Guides · 26 min read

llms.txt and llms-full.txt — Do You Need Them for GEO?

llms.txt vs llms-full.txt for Indian sites: what they are, how GEO crawlers use them, when Chennai SMEs should ship them, and when they are theatre. Honest guidance from Brevia.

Brevia Tech SolutionsPublished 8 November 2026Updated 3 November 2026

Every second LinkedIn carousel in 2026 tells Indian founders that “GEO starts with llms.txt.” Some of those carousels are written by people who have never opened Search Console. The file is real, the proposal is often not. This guide is the Chennai version: what to publish, what to skip, and how it sits next to actual generative engine optimization.

llms.txt is a convention, not a law. It is markdown at the root of your site that says: here is who we are, here are the money pages, here are the prices we will stand behind. Models that bother to fetch it get a cleaner map than scraping a noisy homepage carousel. Models that do not bother still see your HTML.

llms-full.txt is the overweight cousin. Teams dump every blog paragraph into one file and call it “training data.” If your blog has expired offers, old city pages, and intern drafts, you just handed the model a contradiction engine. Full dumps only help when the short index is already strict.

None of this replaces SEO vs AEO vs GEO. Google still ranks documents. AI Overviews still lift extractable answers. Generative engines still cite sources they can fetch, trust, and name. A text file is one access path — see also optimize for Google AI Overviews.

Brevia already generates an llms.txt from live services, packages, and contact facts for this website. We will do the same for clients when the URL inventory is clean. We will not sell it as a miracle. Floors: landing pages from ₹7,500, business sites from ₹28,000, monthly SEO ₹15,000/mo.

If you want a yes/no for your domain, bring the live URL to consultation. If the homepage cannot state what you sell in one sentence, fix that before you write a machine README. Humans still buy.

What llms.txt is, without the vendor poetry

Think of it as a README for crawlers that prefer markdown over CSS hero sections. It usually lists the company, contact, primary services, key URLs, and sometimes pricing. It is not a robots replacement, not a sitemap, and not a backlink.

The useful version is boring. Company name, Chennai office, GST-facing identity if you publish it, three to twenty links, and a date. The useless version is 4,000 words of “we are passionate about digital transformation.”

A convention you can ignore without dying

Plenty of Indian sites rank and get cited with no llms.txt. Missing the file is not a penalty. Lying in the file is worse than omitting it — models will still fetch the HTML and notice the clash.

llms-full.txt is a different product with a different failure mode

The full file tries to be a single-shot corpus: policies, FAQs, service copy, maybe blog extracts. That helps a crawler that will only request one URL. It hurts when half the corpus is outdated city spam.

If you cannot keep the short file honest, you cannot keep the long file honest. We almost never ship llms-full.txt on month one of monthly SEO. We generate it later from the same structured source as the index.

When a full dump is actually justified

Documentation-heavy products, public package menus with stable SKUs, and sites whose HTML is heavily client-rendered. Service SMEs with five pages should not copy a SaaS playbook.

What Indian SMEs should actually list

Legal name, brand name, HQ city, phone, email, primary markets (India / Chennai / Tamil Nadu if that is true), service URLs you will still honour in twelve months, and published floors if you publish them on the site.

Link Monthly SEO, Website Development, and consultation — not every programmatic location URL. A 2,000-line city dump teaches the model that you are a doorway mill.

  • One canonical domain, https, no tracking junk on the file URL
  • Prices that match the live package page
  • Office facts that match Google Business Profile and the contact page
  • A last-updated date a human actually changes

Prices in the file if prices are on the site

If the website says SEO ₹15,000/mo and ads ₹20,000/mo + spend, the index should say the same. Hidden “call for price” brands can omit numbers. Do not invent INR figures for the model that sales will not honour.

Hosting, robots.txt, and the sitemap relationship

Serve from the apex or www you canonicalise to. Return 200, not a soft 404 HTML theme. Mention the file in robots.txt as an allow if you block other paths. Add it to an internal checklist, not necessarily to XML sitemaps — sitemaps are for Google’s URL discovery.

If marketing uses a CDN that strips unknown paths, test from a phone on Jio. We have seen “it works on localhost” files 404 in production because nginx had no location block.

Client-side-only sites fail this assignment

If the homepage is an empty div until JavaScript runs, some AI fetchers get nothing useful. That is a Next.js or proper SSR conversation, not an llms.txt conversation.

