Key Points
- llms.txt is a text file designed to show AI systems which parts of your website are important.
- Google does not use it. Gary Illyes (Google) publicly confirmed this in July 2025. John Mueller compared the file to the long-defunct keywords meta tag.
- The major AI crawlers rarely access the file. Server log analyses from 2026 show that GPTBot, ClaudeBot, and PerplexityBot rarely or never request it—though there are a few exceptions.
- You currently don’t need it for greater visibility in AI searches. What really matters: clean technology, structured data, and content with real added value.
llms.txt: What the File Promises—and What It Doesn't
Maybe you’ve heard the tip before. “Create an llms.txt file, and ChatGPT will find you better.” Sounds simple. Sounds like a quick win.
The problem behind this is subtle but costly. You’re investing time in a file that hardly any AI system accesses. And that’s exactly the time you’re missing out on when it comes to the levers that actually work.
This is exactly where it’s worth taking a closer look at the facts. Because the file currently doesn’t do what many promise.
What llms.txt actually is
llms.txt is a simple text file. It’s located in your website’s root directory, specifically at deine-domain.de/llms.txt.
The idea behind it: You list your most important pages and content in it. This is supposed to help an LLM (Large Language Model—the technology behind ChatGPT, Gemini, and similar systems) understand what your site is about more quickly.
The proposal comes from Jeremy Howard. He published it in September 2024. Important: It is a community proposal, not an official standard. Bodies such as the IETF or W3C (the organizations that set web standards) have not adopted llms.txt.
Where the Hype Comes From
The logic sounds appealing. There are robots.txt for search engines—why not a file for AI?
Then there’s the pressure. AI answers provide direct responses to many questions. Clicks on websites are declining. Anyone who hears about a simple solution is eager to try it.
The problem: A nice idea isn’t necessarily an effective measure. And that’s exactly what llms.txt demonstrates very clearly.
Does Google use llms.txt? The honest answer
No. And Google says so openly.
Gary Illyes, one of Google Search’s most well-known representatives, clarified in July 2025: Google does not support llms.txt and has no plans to do so. The file plays no role in inclusion in AI Overviews (the AI summaries at the top of Google Search) or AI Mode.
John Mueller from Google was even more explicit. He compared llms.txt to the keywords meta tag. That used to be a field where you could enter your keywords. Search engines have been ignoring it for over ten years—because the site owner fills it in themselves, making it easy to manipulate. This exact point also applies to llms.txt.
There is a curious side note. On December 3, 2025, an llms.txt file suddenly appeared in Google’s own developer documentation. The SEO world was puzzled. Mueller commented laconically with “hmmn :-/”. The file had disappeared again by the end of the day. The trigger was an automatic CMS update across many Google properties—the Search team was not involved and removed the file from its docs. Mueller later clarified: This was not an endorsement.
What the data says
You don’t have to believe it just because Google says so. You can look at the numbers. Three independent findings paint a clear picture.
| Question | Finding | Source |
|---|---|---|
| How many websites use llms.txt? | About 10% of 300,000 domains analyzed — the largest sites are the ones that avoid it the most | SE Ranking, November 2025 |
| Does the file bring a measurable increase in traffic? | 8 out of 10 sites tested: no measurable change | Search Engine Land, January 2026 |
| Do AI crawlers even access the file? | Rarely and with no measurable effect. GPTBot appears only sporadically in server logs; no correlation with AI citations could be established | SE Ranking, November 2025 |
The last point is the decisive one. A file that hardly any crawler retrieves can have little effect. Crawlers (programs that automatically scan websites) continue to read your HTML code directly—not an additional file.
OpenAI also refers in its official crawler documentation to robots.txt and static IP addresses. llms.txt does not appear there.
What really matters instead
Here’s the good news. You don’t have to chase after every trend. You probably already know the levers that actually work—and they’re solid best practices.
- Clean crawler control via robots.txt. This file is actually read. Here you specify which bots are allowed to access which areas. That’s the real lever, not a wish list.
