

15 July

Yeah, poor choice of words from me! Thanks for keeping me honest haha

SimonCox

@simoncox.com

15 July

Yes built one for my personal 11ty site and also Google OKF and ADR and am tracking the lot. Wrote some of it up here: www.simoncox.com/post/2026-06...

jeremAIas

@jeremias.codes

15 July

If done correctly, it definitely helps. It’s mainly a performance improvement, agents get faster to the info they are looking for. I recently did a lot of optimizations like that for @supabase.com docs.

Harry Roberts

@csswizardry.com

15 July

Yeah, I recently added this to mine: github.com/csswizardry/...

jeremAIas

@jeremias.codes

15 July

I recommend adding a redirect at server level when “Accept: text/markdown” header is present to your site. That does the trick for a lot of agents, they don’t have to parse and read the html at all.