What is llms.txt and does my dealer website need one?

What llms.txt actually is

llms.txt was proposed in September 2024 by Jeremy Howard (co-founder of Answer.AI). It’s a plain-text file, written in Markdown, placed at the root of your domain — yourdealership.com/llms.txt. The format is simple: a title, a brief description of what your site is, and then curated links to the most important pages on your site, each with a one-line summary.

The idea is analogous to robots.txt but for AI models rather than search crawlers. Instead of telling crawlers what they can access, llms.txt tells LLMs what’s worth accessing — a hand-curated map of your site’s value.

Who’s currently using it

Adoption trajectory in 2026:

• ~10% of top 1,000 websites publish an llms.txt file
• Known adopters: Anthropic, Stripe, Cursor, Cloudflare, Vercel, Mintlify, Supabase
• Sector concentration: technically sophisticated brands (developer tools, SaaS), technology publishers
• Slow adopters: financial services, healthcare, legal, and — relevant here — automotive dealerships

For context: adoption rates are similar across mid-traffic and high-traffic sites (roughly 8-10% either way). It’s not a case where big brands adopted early and small sites are behind — it’s a case where most of the web hasn’t decided it matters.

Do AI models actually read it?

This is the important question and the honest answer is: not really, not yet.

• OpenAI (ChatGPT) — no documented fetching of llms.txt
• Anthropic (Claude) — no documented fetching of llms.txt (despite Anthropic publishing one on their own site)
• Google (Gemini, AI Overviews) — Google publicly said llms.txt is not something they use
• Perplexity — no confirmed fetching
• Microsoft Copilot — no confirmed fetching

A March 2026 study analyzing ~300,000 domains found no measurable correlation between publishing an llms.txt file and being cited by AI engines. A machine-learning model trained to predict AI citation frequency actually improved when the llms.txt variable was removed — the file added noise, not signal.

So why publish one?

Three reasons, in order of strength:

1. Optionality. If llms.txt adoption grows and AI models start fetching it, you’re already prepared. Because generation is essentially free (a few kilobytes of text), this is cheap insurance.

2. Human readability. A well-structured llms.txt is a genuinely useful sitemap for humans who want to understand what your site contains. Developers, researchers, and journalists sometimes use it.

3. It doesn’t hurt. Unlike some tactics that create actual costs or downsides, llms.txt is inert if unused.

Notably absent from this list: “It will get you cited by ChatGPT.” The evidence doesn’t support that claim.

What actually drives AI citations for dealer sites

Since llms.txt currently doesn’t, here’s what does (based on multiple 2026 citation studies):

• Being included in third-party listicles. “Best X for Y” content accounts for ~44% of ChatGPT citations for commercial queries. Get into local “best used car dealers in [city]” articles.
• Structured data on your own domain. AutoDealer, LocalBusiness, and Vehicle schema — these give AI engines the entity information they need to represent your dealership accurately.
• Content answering specific questions. Q&A-formatted content matches how AI engines break down user questions into sub-queries.
• Wikipedia and Reddit presence. These two domains alone account for ~25% of US ChatGPT citations. For dealers, a well-maintained Google Business Profile plus authentic community reviews on car forums matters more than any meta tag.
• Brand mentions in trusted sources. Local news, community publications, industry sites. Unlinked mentions still count — AI engines use entity recognition.

Should you spend time on llms.txt anyway?

If your web platform generates it automatically: sure, use the feature. Costs nothing.

If it would take you or your agency significant time to create and maintain one: probably not. That time is better spent on structured data, content that answers real shopper questions, and getting your dealership included in local “best of” lists.

Zoomshop generates llms.txt automatically for every dealer site — not because we think it’s the primary AEO lever, but because it’s essentially free to include and gives you the optionality bet without the cost. If a future version of ChatGPT or Perplexity starts fetching it, you’re already covered.