Does Your Business Website Need an llms.txt File?

The short verdict for a service-business website

For a typical plumber, law firm, design studio, clinic, or contractor, the answer is optional and low priority. A concise, well-organized website already gives crawlers and visitors direct paths to the home, about, service, location, contact, policy, and article pages. If those pages are weak or hard to discover, another index file does not repair them. Start with the established technical and editorial foundations described in the SEO, GEO, and AI search guides, then decide whether an extra machine-oriented map has enough value to maintain.

What the llms.txt proposal actually describes

The official llms.txt proposal describes a public `/llms.txt` Markdown file intended to help language models use a website at inference time. Its suggested format begins with an H1 naming the project or site, may include a blockquote summary and additional explanation, and then organizes important links under H2 sections. The linked resources may have Markdown versions that are easier for a tool to read than a complex application page. This is most naturally useful for documentation and knowledge sites where a curator can point to canonical manuals, reference pages, examples, and optional material.

The proposal is not an access policy. Putting a link in llms.txt does not grant a crawler permission that robots.txt denies, override a `noindex` directive, bypass authentication, change copyright terms, or establish a license. It also does not prove that any particular model or search feature reads the file. Because the file is publicly reachable, never include private documentation, customer records, credentials, unpublished prices, staging URLs, or anything a normal visitor should not see.

Do not confuse llms.txt with three established controls

robots.txt: The IETF’s RFC 9309 standardizes the Robots Exclusion Protocol used by site owners to tell crawlers which URI paths they may access. The RFC also states that these rules are not access authorization. A disallowed URL is not protected like a password-gated page. Google likewise explains that robots.txt manages crawling, not reliable removal from search results. Use authentication for private content and supported `noindex` controls for crawlable pages that should not appear in search, after considering the platform’s behavior.

XML sitemap: Google’s sitemap documentation says a sitemap helps search engines discover URLs, especially on large, new, media-heavy, or weakly linked sites, but it does not guarantee crawling or indexing. A sitemap should list canonical URLs the business wants discovered. It is not a prose briefing, crawler permission file, or proof that the pages are high quality. The XML sitemap and robots.txt guide walks through these controls together.

Structured data: Schema markup labels entities and page content using a vocabulary such as Schema.org. Google supports specific structured-data features under its own requirements. Markup must match what visitors can see and does not guarantee a rich result. llms.txt does not replace Organization, LocalBusiness, Article, Product, or other appropriate structured data, and adding those types does not remove the need for accurate visible content.

What Google, OpenAI, and Bing currently document

• Google says there are no additional technical requirements, special schema, or special text files needed to appear in AI Overviews or AI Mode beyond normal Search eligibility and best practices.
• Google requires a page to be indexed and eligible for a snippet to appear as a supporting link in its AI features; normal preview controls continue to apply.
• OpenAI documents OAI-SearchBot for ChatGPT search, GPTBot for potential model training, and ChatGPT-User for user-initiated actions, with robots.txt controls and published IP information.
• OpenAI’s official crawler page does not make llms.txt a condition for ChatGPT search inclusion.
• Bing’s official Webmaster Guidelines describe discovery, crawling, indexing, and content quality across Bing search and related experiences without making llms.txt a universal requirement.
• None of these platform documents promises an AI citation because a site publishes an llms.txt file.

Google’s AI features and your website page is the controlling source for claims about Google’s current requirements. It explicitly tells site owners to focus on crawl access, internal links, page experience, indexable text, supported structured data that matches visible content, and current Merchant Center or Business Profile information where relevant. A vendor case study about llms.txt may be an interesting experiment, but it cannot overrule Google’s own documentation or establish a ranking factor.

OpenAI’s crawler overview says OAI-SearchBot is used to surface sites in ChatGPT search and recommends allowing that user agent and the published IP ranges for sites that want to participate. This is a concrete access step. It is still not a selection guarantee, and it is separate from whether the publisher permits GPTBot. Review each user agent by purpose instead of adding a broad rule copied from an old article.

Use a priority ladder before creating the file

For businesses seeking implementation help, SEO services should prioritize crawlability, information architecture, content quality, internal linking, schema accuracy, and measurement before an experimental file. Ask any provider recommending llms.txt which documented platform consumes it, how the provider will test that claim, who will update the file, what happens when URLs change, and which higher-priority work has already been completed. “It might be useful and is cheap to maintain” is a defensible reason. “It guarantees GEO rankings” is not.

If you add llms.txt, keep the experiment clean

Keep a short change log outside or inside the repository so the team knows why each link was selected. If no documented consumer ever requests the file and maintenance repeatedly falls behind, remove it. A missing llms.txt is not a search error. A stale guide that points to retired claims can create confusion. The related guide on getting cited in AI answers focuses on substantive, platform-supported foundations rather than a single proposed file.