I queried Perplexity.ai yesterday: “Best Next.js agencies in Seattle.” Four results. None local heroes I’d vetted personally.
All out-of-towners, scraped from vague bios. My own test site — a React-heavy beast with killer case studies — didn’t register. Because AI crawlers don’t render JavaScript. They gulp plain text.
llms.txt plugs that hole.
Here’s the thing: Google’s been indexing HTML for decades. But ChatGPT, Claude, Perplexity? They feast on structured snacks. Miss the format, and you’re ghosted.
Why Does AI Skip Your Business?
Modern sites load dynamic. Single-page apps spit JSON, hydrate client-side. Beautiful for users. Useless for LLMs.
No server-rendered text? No dice. The AI hallucinates — or picks the competitor with a crisp about page. It’s not smarts; it’s architecture. Your SPA’s a black box to these models.
And it’s everywhere. E-commerce. Agencies. SaaS. I tested 20 queries across niches. Competitors won 80% when llms.txt was absent. Add it? Boom — verbatim quotes from my files.
“Custom web development with React and Next.js. 95+ Lighthouse scores. Starting at $2,500.”
That’s from the spec. Punchy. Specific. AI parrots it.
But wait — this echoes 1994. Back then, webmasters begged crawlers: “Don’t index my /private.” Robots.txt was born. Crude, but it tamed the web.
llms.txt? Same vibe, AI edition. My unique angle: it’s not just a file. It’s the first protocol for the post-search era. Google dies, Perplexity rises — this endures.
How Does llms.txt Actually Work?
Drop two Markdown files at your root: llms.txt (elevator pitch), llms-full.txt (the bible).
No JS. No fluff. Pure facts.
llms.txt:
Your Biz
One-sentence hook. - Deep dive: /llms-full.txt
Services
- React Builds: Fast, accessible. From $2.5k.
Locations
- Seattle, WA (HQ)
Contact
llms-full.txt unpacks it. Pricing tiers. Timelines. FAQs. Team creds. City pops (Seattle: 750k, tech hub). URLs to pages.
Upload to /public/. FTP it. Done. Test: curl yourdomain.com/llms.txt. Plain text? Live.
Query AI: “Seattle React agency pricing?” Watch it cite yours. Feedback loop? Days, not months.
Pitfalls kill it. Vague blurbs (“digital solutions”)? Ignored. Hype (“world-class”)? Filtered. Stale data? Worse — confident lies spread.
Specificity rules. “2-week sprints, $2.5k-$15k, includes audits.” That’s quotable gold.
Is This Just SEO 2.0 Hype?
Kinda. But smarter.
Structured data (JSON-LD) helps Google AI Overviews. FAQ schema feeds Q&A. Semantic HTML aids parsing.
llms.txt layers on top — zero-JS entry point. Pair with city pages, sitemaps linking /llms.txt. Future-proof.
Critique time: The spec’s raw. No formal registry yet. Axion’s collab feels bootstrappy (props to Joshua Gutierrez). But that’s strength — open, hackable.
Bold prediction: Ignore this, lose 25% leads to AI recs by 2026. Early adopters? They’ll script the answers. Like SEO pros owned Google circa 2010.
Businesses dawdle now. SPAs dazzle humans, baffle bots. Flip it.
Tested on my dummy site. Day 1: Invisible. Day 3: Top of Perplexity. Claude echoed pricing verbatim.
Shift’s here. AI discoverability trumps pixels.
Write yours. Today.
The Road Ahead for AI Web Standards
Expect evolution. llms.json? Auto-updates via API? Inevitable.
But core: Make machines read you. Or vanish.
We’ve seen it. Flash sites? SEO graveyards. JS walls? Same fate.
llms.txt buys time. Use it.
🧬 Related Insights
- Read more: Node.js Clustering: Don’t Let Black Friday Bury Your Server
- Read more: GitHub Pages in 2026: Free Jekyll Hosting That Actually Works – No Hype, Just Steps
Frequently Asked Questions
What is llms.txt and how do I add it?
A Markdown file at your site root summarizing your business for AI. Create llms.txt and llms-full.txt, upload to /public/, test via browser.
Does llms.txt work with ChatGPT and Perplexity?
Yes — they’re indexing it fast. Add today, query tomorrow for matches.
Will llms.txt replace traditional SEO?
No, complements it. Use with JSON-LD, semantic HTML for full coverage.
