Does your landing page need an llms.txt file?

Since September 2024, a file with a name almost identical to an old web standard has been showing up in SEO and GEO recommendations: llms.txt, often pitched as "the robots.txt for generative AI." The promise is appealing — give ChatGPT, Claude or Perplexity a condensed, reliable version of a site instead of leaving them to guess from raw HTML. Before adding one to your landing page, it's worth knowing what it actually does, what it doesn't, and what the little available data says about its real-world adoption.

What llms.txt actually is

The format was proposed by Jeremy Howard, co-founder of Answer.AI and fast.ai, in response to a real problem: a language model's context window is too small to ingest an entire site, and turning HTML cluttered with menus, scripts and layout into usable text remains slow and error-prone. His proposal: a markdown file placed at the site's root (/llms.txt), with an H1 title, a short summary as a blockquote, then sections listing key pages in plain language, stripped of a web page's visual noise. The idea isn't to replace the landing page's content, but to offer a "shortcut" a model can read directly — a technical complement to the on-page work covered in our guide to GEO visibility.

What it isn't

The name invites confusion, but llms.txt is nowhere near an official web standard: it isn't ratified by the W3C or the IETF, unlike the robots exclusion protocol that has governed robots.txt since 1994. It isn't a sitemap.xml either — it has no effect on classic Google or Bing indexing — nor a blocking mechanism: adding a page to it neither forces nor prevents any bot from reading it. It's a voluntary convention, adopted on the publisher's side, and read — or not — just as voluntarily by the AI bots on the other end.

What a study on real-world compliance with crawler conventions shows

That last point is exactly where caution is warranted. A study by Taein Kim and coauthors, presented at the 2025 ACM Internet Measurement Conference, tracked the behavior of 130 self-declared bots over 40 days through controlled robots.txt experiments — a three-decade-old standard, widely documented, that virtually every crawling tool knows how to read (study on Google Scholar). The result: compliance with directives is selective, less frequent when the directives are restrictive, and some categories of bots — including certain AI-search-related crawlers — don't even check the file before crawling. If a standard as old and standardized as robots.txt is only followed to a variable degree, there's little reason to expect an informal convention published in 2024, with no standards body behind it, to get more rigorous treatment.

Adoption is rising, but real usage stays marginal

On the ground, several traffic analyses published in early 2026 point the same way: the number of sites publishing an llms.txt is growing — roughly one in ten among the domains those analyses track — but the actual volume of requests AI bots make to that file stays marginal compared to their overall crawling activity. Even Anthropic, which documents llms.txt on its own site, notes that Claude doesn't rely on it systematically for web search or training: the most concrete use case today sits on the side of coding assistants (Cursor, Claude Code) that use it to quickly ingest a project's documentation — a use case close to technical documentation, and fairly far from a standalone marketing landing page.

The real lever is still the content of the page itself

A landmark study on citability by generative engines, led by researchers from Princeton, Georgia Tech and the Allen Institute for AI, measured what actually gets content cited by an AI (study on Google Scholar): verifiable facts, named quotes, clear structure directly in the rendered HTML — not the presence of a separate manifest aimed at bots. In other words, a well-maintained llms.txt doesn't make up for a page whose content stays vague or poorly structured; at best it complements work already done on the page, never replaces it.

So, should you add one to your landing page anyway?

For a standalone marketing landing page — an ebook sales page, a SaaS waitlist, a coaching page — llms.txt isn't a priority: most AI bots don't yet consult it reliably, and the same effort is better spent on the practices covered in our GEO guide, or on properly implemented Schema.org structured data, which is far more widely used today. That said, if it only takes a few minutes to set up and doesn't conflict with anything in classic SEO, there's no reason to skip it either — particularly if the same domain hosts product documentation or a technical blog that a coding assistant might read.

How to create one in a few minutes on a Next.js site

On an App Router project — like the templates LanderKit sells, which already generate their robots.txt and sitemap.xml dynamically — the simplest approach is a dedicated route handler, app/llms.txt/route.ts, returning text/plain with markdown content generated on the fly rather than a static file sitting alone in public/. The content itself should stay minimal: a title, a one- or two-sentence summary as a quote, then three to five links to the pages that actually matter — the homepage, the main offer page, the contact page — written in plain language rather than marketing copy. This isn't the place to paste in the hero's headline.

Mistakes to avoid

• Treating it as an SEO priority when Google and Bing, which drive most organic traffic, don't rely on it for indexing.
• Letting it go stale — a price, an offer or a page that no longer exists, quoted as-is by an AI, hurts trust more than having no llms.txt at all.
• Blocking AI bots in robots.txt while hoping they'll read the llms.txt sitting next to it: the two intentions directly contradict each other.

A concrete case

For a technical waitlist page like the SaaS Waitlist template (demo), aimed at developers and product decision-makers, adding a minimal llms.txt costs little and fits the target audience. For most of the ten other profiles in the catalog — coaching, ebooks, real estate, restaurants — that same time is better spent polishing the FAQ, named testimonials and verifiable facts on the page itself, exactly what our GEO guide covers. The 10 LanderKit templates ($89 each, $229 for the full pack) ship the complete source code for every section: adding an llms.txt route or enriching an FAQ happens directly in the project, with no intermediate back office.