You Don't Need an llms.txt File. Here's What You Do Need.

Priya runs a four-person digital agency in Jaipur. In April she sold a logistics client a "future-proofing" package whose headline item was a text file. The client, a freight firm called Marudhar, paid ₹45,000. The file took eleven minutes to write. When the client's founder asked her in July about llms.txt do you need it or was this a fashion, she did not have a good answer. She does now, and it cost her a refund.

What llms.txt Even Is

It is a proposed convention: a markdown file at your domain root that lists your most important pages in a simplified form, so a language model can read your site without wading through navigation and scripts. Think of it as a sitemap written for a reader rather than a crawler. The idea is reasonable. It is a community proposal, not a standard adopted by a search engine.

That distinction is the whole article. Anyone answering llms.txt do you need it with a five-figure quote is charging you for a convention, not for work. A proposal you can adopt in an afternoon is not a competitive advantage anyone can sell you.

llms.txt Do You Need It? What the Evidence Says

Google's AI optimization guide, published in May 2026 at developers.google.com/search/docs/fundamentals/ai-optimization-guide, addresses this directly. Its position is that AI features in Search draw on the same index and the same signals as ordinary Search, and that llms.txt, AI-specific content rewriting and special AI schema are not required.

What about other assistants? Honest answer: nobody has published a clean, reproducible result showing the file changes citation rates. Absence of evidence is not evidence of absence. It is, however, a very good reason not to bill for it as a ranking service.

Why It Keeps Getting Sold

Because it demos beautifully, and because llms.txt do you need it is a question with a satisfying yes-shaped answer available for sale. It is a file. You can show it to a client. It has a URL. Compare that with "we improved your server response time and rewrote nine pages", which is real work and photographs badly.

Five Things to Do Instead

What Marudhar Freight Did With the Refund

Priya put the ₹45,000 back and rebuilt the scope. Illustrative figures, but the shape is real.

Marudhar had a rate calculator that rendered entirely client-side, so its lane pricing existed nowhere in the HTML. They had one page titled "Services" carrying nine service lines. Their Google Business Profile listed a warehouse address that had moved in 2024. Three fixes, none of them exciting, all of them things a person searching for a Jaipur-to-Surat part-load quote would benefit from.

The founder still asks about llms.txt do you need it every couple of months, usually after a webinar. The answer has not changed. The file sits at their root now, added in ten minutes, unbilled, and nobody pretends it is doing anything.

When Adding It Is Actually Reasonable

There is a real case for it, and it has nothing to do with rankings. Asked narrowly enough, llms.txt do you need it does get a yes.

If you publish developer documentation, an API reference, or a large knowledge base, a curated index of your canonical pages is useful to your own tooling, to partners, and to anyone building an internal assistant over your docs. Maintain it as documentation hygiene. Budget it at the cost of writing a file, which is what it is.

The trade-off worth naming: a stale llms.txt is worse than none. If it lists pages you retired in March, you have published a wrong map of your own site. Nobody who sells the file also sells the maintenance, which tells you how seriously the deliverable is meant.

• Confirm nothing important is blocked in robots.txt
• Get prices, specs and locations into crawlable HTML
• One page, one question, answered fully
• Standard schema that matches the visible page
• Three independent mentions this quarter on sites you do not own

What to Tell the Person Who Read the Webinar Slide

The awkward part of this argument is rarely technical. It is that someone senior attended a session, wrote down a filename, and now expects to see it actioned. Telling them the whole idea is nonsense goes badly. Here is a version that works.

Start by agreeing. The instinct behind it is sound: make your important content easy for a machine to read. Then separate the instinct from the artefact. When a founder asks llms.txt do you need it, what they usually mean is "are we going to be invisible in AI answers", and that is a fair worry deserving a real reply rather than a lecture about specifications.

Second, put the file on the plan and price it honestly. Ten minutes, no retainer, no ongoing fee, reviewed whenever the site structure changes. Nothing is gained by refusing a cheap, harmless file. Everything is lost by billing for it as strategy.

Third, show what the same money buys elsewhere. A single blocked directory found and fixed. Four spec tables lifted out of PDFs. Two omnibus pages split into eight that each answer a real question. Those are demonstrable, and you can show a before and after. Whenever someone raises llms.txt do you need it in a planning meeting, that comparison ends the conversation faster than any argument about standards bodies.

One caveat worth conceding out loud, because pretending certainty here would be its own kind of sales pitch: the ecosystem is young, and a convention nobody requires today could be adopted tomorrow. That is exactly why the correct posture is a cheap file and an expensive focus on fundamentals. If the convention is adopted, you already have it. If it is not, you spent ten minutes. Asking llms.txt do you need it every six months is reasonable. Paying a retainer for the answer is not.