llms.txt — what it actually does for hotels.

A new category of hospitality marketing vendor has emerged in the last 18 months. The pitch sounds like this: "AI is replacing Google. Travelers are asking ChatGPT, Claude, and Perplexity where to stay. If your hotel isn't in those answers, you're invisible. We solve this by adding llms.txt to your website." The price tag ranges from a free generator to thousands of dollars a year. The premise is partially true. The proposed solution is largely not. This page exists because the gap between those two things is wide enough that hospitality buyers are getting hurt — paying for a technical fix that doesn't deliver the AI-search visibility they were promised, while the actual work that would deliver it stays undone. Here is the honest version.

What llms.txt actually is.

In September 2024, Jeremy Howard — founder of Answer.AI and a respected figure in the open-source AI community — proposed a simple convention: a Markdown file at the root of your domain (yoursite.com/llms.txt) containing a curated list of links to your most important content, each with a one-line description. The idea was to give large language models (LLMs) a structured entry point into your site that's easier to parse than crawling and de-noising the entire thing.

The file looks like this:

# Your Hotel Name

> A short description of your property — location, category, what makes it distinctive.

## About
- [About the property](https://yourhotel.com/about): history, ownership, design ethos
- [Location and access](https://yourhotel.com/location): airport distance, transit, parking

## Accommodations
- [Room types](https://yourhotel.com/rooms): suites, junior suites, deluxe rooms
- [Amenities](https://yourhotel.com/amenities): pool, spa, restaurant, business center

## Things to do nearby
- [Local guide](https://yourhotel.com/blog/local-guide): neighborhood overview
- [Seasonal events](https://yourhotel.com/blog/events): annual calendar

That's all there is to it. It's a clean, machine-readable summary of your site. It takes about 30 minutes to create. It has its uses. None of those uses are what most hotel vendors are selling.

What llms.txt is not.

It is not a standard. There is no W3C or IETF specification. There is no enforcement mechanism. There is no requirement that any AI system read it.

It is not — and this matters most — broadly used by the major AI platforms in production. As of mid-2026:

• OpenAI (ChatGPT, GPTBot): occasionally fetches llms.txt files but does not use them as a primary input to responses. OpenAI's documented recommendation for crawler control is robots.txt, not llms.txt.
• Anthropic (Claude, ClaudeBot): publishes its own llms.txt for the Claude documentation site (so coding assistants can navigate the API docs), but has not confirmed that Claude consumes other sites' llms.txt during web retrieval.
• Google (Gemini, AI Overviews): John Mueller has publicly compared llms.txt to the long-deprecated "keywords" meta tag — a vendor-controlled claim about a site's content that the crawler can verify directly from the page itself. Google's AI systems do not act on llms.txt.
• Perplexity: no public commitment to reading llms.txt as a citation input.

One independent study tracked 62,100 AI bot visits to a test domain over 90 days. Of those, 84 requests targeted the llms.txt file — 0.1% of all AI crawler traffic. Another study of 300,000 domains found that among the 50 most frequently AI-cited sites, only one had an llms.txt at all.

These are not anti-llms.txt data points. They are simply the current measurement reality: in mid-2026, AI systems are not measurably treating llms.txt as a meaningful input to the answers they generate.

What the hospitality vendors are selling.

A short, partial list of vendors currently pitching llms.txt as the path to AI visibility for hotels:

• INNsight — markets llm.txt as part of an "AI-ready" PMS and website platform: "we've handled the technical heavy lifting of llm.txt so you can focus on guest experience."
• Visito AI — offers a free llms.txt generator specifically for hotels, framed around making properties "discoverable by ChatGPT, Claude, Grok, and Gemini."
• Lighthouse (Connect AI) — bundles llms.txt generation into a broader "AI visibility" service ("Discoverable. Understandable. Bookable.").
• Hotelogix — generates an llms-medium.txt for clients as part of its hospitality software stack.
• Cloudbeds — published "The Signals Behind Hotel AI Recommendations" research and positions its platform around AI visibility, with llms.txt framed as one component.

None of these vendors are dishonest. llms.txt is a real thing. Generating one is a real service. The technical work is real.

What's misleading is the implication that adding llms.txt will make a hotel discoverable in AI search. The data does not support that claim. The major LLMs do not currently read these files in production. A hotel that pays for llms.txt implementation and nothing else will see no measurable change in AI search visibility, because the file is not the input the AI systems are using to generate hotel recommendations.

What actually drives AI search visibility for hotels.

When a traveler asks an AI system "where should I stay in [destination] for a couples weekend," the system performs roughly the same operation regardless of provider: retrieve candidate web pages, extract specific facts from them, synthesize a response, and cite the pages it pulled from.

The selection pressure that determines which hotels get cited happens at retrieval and extraction. Not at any file-based metadata layer. The systems pull from pages where:

• The hotel's identity and location are clearly stated in extractable prose
• Amenities, room types, and policies are answered in direct language ("Check-in begins at 3:00 PM" beats "We welcome guests with a warm afternoon greeting")
• The property is mentioned in editorial content about the destination — travel guides, "things to do" articles, neighborhood overviews
• Structured data (schema markup) confirms key facts in machine-readable form
• FAQ-formatted Q&A blocks let the system extract specific answers without paraphrasing
• Third-party content (reviews, travel coverage, local press) corroborates the hotel's positioning

None of those signals come from llms.txt. All of them come from the actual content on the actual pages of the actual site. The page is the input. The file is, at most, a table of contents pointing the system to the pages.

