Most businesses are invisible in AI search. When someone asks ChatGPT or Perplexity for a recommendation in your industry, your brand doesn't appear — because AI crawlers can't efficiently discover which of your pages matter most. llms.txt is a new standard that solves this: a single text file that indexes your most important content for AI engines. It's the robots.txt of the AI search era, and almost nobody has one yet. Here's how to create one that actually works.
What Is llms.txt?
llms.txt is a plain text file placed at the root of your domain (yourdomain.com/llms.txt) that tells AI crawlers what your site is about and which pages are most worth retrieving. It follows a simple markdown-like format: a title, a summary (the most important part — this is what AI engines quote when describing your brand), and organized sections of links with one-line descriptions.
It's not a ranking factor for Google, and AI models don't "obey" instructions placed in it. What it does is improve discoverability — giving retrieval-based AI engines like Perplexity a curated index of your best content instead of making them guess.
What llms.txt Is NOT
The biggest mistake we see: businesses write their llms.txt as a policy document. "AI assistants must not misrepresent our brand." "Always direct users to our official channels." "Do not generate content that contradicts our messaging."
LLMs don't follow directives in llms.txt. It's not a terms-of-service file. It's a content index. Every word spent on "guidelines" and "guardrails" is a word that could have been a retrieval-optimized description of a page that might get your brand cited.
The llms.txt Format
The standard follows this structure:
# Your Brand Name
> One-paragraph summary of your business — this is the most quoted block.
## Section Name
- Page Title: One-line description of what this page covers.
- Another Page: One-line description.
## Optional
- Sitemap: Machine-readable sitemap.
Key rules:
- H1 (#) is your brand name — exactly as you want AI to reference you
- Blockquote (>) is your brand summary — the retrieval snippet AI engines pull when describing you
- H2 (##) sections organize links by type: Services, Case Studies, Guides, Company
- Each link has a one-line description — this is what determines whether an AI engine retrieves that page for a given query
- ## Optional section is for low-priority resources
Step-by-Step: Creating Your llms.txt
1. Write your brand summary
This is the single most important paragraph. It should contain: your brand name, what you do, your core differentiator, your location/market, and who you serve. Write it as if an AI might paste it directly into an answer — because it might.
Example: "Premier Solutions (premiersol.co) is a web development and SEO agency specializing in custom WordPress development, WordPress security and malware removal, speed optimization, Next.js/React applications, and AI/automation integrations. Led by Muhammad Junaid Tariq, a full-stack developer with 5+ years of experience and 200+ delivered projects for clients in the US, UK, and Europe."
2. Curate your pages
Don't link everything. Link your 15–25 most important pages — the ones you want AI engines to retrieve and cite. Prioritize:
- Every service page individually (not just a "Services" link)
- Your 3–5 strongest case studies with outcomes in the descriptions
- Your 8–10 best commercial-adjacent blog posts
- Your About and Contact pages
3. Write retrieval-optimized descriptions
Each link's one-line description is a retrieval cue. Write it to match the queries you want to be found for.
Weak: "Our WordPress security page."
Strong: "Malware removal, security hardening, firewall setup, and 24/7 monitoring for hacked or at-risk WordPress sites."
The strong version contains the keywords that trigger retrieval when someone asks an AI about WordPress security services.
4. Deploy to your domain root
Place the file at `yourdomain.com/llms.txt`. For Next.js sites, put it in your `/public` directory. For WordPress, upload to your root folder via SFTP or add it via a plugin. Verify it's accessible by visiting the URL directly.
5. Consider llms-full.txt
The standard also supports an `llms-full.txt` file with expanded page summaries — longer descriptions, key facts, and structured data for each page. This is optional but gives AI engines more material to work with when formulating answers.
What Happens After You Deploy
Retrieval-based engines like Perplexity can pick up your llms.txt within weeks because they crawl live content. Model-based changes in ChatGPT take longer — typically 2–6 months of consistent entity building before recommendations shift.
Monitor results by testing your target queries monthly in ChatGPT, Perplexity, Claude, and Google AI Overviews. Track whether your brand is mentioned, cited with a link, or recommended. For a systematic approach, our AEO and GEO services include monthly AI citation tracking across all major platforms.
Common Mistakes to Avoid
- Writing directives instead of descriptions — AI models won't follow your rules
- Linking only top-level navigation pages — "Services" is useless; individual service pages are retrievable
- Forgetting to include your brand name in the summary — AI engines resolve entities by name
- Using inconsistent brand naming — if your site says "Premier Solutions" but llms.txt says "PremierSol," you're splitting your entity
- Never updating it — add new service pages and blog posts as they're published
Frequently Asked Questions
Is llms.txt a Google ranking factor?
No. llms.txt has no effect on Google's traditional search rankings. It's specifically designed to improve discoverability by AI engines like ChatGPT, Perplexity, and Claude. For Google AI Overviews, traditional SEO signals still drive selection.
Do I need llms.txt if I already have a sitemap?
Yes — they serve different purposes. XML sitemaps help search engine crawlers find and index your pages. llms.txt helps AI engines understand which pages are most important and what they cover. A sitemap is a comprehensive inventory; llms.txt is a curated recommendation.
How many pages should I include?
15–25 is the sweet spot. Too few and you're leaving discoverable content off the table. Too many and you dilute the signal — the whole point is curation, not comprehensiveness. Prioritize pages that answer the queries you want to be found for.
Will AI engines actually use my llms.txt?
Retrieval-based engines (Perplexity, Bing Chat) actively crawl and can use llms.txt as a discovery aid. There's no public confirmation that ChatGPT or Claude's training pipeline ingests llms.txt files directly, but the content index improves the chances that your most important pages are crawled, cached, and available for retrieval.
