Klairia Blog

GPTBot Read Both Our llms Files. Here Is What That Proves.

OpenAI's GPTBot fetched Klairia's llms.txt and llms-full.txt within eight seconds. Here is the evidence, what it means, and what it cannot prove.

GPTBot Read Both Our llms Files. Here Is What That Proves.

We have been monitoring who requests Klairia’s /llms.txt and /llms-full.txt files. First, we saw Googlebot reach the full file. Then it requested the shorter file too.

Now we have a different signal: GPTBot, classified as an AI crawler operated by OpenAI, requested both files within eight seconds.

At 12:34:49 UTC on July 15, 2026, GPTBot requested /llms-full.txt. The server returned 200 with 3,108 bytes. At 12:34:57, it requested /llms.txt. That request also returned 200, with 625 bytes.

These were the only two static-file access events captured in that 24-hour window.

This is the first time our monitoring has seen an AI-specific crawler request Klairia’s llms files. It is stronger evidence of AI crawler discovery than the earlier search-crawler requests. It is still not evidence that the files changed an answer, produced a citation, or influenced how ChatGPT describes Klairia.

That boundary matters.

What we observed

The two requests were:

Time (UTC)PathResponseSize
12:34:49/llms-full.txt2003,108 bytes
12:34:57/llms.txt200625 bytes

Both events were classified as:

  • Bot name: GPTBot
  • Operator: OpenAI
  • Category: ai_crawler

The raw user agent identified GPTBot/1.4 and linked to OpenAI’s GPTBot information page. The file order was full first, short second.

The timing makes the pair notable. It suggests the crawler encountered both URLs during the same brief period. But timing alone does not reveal why it requested them, how it discovered them, or what it did with the responses.

Why this is different from a Googlebot request

Our earlier events came from Googlebot and were classified as search-crawler traffic. Those reads proved that a major search crawler could discover the files, most plausibly through the sitemap where both URLs are listed.

GPTBot is a different class of reader in our telemetry. It is an AI crawler, not a conventional search crawler. Seeing that category reach both files answers a question we could not answer from the Googlebot events: can an AI-specific crawler find and fetch the machine-readable context we published?

For this crawler, on this date, the answer is yes.

That does not mean GPTBot gave the files any special treatment. A crawler can request any public URL. The event proves access, not interpretation.

What the event proves

The evidence supports four narrow conclusions.

1. Both files were reachable

The two 200 responses confirm that neither file was blocked, missing, or failing when GPTBot requested it.

2. The short and full versions were distinct crawl targets

We saw one request for each path. That validates the decision to measure /llms.txt and /llms-full.txt separately instead of collapsing them into one metric.

3. An AI-specific crawler reached the files

The bot name, operator, category, and raw user agent all pointed to GPTBot. This is direct server-side access evidence, not an inference from referral traffic or an external indexing tool.

4. The monitoring loop works

We captured the path, response, byte count, classification, and raw user agent for both requests. That gives us a repeatable way to compare crawler behavior over time.

What it does not prove

The same event does not tell us whether:

  • ChatGPT used either file in a user-facing answer.
  • OpenAI indexed, retained, or trained on the content.
  • GPTBot understands llms.txt as a special format.
  • The files improved Klairia’s mentions, citations, or recommendation rate.
  • Publishing the files caused the crawler to visit.

Those claims require different evidence.

A crawler request is an acquisition signal: a system reached the information. Indexing is a processing signal. A citation or accurate brand mention is an outcome signal. Measuring one layer cannot substitute for measuring the others.

The useful lesson for teams publishing llms.txt

Publishing an llms.txt file creates a new machine-readable surface. Without request tracking, though, you cannot tell whether that surface is being discovered.

At minimum, track:

  1. The exact path requested.
  2. The request timestamp and response status.
  3. The raw user agent.
  4. The bot name, operator, and category.
  5. Whether the short file, full file, or both were fetched.
  6. Repeat access over weeks, not just isolated daily events.

Then connect that crawl data to a separate visibility measurement: how often AI assistants mention the brand, whether they cite it, whether the description is accurate, and how those results change across models, prompts, and languages.

The crawler log tells you whether your context entered the front door. Visibility tracking tells you whether the brand appears in the room.

What we will watch next

One paired request gives us a useful milestone, not a trend.

The next meaningful signals are repeat GPTBot visits, changes in which file it requests first, other AI crawlers reaching the same URLs, and any measurable movement in Klairia’s downstream AI visibility.

We will also keep the short and full files accurate. A crawler returning to stale or inflated claims would be worse than no machine-readable summary at all.

For now, the conclusion is precise: GPTBot requested both of Klairia’s llms files, and both requests succeeded. The AI crawler found the context. What it did after that remains an open measurement question.

FAQ

Does this mean ChatGPT read Klairia’s llms.txt?

No. It means GPTBot requested the file. A crawler fetch is not evidence that a specific ChatGPT response used the content.

Does this prove OpenAI supports the llms.txt format?

No. The crawler requested two public URLs, but the event does not reveal whether it interpreted either file as a special standard.

Why did GPTBot request both files?

We do not know. Both URLs are public and included in Klairia’s sitemap, so ordinary URL discovery is a sufficient explanation. The eight-second timing shows proximity, not intent.

Is this stronger than a search-crawler request?

It is more directly relevant to AI crawler discovery because the reader was classified as an AI crawler. It is not stronger evidence of downstream visibility, citations, or answer quality.

What should happen after publishing llms.txt?

Instrument the files, monitor who requests them, keep the content accurate, and measure actual AI answers separately. Discovery and visibility are connected, but they are not the same result.