We've been hearing this question a lot over the past year: “We pay for SEO but never show up in ChatGPT, is that normal?” The short answer: yes, it's normal, because large language models read sites differently from Google. llms.txt came about to close exactly this gap: a free file that takes less than an hour to set up.
What is GEO and how is it different from SEO?
SEO targets Google's ranking algorithm: keywords, backlinks, page speed, user experience. GEO (Generative Engine Optimization) aims to be cited as a source in the answers produced by ChatGPT, Perplexity, Gemini and Google's own AI Overviews feature. The two don't conflict, they overlap: a clean, clearly written page works for you both on Google and in AI answers. But on the AI side there is one more thing: when a model visits your page, it has to filter the real information out of long, complex HTML. llms.txt shortens that filtering.
What exactly is llms.txt?
A plain-text file in Markdown format that you place in your site's root directory (such as digikeyword.com/llms.txt). What's in it: who the site is, what it does, and links with short descriptions of the most important pages. While robots.txt tells search engine bots “where you may go”, llms.txt tells language models “understand me quickly this way”.
What should an llms.txt file contain?
- A short identity statement: What the business does, in one paragraph, without jargon.
- A list of services/products: Each with a one-line description and a link to the relevant page.
- Important pages: Contact, about us, blog categories.
- What should be left out: Login forms, internal admin pages, duplicate content: leave out anything that would confuse the model.
How did we set it up on digikeyword.com?
When publishing this post we added an llms.txt to our own site as well. The logic is simple: we listed our three services (AI consulting, business automation, digital marketing), the founder's details and the links to our blog posts in short lines. Bullet points instead of long sentences, one piece of information per line. When a language model reads this file, it can answer all three questions, “what does DigiKeyword do, who do I ask, how do I get in touch”, in a single pass.
Is llms.txt alone enough?
No, and this is the point most “GEO guides” skip. llms.txt is a summary card; the real content still has to be on your pages. Three more things matter:
- Schema.org markup (Organization, Service, FAQPage): labels the content for the model as “this is a service description”, “this is a question and answer”.
- Content that answers directly: asking the question and writing the clear answer right below it works better than writing a three-paragraph introduction and getting to the real answer in the fourth.
- A consistent entity definition: let your business name, service description and contact details appear with the same wording everywhere on the site; the model finds contradictory information untrustworthy.
FAQ
Does an llms.txt file replace SEO?
No. It doesn't replace classic SEO (title tags, meta descriptions, sitemap, backlinks); it complements it. Google's ranking algorithm keeps working by its own rules, and llms.txt only helps AI models summarize the site faster.
Without llms.txt, can ChatGPT not see my site at all?
It can: ChatGPT and similar tools can already crawl the web. llms.txt makes that reading easier and tells the model directly which pages really matter; it isn't mandatory but it is a low-cost advantage.
How often should I update the llms.txt file?
There's no fixed schedule. It's enough to go over the file whenever you add a new service, an important page or a blog post.
If you're not sure your site appears in AI search engines
Let's review your processes and your site together and show where you fall short, both on Google and in AI search engines.