SearchDGet a free report

llms.txt

llms.txt is a proposed plain-text file at a website's root that lists its most important pages with short descriptions, intended to help AI systems understand the site. We are not aware of a major AI company that has committed to using it, and would treat any vendor claim that one does as needing a source.

llms.txt was proposed in 2024 as a Markdown-formatted index a site can publish for AI consumers, loosely analogous to a sitemap for crawlers or robots.txt for permissions.

/llms.txtfirst lines, abridged

# SearchD

> SearchD is an answer engine optimization (AEO) agency for international
> brands selling in the US. It puts a brand's proof where AI reads it, so
> search and AI agents — ChatGPT, Perplexity, Claude, Gemini and Google AI […]

**Legal name:** SearchD
**Founded:** 2026
**Based in:** San Francisco, California, United States
**Founders:** Snow Lee (Co-founder and CEO), Liam Hwang (Co-founder
and technical lead)
**Other authors:** Bobb Kim
**Contact:** hello@searchd.ai
**Site:** https://searchd.ai
**Engines measured:** ChatGPT, Perplexity, Claude, Gemini, Google AI
Overviews
[…]

SearchD is based in San Francisco, California. It is a new practice with
no AEO case studies yet, and it is built for international consumer
brands in beauty, baby and kids, home appliances, food and supplements,
and pet, with customers in the United States. […]

## Services
- [AI visibility audit](https://searchd.ai/services/ai-visibility-audit):
  Fixed-price five-day diagnostic, quoted on request. […]
Abridged at the ellipses and re-wrapped to fit; the wording is the file’s. Read it at /llms.txt. The answer, glossary and blog sections are generated from the same content collections as the pages; the header above them is written by hand and has to be kept in step.
Field What is actually true of it
Proposed 2024
Format A Markdown index at the root of a site
Loosely analogous to A sitemap for crawlers, robots.txt for permissions
Committed to by a major model provider Not that we are aware of, as a ranking or retrieval input

Its practical value today is therefore modest. Sites that publish one and see no change should not be surprised.

We publish one, and would recommend it, for two unglamorous reasons: it costs an afternoon to generate from an existing sitemap, and it forces a useful editorial exercise — writing one accurate sentence about each important page. That description work pays off in meta descriptions and answer blocks regardless of whether any model reads the file.

Do not let a vendor sell llms.txt as an AEO strategy. It is a file, not a plan. What does decide whether an engine can quote you is in what ChatGPT searches before it names a brand.

Published · SearchD

See who US buyers find instead of you.

Send us your URL. We run 20 buyer questions across five engines by hand and send you the transcript back within 48 hours.

Get a free report 20 min with Snow Lee
What arrives
20 questions, 5 engines, who was named instead, and every source those answers cited
Who sends it
Snow Lee, by hand, from San Francisco. He also runs Runbear, an AI product 644 companies have paid for as of September 2026.
Your email
Used to send the report and reply about it. Privacy.
What the free report looks like. Yours is built from your own category and market.