Own the content. Own the opinion.

Here is the framing that matters for hospitality marketers thinking about AI search visibility:

When an AI system generates a recommendation about your property, it is, in effect, forming an opinion. That opinion is synthesized from whatever the system found about you during retrieval. The character of the opinion — what gets mentioned, what gets emphasized, what gets left out, what tone the description takes — is determined by the content the system had access to.

If the only substantial content about your property online is the homepage and rooms page on your own site, plus some OTA listings, plus a few travel blog mentions, the AI's opinion about your hotel will reflect that thin source base. You will be described in generic terms. You will be confused with similar properties. You will be left out of specific recommendations because there's nothing to anchor the system to your differentiation.

If your site instead contains:

• A detailed property description in extractable prose
• FAQ-marked answers to every common guest question
• Long-form content about your destination, your neighborhood, your seasonal programming, your nearby attractions
• Editorial pieces describing the property in the voice and detail you want AI to echo
• Specific, factual, schema-marked information about location, amenities, pricing range

...then the AI's opinion about your property reflects your framing, in your language, at your level of detail. The system has no choice but to synthesize from what it found. You control what it finds by controlling what you publish.

This is what "own the content, own the opinion" means in the context of AI search. The opinion belongs to the property that produced the source material the AI extracted from.

A hotel relying on llms.txt as its AI visibility strategy is hoping the AI will read a 30-line index file and form a useful opinion from it. That is not how these systems work. A hotel publishing 100+ pieces of substantive content about itself and its destination is giving the AI a deep well to draw from. Those are completely different positions.

What llms.txt is actually useful for.

None of this is to say llms.txt is worthless. It has real uses, just not the ones being marketed to hotels. Specifically:

What it is not useful for: making your hotel get cited by ChatGPT, Claude, Perplexity, or Google AI Overviews when travelers ask about destinations. That outcome requires substantive content on your site, not a metadata file at your root.

How to actually implement llms.txt (if you want one).

Since the file has legitimate uses and costs almost nothing to ship, here is the honest implementation guide.

The file structure.

An llms.txt file is plain Markdown, saved at the root of your domain. The recommended structure:

1. An H1 title with your business name
2. A blockquote with a one-paragraph description of what you do
3. Optional additional context paragraphs
4. One or more H2 sections grouping related links
5. Under each section, a bulleted list of links with one-line descriptions

Hosting and verification.

The file must live at https://yourdomain.com/llms.txt — root of the domain, not in a subdirectory. After uploading:

• Open https://yourdomain.com/llms.txt in a browser. You should see the raw text content.
• Confirm the file is served with Content-Type: text/plain or text/markdown (most static hosts handle this automatically).
• Reference it in your robots.txt for discoverability: LLM: https://yourdomain.com/llms.txt (an informal convention; not all crawlers honor it).

Maintenance.

Update the file when you add significant content, restructure pages, or change descriptions of your offerings. A stale llms.txt pointing to dead URLs is worse than no file at all. Quarterly review is a reasonable cadence for most hospitality sites.

The optional llms-full.txt.

Some sites also publish a companion file called llms-full.txt — the entire Markdown content of every linked page, concatenated into one file. This lets AI systems ingest the whole site in a single fetch. Useful for documentation-heavy sites; overkill for most hospitality sites. Skip it unless you have a specific use case.

The Digital Fox llms.txt.

We practice what we recommend. Digital Fox publishes an llms.txt at digitalfoxllc.com/llms.txt. It's a curated index of our most important pages — homepage, case studies, the insights essays we want AI systems to associate with our category positioning — each with a one-line description.

It does not — and we don't claim it does — make us magically appear in AI search results. What makes Digital Fox appear in answers to "hospitality SEO consultancy" or "Generative Engine Optimization for hotels" is the 13+ long-form essays on our blog covering those topics in genuine depth, the case studies with real GA4-attributed numbers, and the consistent terminology we use across every page on the site.

The llms.txt is a small accessory to that work. It would be misleading to position it as the work itself.

What to ask any vendor pitching you AI visibility.

If a hospitality vendor is selling you an "AI visibility solution" centered on llms.txt, ask these four questions before you sign anything:

The bottom line.

llms.txt is a useful, low-effort, forward-compatible piece of metadata. It belongs on a serious hotel website. It takes 30 minutes to create. It costs nothing to maintain.

It is not, however, the AI visibility strategy. It is one small element of a strategy whose actual mechanics are:

• Substantive long-form content covering your destination and property at the depth and specificity AI systems can extract from
• Technical SEO and schema markup that helps AI systems identify and categorize your property
• Editorial content that establishes the framing you want AI systems to echo when describing you
• Third-party signals (reviews, press, citations) that corroborate your positioning
• A coherent site architecture that makes the relationships between your pages clear

Vendors selling llms.txt as the answer are skipping the work. The work is harder, slower, more expensive, and dramatically more effective. There is no 30-minute fix that produces AI search visibility. There is only the content infrastructure — built over months, maintained over years — that gives AI systems the substantive source material they need to recommend your property accurately.

Own the content. Own the opinion. The file at the root of your domain is not a substitute for either.

If you've been pitched an AI visibility solution and want to know whether it's the real work or the cosmetic version, our audit will tell you — including whether your existing content infrastructure is what AI systems can actually extract from, and what specifically would move the needle for your property.