LLMs.txt files, an increasingly popular method to improve visibility in AI search, may not be as effective as previously thought.

After reviewing over 300,000 domains, a new analysis from SE Ranking found that LLMs.txt files have no noticeable effect on visibility in major LLMs. 

What Are LLMs.txt Files?

LLMs.txt files are files uploaded alongside your website with the intention of helping AI tools understand your domain and more easily navigate or index your site. In recent months, these..txt files have begun to catch on as a relatively easy way to optimize your site for major AI tools.

They are similar to robots.txt files, which are widely used to optimize websites for search engines like Google or Bing. 

While robots.txt files are well-established as an effective and simple SEO tool, it has been unclear whether any major LLM has begun using LLMs.txt files when indexing websites. 

Slow Adoption

As a new optimization strategy, LLMs.txt files have seen relatively slow adoption. According to SE Ranking’s data, just 10.13% of domains crawled for the study had implemented these files as an optimization strategy. 

Part of the reason for the slow adoption of these files is the lack of research or clear data showing their impact.

Notably, the sites that were most likely to use these files were mid-tier websites likely looking for any edge against high-traffic competition. 

Data Says LLMs.txt Files Are Not Tied To Visibility

For SE Ranking’s study, a team of researchers analyzed how frequently domains were cited across responses from several popular LLMs. 

Ultimately, their data was unable to provide any evidence that the strategy had any real effect, even when using alternate research models to compare.

While the study doesn’t outright discourage people from implementing LLMs.txt files, the researchers conclude fairly bluntly that LLMs.txt “doesn’t seem to directly impact AI citation frequency. At least not yet.”

Robots.txt Is Still Used – Even By AI

While LLMs.txt files had been gradually gaining popularity, the reality is that none of the major AI tools have indicated they use these files when crawling or indexing websites. 

Aside from limited data suggesting GPTBot sometimes fetches LLMs.txt files, there is no evidence of any major LLM using or recommending implementing these files. 

Instead, OpenAI encourages websites to focus on robots.txt files and Google seemingly defers to these types of files when crawling or indexing pages. 

For more about the analysis and how it was conducted, read SE Ranking’s full report here.

Automated tools and audits can be essential parts of your online marketing toolkit when used to get a more complete view of your site, but Google’s Martin Splitt cautions sites from relying too heavily on these tools. 

In a recent Search Central Lightning Talk, Splitt – a member of Google’s Search Relations team – emphasized the importance of using context to make the most effective choices for your site. 

As Splitt said in the video:

“A technical audit, in my opinion, should make sure no technical issues prevent or interfere with crawling or indexing. It can use checklists and guidelines to do so, but it needs experience and expertise to adapt these guidelines and checklists to the site you audit.”

Specifically, Splitt laid out a three-step process he encourages sites to use to best assess a website’s performance and its needs. 

How To Audit Websites According To Martin Splitt

Splitt breaks down his website auditing process into three parts:

  1. Understand the Website
  2. Identify Potential Issues
  3. Make Sense of The Data

Before you ever start auditing a website directly, Martin Splitt says the first step is to take a moment and be sure you fully understand the site and its technical needs. This includes answering questions like what the website does, what technology it is built on, what sort of technical resources it uses, and how it is structured. 

This is essential to being able to effectively use tools later in the process to identify problems and spot the most effective areas for improvement. Without this, you may not even be able to identify the most effective tools to use for your site’s specific needs.

Once you have a broader view of a website, you can then start digging into data using relevant tools and analytics. Since you already understand the technical structure and needs of a site, you will be able to more effectively and efficiently understand the data these tools provide, filter out what isn’t relevant for your site, and hone in on the data that’s most important. 

Real-World Examples of Misleading Tools

Splitt used a real-world example to show how tools can be misleading if you don’t have a proper understanding of a site you are evaluating. He pointed to a high-number of 404 errors within a website as something that many tools will flag as an issue, even though they are often a normal consequence of website maintenance:

“A high number of 404s, for instance, is expected if you removed a lot of content recently. That’s not a problem. It’s a normal consequence of that. But if you have an unexplained rise in 404 responses, though, that’s something you want to point out and investigate…”

Context Matters

Ultimately, Splitt’s message is to rely on context to understand the tools you use online, rather than blindly relying on tools to understand your site. 

While tools can be hugely useful to get the best view of your site’s performance, they are unable to understand the unique details of your site and may not always have the best advice for your specific website.

As Martin Splitt stressed:

“Please, please don’t follow your tools blindly. Make sure your findings are meaningful for the website in question and take the time to prioritize them for maximum impact.”

Watch the full video from the recent Search Central Lightning Talk below: