The forgotten llms.txt still describing your old website
When we rebuilt this website, its llms.txt still described the old one: services we no longer offer and pages that no longer exist. It had been written by a plugin that was no longer installed.
What /llms.txt said, measured on this site
What we found
When we rebuilt onintel.com, there was an llms.txt file at the root of the site. It described the old website: services we no longer offer, such as viral marketing and social media, and links to pages that no longer existed.
It had been generated by an SEO plugin that had already been removed. The plugin was gone. Its file was not.
What llms.txt is
llms.txt is a plain-text summary placed at /llms.txt, proposed so that language models can quickly understand who a site belongs to and where its important pages are. It is not an official standard, and there is no guarantee that a given assistant reads it. More and more AI tools do look for it, and when they find it, it is written by you, so they tend to trust it.
That is why an outdated llms.txt is worse than none: it gives AI a clear, confident and wrong description of your business.
Why an uninstalled plugin keeps talking
The plugin did not generate the file on each request. It wrote a physical file to the server. Uninstalling a plugin removes its code, not the files it created.
A physical file is also served directly by the web server, before WordPress runs. Even if you install a new tool that generates llms.txt from your current pages, the old file stays in front of it and nobody sees the new one.
How to check yours
- Open
/llms.txton your site. Does it list the services you sell today, with working links and in the right language? - Run
curl -I https://yoursite.com/llms.txtand look atLast-Modified. A date from before your last redesign is a warning sign. - If a plugin should be generating the file, look in the root folder of your hosting for a real
llms.txtfile. If one exists, it is the one being served.
How we fixed it
We deleted the static file. llms.txt is now generated from the pages that are actually published on this site, in both languages, and it changes when a page is edited. Pages marked as not indexable stay out of it.
The same pages also produce /services.json, the list of services in a format made for machines. Neither file invents anything that is not on the site: where there are no public prices, they say so.
The wider lesson
Everything written for machines ages in silence: llms.txt, robots.txt, old sitemaps, structured data hard-coded in a theme. None of it shows on screen, so none of it gets reviewed. After every redesign or migration, open each one and read it as a machine would.
Want us to check this on your site?
Get in touch