What Is llms.txt?

The one-sentence definition

llms.txt is a plain-text file placed at your website root that tells ChatGPT, Claude, Perplexity and other large language model crawlers who your brand is, what it offers and which pages are worth citing. It is written in Markdown and always lives at https://yourdomain/llms.txt.

When an AI engine answers a question like "which agencies in Kaohsiung build AI-driven websites", it has to judge a site's positioning and credibility very quickly. Crawling page by page is slow and easy to misread; a well-structured llms.txt hands the answer over directly.

You can inspect our own file at sunshinetaiwan.com/llms.txt.

llms.txt vs robots.txt vs sitemap.xml

All three live at the site root, but they serve different readers. In short: robots.txt governs "may you fetch", sitemap.xml lists "which URLs exist", llms.txt explains "who you are and what to cite".

The three complement each other. A missing llms.txt will not hurt Google indexing, but it forces AI engines to take the long way round when interpreting your brand.

Six essential blocks

The value of llms.txt lies in density, not length. These six blocks cover what AI engines most often need when answering brand questions.

• One-sentence positioning: what the brand does, for whom, and where. This is often the sentence AI quotes verbatim.
• Company details: legal name, address, phone, email and business hours, matching the website and Google Business Profile word for word.
• Service list: service names with a single line of explanation each, free of marketing adjectives, so AI can map user intent directly.
• Key page index: the URLs you most want cited, each with one line describing the question that page answers.
• Condensed FAQ: question-and-answer pairs drawn from FAQs that actually exist on the site; AI strongly favours this format.
• Last updated and usage notes: a maintenance date and citation guidance, which raises source credibility.

Five steps to build your own

1. Inventory the facts — list legal name, address, phone, email, hours and services, then reconcile them against the website, Google Business Profile and social bios so no version drifts.
2. Write the positioning sentence — one adjective-free sentence, for example "Sunshine Pictures Technology is a Kaohsiung-based brand marketing company offering AI digital engine, video production and influencer integration services".
3. Build the Markdown structure — open with # Brand name, use ## for services, pages, FAQ and contact blocks, and format links as - [Page name](URL): one-line description.
4. Publish at the root and verify — open https://yourdomain/llms.txt and confirm it renders as plain text with no redirect and no 404.
5. Keep it in sync — whenever services, contact details or key pages change, update llms.txt, the website copy and the structured data together.

Five common mistakes

1. Pasting marketing copy — unverifiable claims such as "industry leading" tend to be skipped, diluting the density of citable facts.
2. Details that contradict the site — a phone number that differs from the contact page signals an unreliable source and lowers citation willingness site-wide.
3. URLs without descriptions — with no one-line summary, AI cannot tell what a page answers, wasting the exposure.
4. Placing it in a subdirectory — llms.txt must sit at the root; most crawlers will not look for /docs/llms.txt.
5. Writing it once and forgetting it — stale details keep being cited long after they change.

Frequently asked questions

Low cost, long payback

llms.txt is not a ranking shortcut. It is a factual record that helps AI understand your brand correctly. The build cost is minimal — usually just a tidy set of company facts — but the cost of getting it wrong is high, because once an AI has learned stale or contradictory details, correcting them takes far longer.

Further reading: SEO vs GEO vs AI Search: the 3-layer strategy for 2026.