Two years ago nobody knew what a llms.txt was. Today, in 2026, it's the file that decides whether ChatGPT understands your business or ignores it entirely. And most websites still don't have one — or worse, have one that's doing nothing useful.
If you've been doing SEO for a while, this should sound familiar: when robots.txt showed up in the mid-nineties, it was considered a nerd's concern. A decade later it was mandatory. The llms.txt story is following the same arc, just compressed — what took robots.txt ten years is playing out in eighteen months.
The difference is that this time, arriving late doesn't cost you a ranking position. It costs you an entire channel — the one where ChatGPT, Claude, Perplexity and Gemini make recommendations to tens of millions of users every day.
What a llms.txt is (and what it isn't)
In plain language, a llms.txt is a cover letter your website leaves signed for AI models. It tells them who you are, what you do, what products or services you offer, who you serve, and which content on your site actually matters.
It's not a magic bullet. It's not SEO. It's not schema. It's something else entirely — complementary: a document designed so an AI model, crawling thousands of sites a day with a limited attention budget, can understand your business in thirty seconds without having to infer it from raw HTML.
And here's the first misunderstanding: having the file isn't enough. We've seen it dozens of times. Agencies that upload a generic template with the company name and five bullet points — then wonder why the results never materialize.
Who actually needs one in 2026
Not everyone needs a llms.txt with the same urgency. A personal blog with 20 daily visitors can live without one. An e-commerce store, a clinic, a professional practice, a restaurant, a real estate agency, a B2B brand, or any business whose typical customer opens ChatGPT to ask for recommendations — yes, you need one. Now.
It's especially urgent for sectors where search has shifted fastest toward AI assistants: tourism, hospitality, professional services, healthcare, education, and anything tied to purchase decisions with medium-to-high friction. Nobody asks ChatGPT where to buy diapers. Everyone asks it which architect to hire.
What happens in 2026 if you don't have a good one
Nothing dramatic overnight. What happens is silent and cumulative: your competitors start appearing when a potential client asks about your sector, and you don't. The AI models don't cite you. Not because they have anything against you — because they genuinely don't know what you do, or they have a vague impression and prefer to cite someone with a clearer identity.
In local markets this becomes very visible. We've seen clients go from zero citations to appearing by name in 60–70% of relevant prompts in their area within three months. And we've seen the opposite: technically impeccable websites, invisible in AI because no one ever explained to the models who they were.
Why doing it yourself rarely works
Let's be straight about this. There are llms.txt templates circulating on LinkedIn, Medium, and GitHub. Many are well-structured. But copying a template and filling in your details is like buying a guitar without knowing how to play: you have the instrument, you're missing the music.
A llms.txt that actually works isn't a technical document. It's an editorial decision. It requires having analyzed which prompts matter to your business, what your competitors are getting cited for, which content on your site deserves prominence, and how your value proposition is phrased in the language AI models are using to answer. That doesn't come from a template. It comes from an audit.
With Menorca Studio, the architecture practice, the llms.txt was more technical — the sector demands it — but no less considered. We paired it with sector-specific schema and a llms-full.txt with their flagship projects described in detail. Result: top position in ChatGPT and Perplexity for "architect in Menorca" and related queries.
Two signals that yours isn't working
The first is obvious: you've had a llms.txt live for months and you're not seeing any change in citations. If you've tested it against a real set of prompts — commercial intent, comparative queries, long-tail — and nothing moves, something is wrong.
The second is subtler: when you ask ChatGPT about your own sector in your area, the AI cites your competitors accurately but describes you in vague or incorrect terms. That's not bad luck — it's a llms.txt that isn't doing its job. It's like having a business card with a typo in your own name.
At A-Digital we work with clients in Menorca, across Spain, and throughout Europe. The pattern repeats everywhere: a well-crafted llms.txt isn't about length, it's about knowing what to put inside. What works for a brand in Mahón works, with the right adjustments, for one in Madrid or Munich.
Is your llms.txt getting you cited — or getting you ignored? We'll run a free audit and tell you exactly what changes, with real prompt examples from your sector.
Request your free audit →