Large Language Models (LLMs) like ChatGPT, Claude, and Perplexity are changing how people interact with the internet. You've probably noticed the AI-generated summaries or recommendations above the traditional search results when you're googling. They're helpful for quick answers, but for marketers this could result in an estimated 30% reduction in organic traffic.
This leads to two questions I've been asked repeatedly over the last few years: How do we get into the ai-generated result page and how can we control what it says about us?
While there are plenty of theories out there of how to get included in the results, we stand by putting out quality content that pushes the fold. If you can resonate with your target audience, you're definitely headed in the right direction.
As for how to control how LLMs perceive your site, the LLMS.txt is a newly proposed solution to accomplish just that.
What is LLMS.txt?
LLMS.txt is a plain text file that lives in the root directory of your website (at /llms.txt). It’s designed to help LLMs understand what your site is about, who it's for, and how you’d prefer its content be used or summarized. Think of it as a cousin to robots.txt, but aimed at AI models instead of search engine crawlers.
The format was first proposed by Answer.ai and has been lightly adopted by a few early AI companies including Anthropic. It’s not yet a standard, but it is gaining interest in the dev, marketing, and AI communities.
Why LLMS.txt Exists
Unlike traditional search engines, most LLMs don’t rely on a fixed crawl or centralized ranking system. Instead, they operate using pre-trained knowledge from a range of sources, including web snapshots, and in some cases, use real-time retrieval tools to fetch data on demand.
When a user asks a question, the AI system quickly pulls together the most relevant information it can find to answer. But there’s a limit to how much it can take in at once, so pages that are cluttered or too complex might get overlooked or misunderstood.
LLMS.txt addresses this by offering a clean and defined roadmap to the most important content on your site. It’s meant to help surface what your site is about, which pages matter most, and how you want your brand or services to be framed in AI-generated outputs.
What Can It Do?
While still early, they show potential to:
- Highlight key pages and prioritize content-focus areas
- Offer plain-language instructions to LLMs (e.g., “Focus on service detail pages, ignore cookie banners and nav”)
- Provide licensing or usage preferences for AI-generated summaries
- Clarify brand voice and target audience to help preserve positioning in outputs
In theory, if a model references your site for answers or brand mentions, this file could help shape how it describes your site in its response. If adopted more widely, it could become a new layer of control in AI visibility.
Here's a few examples that might help put this into perspective:
- Someone asks an AI chat how your pricing works or what your return policy is, you could link directly to markdown versions of those pages to ensure it is pulling clean and up to date responses rather than potentially blending data from other sources or outdated site content.
- A user searches for the best tech solution that can solve their business's problem. Your LLMs.txt could explain your audience, values and areas of specialization which could be surfaced to differentiate from your competitors or appeal to their specific niche.
- You notice your blog content or thought leadership content is being used in AI summaries without context or clarity. Your LLMS.txt could include content license language to set expectations for reuse, attribution, or fair use.
Why It’s Still Experimental
There’s no guarantee that models are actively using LLMS.txt for training or live query responses. While Anthropic has acknowledged it and some LLM directories are emerging on GitHub, major players like OpenAI or Google haven’t formally adopted it.
That said, it's a relatively low time investment and if this format does become more widely supported, early adopters could benefit from having already structured their sites accordingly. We wrote an article detailing how you can create one an llms.txt, add it to your site, and add it to the llms.txt directory.
What We’ve Seen So Far
At DMG, we’ve just started testing out the new file with ourselves and some of our clients. The process of creating and implementing a file is fairly straightforward and easy to replicate across Webflow and other CMS setups.
We’re closely monitoring how others are adapting and using it as well as any potential evidence that it is working in our favor. Since it’s still early, we haven’t seen any data yet that its working, but as with anything in SEO I imagine it just takes time.
The bottom line is that LLMS.txt may not change how your site is represented in AI tools overnight, however it's a low-risk step forward in preparing for the future. For marketers focused on SEO and brand representation, it offers a layer of structure and clarity that could prove valuable as AI continues to shape how content is discovered and summarized.
If you’d like help drafting or implementing your LLMS.txt file, reach out to our team.