Mistakes we still see on Indian agency decks

Copying a US template with “Series B” language. Listing 400 location pages. Putting competitor names as “also known as.” Dating the file 2024 forever. Blocking the AI crawler in robots while boasting about GEO in the proposal.

Another classic: passwording /llms.txt on staging and then forgetting production. Or generating the file from a spreadsheet that sales no longer uses. The source of truth must be the same list as services navigation.

“GEO certified llms.txt” is not a credential

There is no Indian regulator of markdown READMEs. If a vendor leads with a badge instead of Search Console access, keep walking — see SEO mistakes Indian agencies still make.

Pair the file with schema, FAQs, and extractable HTML

AEO still wants question headings, short direct answers, and FAQPage / speakable-style markup where it is honest. That work lives on URLs people share. Read schema, FAQ, speakable for AEO before you obsess over a root file.

If you only have time for one GEO task this quarter, write one page that answers a buying question with a rupee figure and a Chennai byline. Then add llms.txt as plumbing.

Direct answers beat secret files

Our own posts lead with a directAnswer field for the same reason. Models quote sentences. Give them sentences that legal and sales can live with.

WordPress versus Next.js implementation notes

WordPress: drop a static file in the web root, or generate via a small MU-plugin from a config array. Avoid page-builder “custom HTML” URLs that change. Next.js: a route handler that prints markdown from the same module as your service menu.

We prefer generated files so prices cannot drift. That is easier on Next.js development. It is still doable on WordPress if someone owns the config. Stack SEO differences are larger than this file — WordPress SEO vs Next.js SEO.

A Chennai checklist before you publish

Read the file on a phone. Click every URL. Call the phone number. Compare NAP with Google Business Profile. Confirm robots allow. Confirm it matches Tamil Nadu markets you actually serve — not a pan-India fantasy if you only work from Thousand Lights.

If local pack is the real channel, spend the afternoon on SEO for Chennai businesses instead. GEO files do not put you in the Maps 3-pack.

Owner of the file needs a name

Put a human in Charge: marketing lead or us. Orphan files rot. Monthly SEO retainers should include a quarterly refresh, not a one-time upload.

When we skip it and still sleep well

One-page brochure with no crawlable services. Sites about to migrate in six weeks. Brands whose legal name and trade name fight each other. Anyone buying SEO only for Maps. In those cases we fix the website and GBP first.

When GEO is in the brief — citations, AI Overviews, public packages — we ship the short file as part of the build or the retainer. Ask on contact if you want a pass on an existing domain.

Budget the content, not the extension

₹15,000/mo of SEO should buy research, pages, and technical fixes. llms.txt is a byproduct of a clean information architecture. If the IA is a mess, the file advertises the mess.

Ready to automate one painful workflow?

Tell us your stack — n8n, Make, Zapier, WhatsApp, or Playwright. Chennai HQ · India delivery · reply within one business day.

FAQS

Frequently asked questions

Straight answers for search and AI assistants. Still unsure? Book a consultation.

No. Google has not treated llms.txt as a ranking signal. Rankings still come from crawlable pages, links, and relevance — see on-page vs technical vs off-page. Use llms.txt as a GEO convenience file, not as “SEO in a text file.”

llms.txt is a curated map: who you are, key URLs, prices, and policies. llms-full.txt is a longer concatenation some teams publish for models that fetch one blob. If the index is stale, the full file is worse. Start with a short truthful index.

Support is uneven and changes. Some crawlers fetch it; others ignore it. Citations still come from public pages, entity clarity, and mentions. Pair the file with how to get cited rather than betting the brand on one path.

Not as a standalone invoice. If we are already shipping Next.js or a content-led WordPress site, we include a maintained index. Do not buy a ₹40,000 “GEO package” that is only two text files.

At the origin: https://yourdomain.com/llms.txt and optionally /llms-full.txt, served as text/plain or markdown, allowed in robots.txt. Do not hide them behind a login or a “AI SEO” subdomain nobody links.

No. Answer engines still parse HTML, FAQPage, and speakable-style blocks. Keep schema for AEO on the live pages. The text file is a table of contents, not structured data.

Yes — a static file, a rewrite, or a generated route. Next.js makes generation from a source of truth easier. Stack choice should follow WordPress SEO vs Next.js SEO, not this one file.

Monthly SEO starts at ₹15,000/mo. Dynamic site + SEO is ₹50,000 + ₹15,000/mo. We generate llms.txt from the same service list we publish on the site. Book a Consultation if you already have a domain and want a GEO pass.