Published August 30, 2026 · TruePrime AI
The llms.txt specification has been live for over a year now. Thousands of businesses have created these files. But most of them aren’t doing what their owners think they’re doing — because the gap between “having a file” and “having a file that AI assistants actually use” is wider than it looks.
This post explains what happens on the other side: when an AI assistant visits your llms.txt, what it does with the information, and why the common mistakes make the entire effort pointless.
When ChatGPT, Perplexity, Claude, or another AI assistant encounters a question about a business category, it looks for authoritative sources to answer from. Your llms.txt file is one potential source. But here’s the part most guides skip: AI assistants don’t just read your llms.txt file in isolation. They’re cross-referencing it against everything else they know about you.
Think of it like a journalist fact-checking a press release. The press release (your llms.txt) provides the company’s version of the story. But the journalist also checks third-party sources, reviews, competitor coverage, and public records before deciding what to publish. AI assistants work the same way.
This means a llms.txt file that says “we are the leading provider of X” but has no corroborating evidence anywhere else on the web gets treated with the same skepticism a journalist would apply. The file is a signal, not a shortcut.
The most widespread mistake: treating llms.txt like a SEO keyword dump. Businesses stuff their files with keyword-dense marketing copy optimized for Google. AI assistants don’t process text this way. They parse for factual claims, structured information, and clear descriptions of what a business does, for whom, and where. A file written in clean, factual prose outperforms one stuffed with keywords every time.
A llms.txt file says “we serve dentists, lawyers, and contractors.” But the website has no service pages for those verticals, no case studies, and no schema markup confirming those service areas. The AI assistant has nothing to cross-reference — so the claim gets low confidence weight. Every claim in your llms.txt should have a corresponding page on your site that backs it up.
Many businesses created their llms.txt once and never touched it again. Meanwhile, they’ve changed pricing, added services, shifted focus. The file now contradicts their actual website. AI assistants notice contradictions — they reduce confidence in sources that disagree with themselves. If your site has changed, your llms.txt needs to change too.
Some llms.txt files run thousands of words, covering every possible topic the business might want to rank for. AI assistants have context windows, but they also have relevance filters. A sprawling file that covers everything equally ends up emphasizing nothing. The best llms.txt files are focused: they tell the AI exactly what the business does best, with specifics, in under 500 words.
The llms.txt specification is just one piece of the AI readability puzzle. A companion brand-facts.json file provides machine-readable structured data — pricing, locations, services, hours — that AI assistants can parse directly. Businesses that have llms.txt without brand-facts.json are giving AI assistants a narrative without data. Both are needed.
When done correctly, llms.txt does three things:
Is it worth hiring someone to build and maintain your llms.txt? That depends on two things: how important AI search visibility is to your business, and how much you’re willing to learn about the technical details.
For businesses that get most of their customers from online search — dentists, lawyers, contractors, real estate agents, local service providers — AI visibility is rapidly becoming as important as Google visibility. The businesses that set this up now are establishing themselves in AI assistants’ understanding before competitors do. That early-mover advantage compounds.
For businesses that rely primarily on referrals or don’t compete online, the ROI is lower. It’s still worth having a basic file, but the managed-service approach may be overkill.
The middle path: start with a basic llms.txt (our DIY guide walks through the setup in 30 minutes), then evaluate whether managed maintenance makes sense based on the results you see. For a detailed breakdown of what managed llms.txt services cost and what they include, see our llms.txt service guide.
Every TruePrime Go client gets a llms.txt file built and maintained as part of their plan. We also build a companion brand-facts.json with structured business data, and monitor both for consistency with the client’s live site. When pricing, services, or locations change, the files update in the same deploy cycle.
We run this on our own companies first. ShopProp Realty (est. 2007, 8 states) and AskBeforeYouEat both have llms.txt files maintained by TruePrime Go, which means the process is tested on real businesses before any client’s file is touched.
Plans start at $499/month, first payment after 30 days, cancel anytime. The llms.txt work is one piece of a full AI visibility program that includes AEO, content, and lead capture.
See PricingTruePrime Go · AI marketing that includes llms.txt, brand-facts.json, and full AEO optimization — $499/$799/$999/month, first payment after 30 days, cancel anytime.