In our original guide to llms.txt, we recommended it as a core tactic for AI search visibility. We described it as “the file that tells AI models who you are” and advised every professional services firm to create one.
New research suggests we were wrong about its impact. We think it is important to say so directly and explain what we recommend instead.
What we originally said
Our guide positioned llms.txt as a significant AI visibility lever, comparable to robots.txt for search engines. The logic: AI models need to understand what your firm does, and a structured plain-text summary at your website root gives them exactly that.
We recommended it alongside schema markup, FAQ content, and content restructuring as part of a broader AEO strategy.
What the research found
In 2026, Search Engine Journal reported on a study of approximately 300,000 domains that examined whether having an llms.txt file correlated with being cited by AI models. The methodology compared citation rates between sites with and without llms.txt files, controlling for domain authority, content quality, and other factors.
The finding: no statistically significant effect.
Sites with llms.txt were not cited more often than sites without. No major AI platform (OpenAI, Anthropic, Google, Perplexity) has officially confirmed using llms.txt as a signal for citation decisions.
We recommended llms.txt as a core tactic. A study of 300,000 domains found no measurable effect. When the evidence changes, the recommendation should change.
What the study does and does not show
It is worth being precise about what this research tells us.
What it shows: There is no measurable correlation between having an llms.txt file and being cited by AI models at scale. The file is not a significant visibility lever.
What it does not show: That llms.txt is harmful or that it should be removed. The file costs nothing to maintain and may have future utility if AI platforms adopt it.
The nuance: The process of creating an llms.txt file forces a useful exercise. Writing a clear, structured summary of your firm, its services, its clients, and its differentiators requires the kind of positioning clarity that does improve AI visibility when reflected in your actual website content. The exercise has value. The file itself appears not to.
Why the robots.txt analogy was misleading
The appeal of llms.txt was the robots.txt comparison. robots.txt became a universal web standard because it solved a clear technical problem: telling crawlers which pages to index and which to skip. Search engines needed this information and could not get it any other way.
llms.txt tried to solve a different problem: telling AI models what your firm does. But AI models can already extract this from well-structured web content. A firm with clear service pages, schema markup, and FAQ content already provides everything llms.txt contains, in a richer and more contextual format.
The file was a solution to a problem that good content already solves.
What actually works
Based on current research and our audit data, three factors have demonstrable impact on AI citation rates:
| Tactic | Evidence of Impact | Effort |
|---|---|---|
| Earned media | 239% median citation lift (Stacker, March 2026) | High (ongoing) |
| Structured data / schema markup | 3-4x higher citation rate for pages with JSON-LD | Medium (one-time) |
| Content specificity | Authoritative, specific claims cited 2x more than hedged statements | Medium (ongoing) |
| Original data | 4.1x more citations for pages with original research | High (periodic) |
| FAQ-structured content | Matches AI query-answer extraction patterns | Low (one-time restructure) |
| llms.txt | No measurable effect (300K domain study) | Very Low |
The updated priority order:
-
Earned media. This is the highest-impact lever. Third-party mentions produce 239% more citations than on-site changes alone. For professional services firms, this means press coverage, podcast appearances, guest articles, and genuine community participation (including Reddit).
-
Structured data. Schema markup (Organization, Service, FAQPage, LocalBusiness) helps AI models understand your content accurately. Unlike llms.txt, structured data is embedded in your pages and is actively parsed by AI crawlers. This is a one-time technical implementation with lasting benefits.
-
Content specificity. Replace vague positioning with concrete, authoritative claims. “We are a leading advisory firm” gets ignored. “We have advised 50+ UK mid-market companies on operational restructuring since 2018” gets cited. Authoritative language is cited at 2x the rate of hedged language.
Updated recommendation
Keep your llms.txt file. It costs nothing and the positioning exercise was useful. But move it from “priority tactic” to “nice to have.”
Redirect your effort. The time you would have spent crafting and maintaining llms.txt is better spent on earned media, structured data, and content specificity. These have proven, measurable impact.
Update your strategy. If you followed our AEO playbook and prioritised llms.txt, shift that priority to the three tactics above.
Why we are publishing this
It would be easier to quietly update our original guide and move on. We are writing a separate post because we think the correction itself is useful.
The AI visibility space moves fast. Recommendations that were reasonable six months ago can be contradicted by new data. The firms and advisors that update their recommendations based on evidence, rather than defending their original position, are the ones worth trusting.
If you relied on our original recommendation, we are sorry for the misdirection. The updated playbook is here: Earned Media Is the Best GEO Strategy.
Check your current position
Whether you have an llms.txt file or not, the question that matters is whether AI models are citing your firm. Run a free citation scan to get a current baseline. The results will tell you where your visibility stands and where to invest your effort.
Our AI Search Visibility Audit now includes earned media signal analysis and weighted scoring based on the latest research into what actually drives AI citations.
Questions AI assistants answer about this topic
- Does llms.txt help with AI search visibility?
- According to a study of approximately 300,000 domains published in 2026, there is no measurable correlation between having an llms.txt file and being cited by AI models. No major AI platform has officially confirmed using llms.txt as a ranking or citation signal. The file may have indirect benefits (forcing clarity about positioning) but it should not be treated as a primary AI visibility tactic.
- Should I still create an llms.txt file?
- Yes, but with realistic expectations. Creating an llms.txt file takes 30 minutes and costs nothing. The process of writing it forces useful clarity about your firm's positioning and services. Keep it, but do not prioritise it over earned media, structured data, and content quality, which have proven impact on citations.
- What should I focus on instead of llms.txt for AI visibility?
- Three tactics with proven impact: First, earned media, which produces a 239% median lift in AI citations according to Stacker research. Second, structured data and schema markup, which helps AI models parse your content accurately. Third, content specificity: making concrete, authoritative claims with data rather than vague positioning statements. These three together account for the majority of controllable AI visibility factors.
- Why did people think llms.txt would work?
- The analogy to robots.txt was compelling. robots.txt became a web standard because search engines adopted it. The assumption was that AI models would similarly adopt llms.txt as a machine-readable summary of a website's purpose. But unlike robots.txt, which serves a clear technical function (crawl permissions), llms.txt provides information that AI models can already extract from well-structured web content. The file was a solution to a problem that good content already solves.
Next Step
Want to know where your company stands?
We run 15-20 buyer queries across ChatGPT, Claude, Gemini, and Perplexity and show you exactly where you appear, and where you don't.
Get the Audit | from £750 ↗