A plain-text file, similar to robots.txt, that tells AI language models what’s on a website and which pages are worth reading. For agencies, it is a small, concrete step clients can actually take toward AI search visibility.
LLMs.txt is a proposed convention where a website places a plain markdown file at its root, yoursite.com/llms.txt, that summarizes the site’s purpose and links to its most important pages, written specifically for AI models to read rather than for human browsers or traditional search crawlers. The idea mirrors robots.txt, but where robots.txt tells crawlers what not to index, llms.txt tells language models what’s actually worth paying attention to.
Adoption is still early and not every AI platform reads it the same way, but it picked up real momentum through 2026 as AI crawlers started generating meaningful traffic to sites, in some cases more than traditional search crawlers. It’s not a guarantee a model will cite a site, but it removes a real obstacle: models generating long, poorly structured pages by guessing instead of being pointed to a clean summary.
Most AI-search advice right now is strategic and hard to measure. LLMs.txt is one of the few pieces of it an agency can actually build and ship this week.
It is a concrete deliverable for an ambiguous problem. Clients are asking what to do about ChatGPT and AI Overviews cutting into their organic traffic, and an llms.txt file, unlike most AI-search advice, is a file an agency can hand over and point to as done.
It forces a useful content audit. Writing a good llms.txt means deciding which pages on a site actually represent the brand accurately, which is a worthwhile exercise on its own, AI models aside.
It’s cheap insurance, not a silver bullet. It won’t outrank a competitor with better content or more authority, but it costs almost nothing to implement and there’s little reason not to have one once a site’s content is in reasonable shape.
A B2B software client asks their agency why ChatGPT keeps describing their product incorrectly when prospects ask about it. The agency audits the client’s site, finds the pricing and feature pages are buried under a confusing navigation structure, and builds an llms.txt file that points directly to the accurate, current versions of those pages. It’s not a guarantee the model updates its answers immediately, but it’s a five-hour task the agency can point to as a real step, sitting alongside the longer-term work of cleaning up the underlying content itself.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.