An llms.txt file is an optional text directory that points readers and software to useful pages on your site. Keep one if it helps people or assistants find your material and you can maintain it. Google does not require it for its AI search features, and adding the file is not evidence that your business will be cited or recommended.
What belongs in the file?
Start with a short, factual description of the business or product. Add links to the pages that explain it best: services, product documentation, support, and contact details. Each entry should say what the destination actually contains.
For an illustrative deck-building business, a useful entry might read:
Composite deck installation in Bellevue: materials, project examples, and the information to send with an inquiry.
Open that page before publishing the entry. If it contains only a general introduction, either improve the page or narrow the description. The file should help a reader choose a destination.
Which checks come first?
Confirm that your important pages are public, reachable, and indexable where appropriate. Check whether the initial HTML contains their main content and whether normal links lead to them. Inspect crawler rules separately from firewall responses.
For ChatGPT search, OpenAI distinguishes OAI-SearchBot from GPTBot, its training crawler. A blanket “allow AI” or “block AI” setting can hide that distinction. Use the provider's current documentation when choosing a policy.
Next, check the pages themselves. Can a visitor identify your services, assess real work, and contact you? An extra index is useful only when its destinations answer those questions.
Keep it small enough to maintain
Use canonical URLs. Remove links to retired services and correct changed product availability. If a product has its own website, link there directly. Avoid copying long passages that will drift as soon as the page changes.
Include the file in your normal release checks: follow its links, compare its descriptions with the visible pages, and review it when an offering changes. A simple directory with accurate destinations is easier to maintain than a second version of the website.
How can you tell whether it helps?
Server logs can show whether a verified client fetched the file. That proves a request occurred. It does not show that the file caused a citation, a recommendation, or an inquiry.
Track those outcomes separately. For search, watch indexed pages and relevant queries. For AI answers, keep the exact prompts, cited URLs, dates, and repeated observations. For the business, count qualified inquiries and ask how people found you.
If you are choosing between improving a vague service page and expanding an llms.txt file, improve the service page first. Our website work includes both the pages customers use and the search structure that helps them be discovered.
Sources
- The llms.txt proposal, for the format and intended use.
- Google: AI features and your website, checked October 21, 2026: no special AI text files or markup are required.
- OpenAI crawler documentation, checked October 21, 2026.