Google has now formalized its position on llms.txt in its official developer documentation. The position I took when I first published this is unchanged. The evidence has only gotten stronger. Updates throughout this piece reflect what we know now.
There's a lot of noise around llms.txt. I want to give you a more honest answer than most of what's circulating.
I actually have an llms.txt on my own site, and I'll explain why I kept it below. But I've watched too many clients chase shiny new formats while ignoring the things that actually drive visibility. So let me walk you through what this is, what it's worth, and when it actually makes sense to prioritize it.
What is llms.txt?
An llms.txt file is a plain markdown file that lives at the root of your website, usually at /llms.txt. The idea is to give AI systems a cleaner summary of your site and a curated list of pages they should pay attention to. Think of it as a simplified index written specifically for machines rather than people.
It's a proposed standard, not a formal requirement. No one is penalizing you for not having one.
Why did people start paying attention to it?
Because AI search and citation visibility became a serious topic in 2025 and 2026. The pitch makes intuitive sense: if AI tools need cleaner inputs to understand your site, maybe a cleaner file helps. The problem is the evidence hasn't caught up to the enthusiasm.
Here's where things stand now:
- Google has formally said it doesn't use llms.txt. Its May 2026 developer guide on optimizing for generative AI search tells site owners they don't need to create AI-specific files, markup, or markdown to appear in AI search results.
- The largest independent dataset on the question found no measurable effect. SE Ranking analyzed 300,000 domains and found that llms.txt had no statistical effect on how often a domain was cited by LLMs. When researchers removed the variable from their predictive model, its accuracy actually improved.
- Ahrefs went on record from the practitioner side. Its analysis of llms.txt noted that no major AI provider has committed to parsing the file, and server logs show major AI services don't even check for it.
- The counter-evidence is thin. A few practitioners have reported bot traffic increases after implementation, but no one has confirmed those bots actually fetched the llms.txt file or that the file caused the increase.
That doesn't make llms.txt worthless. It does mean we should be honest about what it is and what it isn't.
Do you need an llms.txt file right now?
For most sites, no. Not as a priority.
That doesn't mean it's useless. It means there's a difference between a tidy experiment and a must-have. Right now, it's closer to the former.
- Your core pages are already strong
- Your site is technically sound
- You have capacity to experiment
- Your site has deep docs, product libraries, or technical content
- You want a structured summary of your most important pages
- Weak or thin service pages
- Poor internal linking
- Crawl errors or indexing issues
- Missing schema where it actually matters
- JavaScript-heavy content search engines struggle to read
Those underlying issues are what actually limit AI visibility. An llms.txt file isn't going to paper over them. Missing or malformed llms.txt is one of several patterns I see on newly built sites that aren't ranking. The rest are usually more load-bearing: read about the others here.
What are the real pros?
It is genuinely easy to create
For a small to mid-sized marketing site, we're talking 30 to 60 minutes if your content is already organized. It's just markdown: a short summary of your business, your key pages, and some brief context about each. The exercise itself can be useful even before you publish it, because it forces you to articulate which pages actually matter and how you'd describe your brand to a machine.
It may become more useful over time
AI platforms are still evolving. The file that has no formal backing today could become something more meaningful in 12 months. Building it now, while it's low effort, means you're not scrambling later if the landscape shifts.
What are the real cons?
No strong proof it improves AI visibility
This is the honest headline. The 300,000-domain SE Ranking study is the strongest dataset to date, and it found no effect. Independent practitioner testing has also not found a correlation between adding llms.txt and improved citations, Share of Voice, or referral performance from AI platforms.
The biggest platforms have not backed it
Google has now formally said in its developer documentation that it doesn't use llms.txt. OpenAI, Anthropic, and Perplexity have not issued clear guidance endorsing it either. If the major platforms were treating it as a real signal, we'd have heard by now.
It can become a distraction and go stale
If you build it and never maintain it, the value drops even further. And if you're spending time on llms.txt while more important issues sit unfixed, that's the wrong order of operations. The fundamentals still matter more: strong pages, crawlable structure, good schema, and content that actually earns citations. The off-site equivalent is just as load-bearing: LinkedIn is the #2 most-cited domain in AI search, and that publishing layer is doing more for B2B AI visibility right now than any single file-level signal. If you want to understand what GEO and AI visibility work actually involves beyond file-level signals, that's worth understanding before you decide where to invest.
Why I still have one
Fair question, given everything above. The honest answer: it took about 45 minutes to write, it points to the same canonical pages I keep current everywhere else on the site, and if any AI-access standard ever does get formal adoption, I'd rather already have the file in place than scramble to build one. The argument is optionality, not performance. I don't measure it, I don't claim it drives citations, and if a client asked me to build them one as a standalone deliverable, I'd say no.
How do you implement an llms.txt file?
Homepage, about page, key service or product pages, high-value content, contact page, and documentation if you have it. Be selective. If everything is a priority, nothing is.
Site name, a short description of what you do and who you help, and grouped links to important pages with brief notes. Keep it honest and specific.
Place the file at the root of your domain so it resolves at https://yourdomain.com/llms.txt. Test that it loads correctly before moving on.
Add it to whatever process you use for site updates. When you add a major new page or change how you describe your services, update the file too.
Final take
If you have the time and your foundation is solid, go ahead and add one. If you don't, you're not behind.
llms.txt is a low-effort, reasonable experiment for sites that are already in good shape. It's not a shortcut, and it's not a substitute for the foundational work that actually drives visibility. Google has now formalized in its documentation what was already clear from the independent data: this file is not how you win AI search. Fix the important things first. Then add the nice-to-haves.
Frequently asked questions
An llms.txt file is a markdown file at the root of a website, usually at /llms.txt, that gives AI systems a simplified summary of the site and links to important pages.
Not right now for most sites. If your technical and content foundation is already solid, it's a reasonable experiment. If not, focus on the fundamentals first.
No, based on the largest independent study to date. Strong, crawlable, well-structured pages with good schema still do the heavy lifting. Google has also formally said in its developer documentation that it doesn't use the file.
Write a markdown file with a short site summary and links to your most important pages, then publish it at /llms.txt on your domain.