llms.txt is a proposed plain-text file, placed at a site’s root, that lists the pages a business wants language models to read. Vendors increasingly sell it as a shortcut to being cited by ChatGPT, Perplexity, and Google. The best large-scale evidence available says that promise is unproven, and Google says the file is not needed.
- SE Ranking’s analysis of about 300,000 domains found no significant correlation between having an llms.txt file and how often a domain is cited in LLM answers (reported by Search Engine Journal, November 2025).
- Only 10.13% of those domains had the file, and removing it as a variable made SE Ranking’s prediction model more accurate, not less.
- Google’s documentation states that no new machine-readable files, AI text files, or special markup are needed to appear in AI Overviews or AI Mode.
- The file is cheap to add, so the risk is not harm. The risk is paying for it, or reporting it to your board, as a proven visibility lever.
What llms.txt is claimed to do
The pitch is simple. llms.txt: a markdown file at yourdomain.com/llms.txt that gives AI systems a curated map of your most important pages, much as robots.txt gives crawlers rules and a sitemap lists URLs. The implied promise is that a cleaner map leads to more accurate retrieval and more citations.
That is a hypothesis about how AI products choose sources, and it can be tested. For a buyer, the useful question is not whether the idea sounds sensible but whether any platform reads the file when it decides what to cite. The next two sections look at the measured evidence and at what the platforms themselves have published.
What the 300,000-domain study found
The strongest public test so far comes from SE Ranking, an SEO software company. According to Search Engine Journal’s report of November 20, 2025, SE Ranking crawled roughly 300,000 domains, checked each for an llms.txt file, and compared the result with how often each domain was cited across responses from major LLMs. It used both correlation tests and an XGBoost machine learning model to estimate how much each factor contributed to citations.
The finding was negative. Simple statistics showed no significant link between the file and citations. When SE Ranking removed llms.txt from the model, the model’s accuracy improved, meaning the file added noise rather than signal. SE Ranking’s own conclusion was that llms.txt “doesn’t seem to directly impact AI citation frequency. At least not yet.” Adoption was 10.13% of domains, spread fairly evenly across traffic tiers.
The implication is that a vendor claiming a citation lift from llms.txt should be able to show a controlled before-and-after result. Absent that, the claim is marketing, not evidence.
| Source | What it says about llms.txt | Type of evidence |
|---|---|---|
| SE Ranking via Search Engine Journal (Nov 2025) | No significant correlation with LLM citations across about 300,000 domains | Observational, large sample, correlation and XGBoost model |
| Google Search Central, AI features documentation | No new machine-readable files or AI text files needed for AI Overviews or AI Mode | Official platform guidance |
| OpenAI crawler documentation | Recommends allowing OAI-SearchBot in robots.txt; does not describe llms.txt as a ranking or citation input | Official platform guidance, as summarized by Search Engine Journal |
What the platforms say
Google is the most direct. Its documentation on AI features in Search states: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features.” It adds that there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary,” and that a page must be indexed and eligible to be shown in Google Search with a snippet.
OpenAI’s published guidance points elsewhere. Its crawler documentation says sites that opt out of OAI-SearchBot “will not be shown in ChatGPT search answers,” so the control that matters is robots.txt access, not a summary file. SE Ranking noted that some server logs show GPTBot occasionally fetching llms.txt, but infrequently and without any link to citation outcomes. For access rules in detail, see which AI bots to allow in robots.txt.
How to evaluate a vendor selling it
Agencies and tools that bundle llms.txt into a “GEO package” are not necessarily wrong to include it. The cost is a few hours at most. The problem arises when the file is priced as a deliverable or presented as the reason citations rose, because other changes usually ship at the same time. Our guide to choosing an AI marketing agency covers the wider vetting process. These questions apply specifically here.
- Ask which platform documentation says the file is read, and request the link.
- Ask what else changed in the same period, such as new content, fixes to crawl access, or PR coverage.
- Ask for citation tracking from before the change, using repeated prompts, not a single screenshot.
- Ask what the vendor would report if citations did not move.
Where the effort is better spent
If llms.txt does not reliably move citations, the practical question is what does. The evidence reviewed here does not rank those factors, and SE Ranking’s report as covered did not name the drivers it found. What can be said is that Google ties AI feature eligibility to ordinary indexing and snippet eligibility, and OpenAI ties ChatGPT search inclusion to crawler access.
That points to a short, checkable list: confirm the pages you want cited are indexable, confirm search crawlers are not blocked, write passages that stand alone when quoted, and measure results over repeated runs. Our piece on measuring AI search visibility when answers change every run explains the measurement side, and what AI search means for SEO covers the overlap with classic ranking work.
What the evidence doesn’t support
A correlation study cannot prove the file never helps. SE Ranking’s own wording leaves room with “at least not yet,” and platforms can change what they read. The study also measures domain-level citation frequency, so it would miss a small effect on specific pages. Search Engine Journal’s coverage describes adding the file as low effort and unlikely to cause technical harm, and advises testing it where it fits your workflow without presenting it internally as a proven lever. We looked for a Search Engine Land 10-site test cited by secondary blogs, could not verify it from a primary source, and left it out.
Frequently Asked Questions
Should my business add an llms.txt file?
It is optional. The file takes little time and is unlikely to cause harm, but current evidence shows no citation benefit. Add it if it is easy, and do not pay extra for it.
Does Google use llms.txt for AI Overviews?
Google’s documentation says no additional files, AI text files, or markup are needed for AI Overviews or AI Mode. Eligibility depends on a page being indexed and snippet-eligible.
What should I check before buying a GEO package?
Ask for platform documentation behind each tactic, a baseline of citation tracking taken before work starts, and a clear account of what else changes during the engagement.
Telling a documented tactic from a fashionable one is most of the work in AI search. At Grow with Hafsa we help teams build AI marketing systems where every tactic has a baseline, a test, and a result you can check.
For a related tactic that vendors also sell as a citation lever, read what the evidence supports on schema markup for AI search.
For a signal with stronger evidence behind it, see our review of brand mentions versus backlinks in AI search visibility.