TL;DR
In 2026, prospects increasingly ask ChatGPT, Perplexity or Copilot about companies instead of typing queries into Google. The llms.txt file is a simple, model-friendly index of your site content — it complements your sitemap and JSON-LD with a layer of “company knowledge” in Markdown. Below: what to put in the file, how to serve it from Next.js on DevStudioIT Cloud, and how to connect it to Branchly data without duplicating chaos.
Who this is for
- Owners of corporate sites whose brand “does not exist” in AI answers
- Marketing teams planning visibility beyond classic SEO
- Next.js developers looking for a lightweight standard without a heavy CMS
- B2B service firms with offers, FAQ and case studies to organize
Keyword (SEO)
llms.txt corporate website, ai search optimization 2026, chatgpt site indexing, perplexity crawl, structured site knowledge
What llms.txt is and how it differs from robots.txt
robots.txt tells bots what they may crawl. llms.txt (a community proposal from around 2024–2025) tells AI models and agents what they should know about your company — in one Markdown file at /llms.txt or /llms-full.txt.
Typical structure:
# Company name
> Short description (1–2 sentences) — elevator pitch.
## Offer
- Service A — for whom, outcome
- Service B — scope, indicative price if public
## FAQ
- Question → answer in 2–3 sentences
## Contact
- Email, region, supported languages
## Optional: links to full content
- /en/services
- /en/case-studies/client-xIt does not replace Schema.org or your sitemap — it adds a layer optimized for LLMs that prefer concise, hierarchical text over parsing the entire DOM.
Why it matters in 2026
- AI Overviews and search chatbots cite sources — if your site is hard to summarize, a competitor with a clear
llms.txtwins - Perplexity, ChatGPT Browse, Bing Copilot visit domains more often for up-to-date facts
- Hallucinations about your company — without explicit knowledge, the model “fills in” from a blog fragment from three years ago
- Multilingual sites — a separate file or section per locale reduces PL/EN mixing in answers
There is no official “llms.txt Google ranking factor”, but controlling the narrative in AI channels is a real advantage — especially for local services and SaaS.
What to include — content checklist
| Section | Purpose | Data source |
|---|---|---|
| Company description | Identity, USP | Homepage, pitch deck |
| Offer / products | What you sell | /services, public pricing |
| Personas / industries | Who you work for | Case studies |
| FAQ | Objections, process, timelines | Forms, support |
| Engagement process | Steps from brief to launch | Internal playbook |
| Case studies (short) | Social proof | Portfolio |
| Contact and area | How to reach you | Footer, GMB |
| Policies | GDPR, cookies — summary | /privacy-policy |
| Changelog | Last update date of the file | CI / manual |
Avoid: walls of text, duplicating the entire blog, sensitive data (NDA B2B pricing), stale fixed prices without a date.
Next.js implementation (App Router)
A static file in public/llms.txt works immediately, but a dynamic version lets you pull current data from an API:
// app/llms.txt/route.ts
import { NextResponse } from 'next/server';
export const revalidate = 3600;
export async function GET() {
const body = await buildLlmsTxt(); // fetch from CMS / Branchly / markdown
return new NextResponse(body, {
headers: {
'Content-Type': 'text/plain; charset=utf-8',
'Cache-Control': 'public, max-age=3600, s-maxage=3600',
},
});
}Multilingual variant: /en/llms.txt and /pl/llms.txt or one file with # English / # Polski sections.
Branchly (branchly.cloud) as source: store FAQ, service descriptions and metadata in tables — the Route Handler assembles llms.txt at build time (SSG) or hourly (ISR). Single source of truth, no manual sync between site and AI file.
Hosting on DevStudioIT Cloud (devstudioit.cloud): serve the file from the same domain as the site (not a CDN subdomain without canonical), with correct Content-Type and no block in robots.txt.
Linking with JSON-LD and sitemap
Consistency triangle:
- JSON-LD (
Organization,LocalBusiness,FAQPage) — for Google Rich Results - sitemap.xml — URL list for classic index
- llms.txt — condensed knowledge for LLMs without crawling 200 pages
The same facts (address, public company ID, service scope) must match. Discrepancies mean the model picks a “newer” blog fragment and confuses the customer.
Optional <head> link (not required, but helpful for tools):
<link rel="alternate" type="text/plain" href="/llms.txt" title="LLM site summary" />llms-full.txt and URL mapping
For larger sites we use two files:
/llms.txt— summary ~800 words, updated weekly/llms-full.txt— index with links: “Full description of service X →/en/services/x”
The AI crawler can go deeper but gets a priority map — what is canonical vs archive.
Testing and maintenance
- Open
/llms.txtin the browser — readable Markdown, UTF-8, no BOM - Ask Perplexity: “What does [your company] do? Source: domain” — compare with intent
- In ChatGPT (Browse) check whether it cites current offers
- Date in header:
Last updated: 2026-07-20— models and humans know the file is fresh - CI hook: diff
llms.txtwhen/serviceschanges in the repo
Process: marketing edits content in Branchly → webhook → rebuild llms.txt fragment → deploy on DevStudioIT Cloud. No manual copy-paste from Notion.
FAQ
Does llms.txt replace SEO?
No. It complements AI and agent channels. Classic SEO (content, Core Web Vitals, links) still drives organic traffic.
Should I block AI in robots.txt?
Separate decision. You can publish llms.txt even with restrictive robots — you consciously share selected facts. Full crawl is still controlled per bot.
Does Google officially support llms.txt?
As of 2026: no official Google documentation for llms.txt. The standard grows ecosystem-wide (Perplexity, dev tools). Treat as emerging best practice, not a ranking guarantee.
How to handle multiple languages?
Separate files per locale (/en/llms.txt, /pl/llms.txt) with hreflang on the homepage, or one file with clear language sections and a link to the canonical offer page.
Should I include prices?
Only public ones, with a validity date. For custom quotes: “from $X” + “current pricing: /en/pricing”.
CTA
Want ChatGPT and Perplexity to describe your company accurately — with up-to-date offers and FAQ?
- Talk about AI visibility — content audit,
llms.txt, JSON-LD and deployment on DevStudioIT Cloud - Websites — Next.js, Branchly, knowledge structure for SEO and AI
About the author
We build fast websites, web/mobile apps, AI chatbots and hosting setups — with a focus on SEO and conversion.
Recommended links
From theory to production — Branchly, our hosting stack and shipped work.