Your site probably has great SEO for Google — and is completely opaque to the AI assistants your customers now ask first. The fix takes about 20 minutes. Here's the whole checklist with copy-paste snippets.
1. Unblock AI crawlers (2 min)
Check yoursite.com/robots.txt. Many templates silently block AI bots. Make sure these are allowed:
User-agent: GPTBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: PerplexityBot
Allow: /
Sitemap: https://yoursite.com/sitemap.xml
If an AI crawler can't read you, nothing else matters.
2. Publish llms.txt (10 min)
llms.txt is a markdown file at your web root — a curated summary AI systems can quote. Think robots.txt, but for meaning instead of permissions:
# Acme Dental — family dentistry in Austin, TX
> Family dental clinic in Austin since 2011. Same-day emergency
> appointments, transparent pricing from $90.
## Facts
- Founded: 2011
- Services: cleanings, implants, emergency care
- Address: 12 Main St, Austin, TX
- Price range: $90–$1,200
## Pages
- [Services](https://acmedental.com/services): full price list
- [Book](https://acmedental.com/book): online booking
Rules: concrete facts only. AI engines quote specifics, never slogans.
3. Add JSON-LD (5 min)
Machine-readable truth in your <head>:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Acme Dental",
"address": { "@type": "PostalAddress", "addressLocality": "Austin", "addressRegion": "TX" },
"priceRange": "$90-$1200",
"foundingDate": "2011"
}
</script>
For SaaS use SoftwareApplication, for content use Article, and add FAQPage schema to your FAQ — assistants live in question-answer format.
4. Verify (3 min)
-
curl yoursite.com/llms.txt→ 200 - Google Rich Results Test → JSON-LD parses
- Ask ChatGPT: "What do you know about {your business} at {your URL}?" — with browsing on, it should now cite your facts.
The measuring problem
The annoying part is step 5: knowing whether any of this changed what AI engines actually say about you — answers vary by phrasing, region and model version. Full disclosure: that's the problem I built AnswerRank around — it asks ChatGPT/Claude/Perplexity/Gemini your buyers' questions and tracks whether you're in the answers (free check, no signup). But everything above works with or without it. Ship the files first.
Top comments (5)
didn't know llms.txt was becoming a thing, definitely gonna try adding that to my project today
Thanks! Two gotchas worth knowing before you ship it: (1) keep it to verifiable facts — founding year, services, prices, location. Assistants quote specifics and skip marketing copy. (2) After deploying, run curl yoursite.com/llms.txt and check it actually returns 200 as plain text — a lot of frameworks route unknown paths to a 404 page or serve it as HTML, and then nothing reads it. Drop your URL here if you want a second pair of eyes once it's live.
didn't know about llms.txt yet, definitely gonna add that to my project tonight. thanks for the tips
Glad it was useful! Same tip as in the thread above: facts over adjectives, and curl the file after deploy to be sure it's served as plain text. If you also add FAQPage JSON-LD to an existing FAQ page, that's usually the highest-leverage 10 minutes after llms.txt.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.