If you have read anything about AI search optimization in the past year, you have probably been told to add an llms.txt file to your site. It is presented as the robots.txt of the AI era: a simple text file that tells language models how to understand your content. It takes about twenty minutes to create. The honest question is whether those twenty minutes buy you anything.
No major AI provider has committed to reading llms.txt. Adding one is cheap and harmless. Expecting it to change your AI visibility is not supported by anything currently observable.
Primo Collab
What it is supposed to do
The proposal is straightforward: a markdown file at the root of your domain that gives AI systems a curated map of your site. Rather than crawling everything and inferring what matters, a model could read your llms.txt and get a clean summary of your key pages, what your business does, and where the important content lives.
As an idea it is sensible. Structured, deliberate context beats inference.
The adoption problem
Independent tracking put llms.txt adoption at roughly 8.7% of the top 1,000 websites as of mid 2026. That is not nothing, and it has grown. But adoption by publishers is only one half of the equation, and it is the less important half.
The half that matters is consumption. A file only works if something reads it, and the major AI providers have not committed to reading llms.txt. Without that, publishing one is closer to leaving a note than to configuring a system.
What actually drives AI visibility right now
The things that observably affect whether AI systems can find, understand, and cite your content are unglamorous and mostly predate AI entirely:
- Clean, crawlable HTML. Content that requires JavaScript execution to appear is harder for everything to read, not just AI.
- Real structured data. Schema.org markup for your organization, products, articles, and FAQs is read by systems that definitively do consume it.
- A working sitemap and sane robots.txt. Basic, boring, and load-bearing.
- Clear headings that match real questions. This is what actually gets passages extracted.
- Answers stated plainly near the top of the page. Consistently the strongest single factor in whether a passage gets pulled.
- Content that is actually specific. Vague copy does not get cited because there is nothing quotable in it.
None of these are speculative. All of them help traditional search too, which is why they are a better use of the same twenty minutes.
So should you add one?
Yes, if you understand what you are buying. It costs almost nothing, it carries no risk, and if consumption does materialize you are already set up. Treat it as a cheap option on a possible future, not as an optimization.
No, if adding it would come at the expense of the fundamentals above, or if you would then report to yourself that AI visibility is “handled.”
If you do add one, do it properly
- Put it at the root: yourdomain.com/llms.txt.
- Write it in plain markdown. A short description of what the business does, then a linked list of your most important pages with one line of context each.
- Keep it to genuinely important pages. A dump of every URL defeats the purpose.
- Update it when your site structure changes, or do not bother having one.
# Primo Collab
Web development studio: WordPress, Shopify, SEO, CRO, and maintenance
for small and mid-sized businesses.
## Services
- /services/web-maintenance-security: Ongoing maintenance, security, and support
- /services/conversion-rate-optimization: Conversion audits and implementation
## Key resources
- /insights: Practical guides on web development, SEO, and site performance
- /faq: Answers to common client questionsThe mistake to avoid
Adding llms.txt and treating AI visibility as done. The file is a twenty minute task with an uncertain payoff. Writing pages that answer real questions clearly, marking them up properly, and making sure they load fast is a longer task with a payoff you can actually observe in your Search Console data. Do the second one first. Add the file afterward if you want the option.