- Structured data with Schema.org. FAQ, Article, and Organization markup (technical markup of your content) helps machines classify your site correctly. This applies to Google just as much as it does to AI systems.
- Content with real added value. Google has been saying the same thing for months: What gets you ahead in classic search also gets you ahead in AI answers. Firsthand experience, clear facts, verifiable evidence.
- Technical foundation. Fast loading times (Core Web Vitals), clean page structure, correct language markup. First the foundation, then visibility.
To understand what an AI actually reads from your website—and what it doesn’t:
| Action | Used by AI crawlers? | Effect |
|---|---|---|
| llms.txt | Hardly ever—rarely accessed, no documented effect | Currently no measurable visibility |
| robots.txt | Yes — is reliably read | Controls which bots see which areas |
| Schema.org | Yes — evaluated by Google and AI systems | Helps machines correctly categorize content |
Be honest with yourself: These four levers require more work than a quickly created text file. But they are the only ones with measurable results. That’s exactly what we check before recommending a measure to a client—and we decline if a tool won’t deliver results.
When llms.txt can still be worthwhile
Let’s be honest: llms.txt isn’t completely worthless. There is a useful application for it.
If you run a large documentation site or a technical help portal, the file can help an AI assistant navigate your site. Anthropic (the company behind the AI model Claude), for example, uses llms.txt in its own documentation.
But that’s different from “greater visibility in AI searches.” It’s about providing guidance for assistants working directly on your site—not about your ranking in major AI responses.
For the typical B2B SME, this means: The file doesn’t hurt. It’s quick to set up. But it doesn’t replace any of the four things listed above. Focus your time first on where it has a measurable impact.
Next Step
Want to know if your website even shows up in AI searches—without relying on hype tools? Then let’s take a look at your crawler management, your structured data, and your technical foundation together. Honestly, with no sales pressure.
Frequently Asked Questions (FAQ)
Was ist llms.txt?
llms.txt ist eine Textdatei im Wurzelverzeichnis einer Website. Sie soll KI-Systemen zeigen, welche Inhalte wichtig sind. Der Vorschlag stammt aus dem September 2024. Es ist ein Community-Vorschlag, kein offizieller Web-Standard.
Nutzt Google llms.txt?
Nein. Google hat im Juli 2025 öffentlich bestätigt, dass es die Datei nicht unterstützt und das auch nicht plant. Für AI Overviews und den AI Mode spielt llms.txt keine Rolle. Maßgeblich bleiben normale SEO-Signale.
Bringt llms.txt mehr Sichtbarkeit in KI-Suchen?
Dafür gibt es keinen Beleg. Ein Test mit zehn Seiten zeigte bei acht keine messbare Veränderung. Hinzu kommt: Server-Log-Analysen zeigen, dass die großen KI-Crawler die Datei kaum abrufen.
Soll ich llms.txt trotzdem anlegen?
Sie schadet nicht und ist schnell erstellt. Für große Dokumentationen kann sie KI-Assistenten Orientierung geben. Sie ersetzt aber keine saubere robots.txt, keine strukturierten Daten und keine guten Inhalte. Setz deine Zeit zuerst dort ein.
Was hilft stattdessen für KI-Suchen?
Vier Dinge: saubere Crawler-Steuerung über robots.txt, strukturierte Daten mit Schema.org, Inhalte mit echtem Mehrwert und ein technisches Fundament mit schnellen Ladezeiten. Das wirkt in der klassischen Suche und in KI-Antworten.
Sources
- https://llmstxt.org/ — Original proposal llms.txt (Jeremy Howard, starting September 2024)
- https://developers.google.com/search/docs/crawling-indexing/robots/intro — Google: robots.txt (what Google actually reads)
- https://developers.google.com/search/docs/appearance/ai-features — Google: how content appears in AI features
- https://searchengineland.com/no-llms-txt-is-not-the-new-meta-keywords-458199 — Analysis of public Google statements (Illyes, July 2025; Mueller), Search Engine Land