TL;DR
The llms.txt questions started landing in our inbox in late 2025, and they're still coming in.
Clients forward audit alerts, partner agencies ask what we're doing, our own team asks what to say on the next client call.
Source: StudioHawk Slack and HubSpot email records, October 2025 to November 2026. Names and client details removed.
This week a client's in-house SEO asked us for one, along with case studies showing it works.
We went looking and couldn't find a single one.
We gave the answer we've given since 2025: it's fine to put on a website, it doesn't hurt, and it doesn't cause any significant uplift.
Add it if it takes minutes, then focus on the bigger fish to fry.
Start with what the file is.
llms.txt is a Markdown file at the root of a website that lists the pages an AI tool should read first.
Jeremy Howard proposed it on September 3, 2024. He wanted a way "to provide information to help agents use a website."
The file holds the site name as a heading, a one-line summary, then lists of links with a note after each.
It's still a proposal, and no search engine or AI company has adopted it as a standard.
The proposal's own site says the files "are used most heavily for software documentation, where coding agents follow them."
Our Hawk Academy glossary entry on llms.txt walks through the format line by line.
The more useful question is how a file built for coding tools ended up in marketing audits.
Clients ask for an llms.txt for one of four reasons. We've seen all four in our client conversations since mid 2025:
All four start from how a site looks to a tool or a rival.
None of them starts from a result someone measured.
Attention ran two years ahead of the evidence.
Tools, platforms and brands adopted llms.txt one after another until it looked like a standard.
No search engine or AI company ever said it reads the file.
Source: llmstxt.org, Search Engine Journal, Chrome for Developers, Shopify developer changelog and Google Search Central (guide updated July 10, 2026). Dates checked November 3, 2026.
Google said no the whole way through. In April 2025, John Mueller compared llms.txt to the keywords meta tag, a tag Google has ignored for ranking since at least 2009.
Even the Lighthouse check marks a missing file as not applicable, because providing it is "optional at the moment".
The evidence on what the file does for search and AI answers points the same way.
No source we can find shows llms.txt changing rankings, citations or AI traffic. Here is the evidence, strongest first.
Source: StudioHawk analytics data, 394 websites, October 2025 to September 2026; llms.txt checked November 2, 2026. Full method in the Hawk Academy guide.
Shopify and the proposal's authors keep changing the file anyway, because its job is changing.
llms.txt is turning into instructions for AI agents, the tools that browse and buy on a person's behalf.
Two changes in 2026 pushed it there.
First, version 2 of the proposal, published in August 2026, lets one site carry several files. In its words, "a file covers the pages under its path, and the most specific file applies."
That suits large documentation sites, where a coding agent needs the right section and nothing else.
Second, Shopify's default file stopped being a link list.
A Shopify store's /llms.txt now serves the same content as its /agents.md, a file of instructions for shopping agents.
Shopify's changelog of May 28, 2026, says the paths "/llms.txt and /llms-full.txt also point to this content by default."
You can see it on our client Sylvane's store.
Its llms.txt opens with the heading "Agent Instructions" and reads like a manual for a shopping agent: how to search products, build a cart and check out with Shop Pay.
Most clients assume the file tells AI who the brand is, this one mostly tells an agent how to buy.
Shopify store owners should open yourstore.com/llms.txt before writing anything.
You almost certainly have one: 121 of the 125 Shopify stores in our sample served a file.
Keep any edit small, and leave the agent instructions working.
Writing an llms.txt takes minutes. Getting it served is where the time goes, we've planned or shipped the file on more than 20 client sites.
Each time, the hard part was access:
That last one is ours, our own llms.txt is built as a HubSpot page, so HubSpot serves it as text/html.
The proposal names no content type, no AI company documents reading the file at all, so none says what format it needs.
The file still loads, and nothing we can measure depends on it.
Keep the file accurate if you run one.
We've had to strip old company names and retired URLs out of client files, and out of our own.
That upkeep matters less than the fixes below.
Fix the things AI systems demonstrably read before you touch llms.txt.
Our short list:
None of this is new. It's the same groundwork that gets you into Google's AI Overviews. Of the six, the AI info page needs the most explaining.
An llm-info page is worth building if AI tools describe your business wrongly. It's an ordinary HTML page of plain facts: what you do, who runs it, where you operate. AI crawlers read it under the same robots.txt rules as any other page.
That's why it gets read.
We tested one in September 2025 by placing a unique marker on a client's AI info page, then asking ChatGPT about the brand. ChatGPT reached the page through a live web search, because the page ranked for the brand name. It worked because it was a normal page that ranked. No special file was involved.
Ours sits at studiohawk.com.au/llm-info.
Put the same facts on your About page if AI tools already describe you correctly.
People read that page too.
The Hawk Academy glossary entry on llm-info pages covers what to include.
No. An llms.txt can't hurt your rankings in Google, because Google Search ignores the file. Google's guide says it "will neither harm nor help" your visibility in Search.
Yes. Shopify adds an llms.txt to stores by default, and since May 2026 it mirrors the store's agents.md, a file of instructions for shopping agents. Check yours at yourstore.com/llms.txt before building a new one.
No. You can't block AI bots with llms.txt, because the file has no rules for blocking anything. Use robots.txt, which OpenAI, Anthropic and Perplexity all document for controlling their crawlers.
No documented rule says it matters. The proposal names no content type, and the Lighthouse check only flags server errors. Ours serves as text/html through HubSpot, and the text inside is still plain Markdown.
Add the file if it takes minutes.
Then spend the real time on crawler access, starting with the free AI bot access checker, which shows which AI crawlers your robots.txt lets in.
For the wider picture, download our AI search whitepaper, or talk to us about AI SEO.