Why AI Assistants Get Your Product Wrong
Answer engines read whatever they can scrape. Without a curated source they assemble your positioning, pricing and feature list from marketing pages written for humans.
No Curated Source to Cite
An assistant asked about your product stitches an answer together from nav labels, blog intros and third-party listicles. You have no file that says, plainly and in one place, what the business does.
Hand-Written Files Go Stale
A llms.txt written once and committed is wrong the week you publish three articles and rename a page. Nobody remembers to regenerate it, so it quietly describes last quarter.
Scaffolds Get Shipped by Accident
A half-filled knowledge bundle served to an answer engine is worse than none. Placeholder text and broken cross-links are read as fact by a model that has no way to tell.
How SEOAgent Makes a Site AI-Readable
The knowledge is already in your repo. SEOAgent turns it into the two formats answer engines actually fetch, then keeps both current as the site changes.
llms.txt Built From Real Site Data
`seoagent llms` assembles the markdown map from your page inventory, your published articles, the titles and meta descriptions crawled during the audit, and your business context. Drafts and private paths stay out.
An Open Knowledge Format Bundle
OKF is the open standard Google published in 2026 for packaging an organization’s knowledge as markdown with YAML frontmatter. `seoagent okf scaffold` lays out the directory; your coding agent fills it from your context, strategy and published content.
Validation Before Anything Ships
`seoagent okf validate` checks frontmatter types and cross-links. A bundle that still carries scaffold placeholders, or fails validation, is never published.
Published Where Crawlers Look
`seoagent sync` mirrors the sources into your static directory: `llms.txt` at the site root and the bundle at `/.well-known/okf/`. The public directory is detected from your framework.
Regenerated on Every Sync
llms.txt is rebuilt each time you sync, so it cannot drift after a publish. Every write is content-addressed, so an unchanged file leaves no noise in your git diff.
The Audit Checks the Live Site
The free audit raises `llms_txt_missing`, `okf_bundle_missing` and `ai_files_unpublished`. Having the files in `.seoagent/` is not a pass; only what your deployed site serves counts.
How AI Search Readiness Works
Scaffold and Detect
`seoagent init` creates `.seoagent/okf/` and detects your public directory (`public` for Next.js, Vite and Astro; `static` for SvelteKit, Gatsby, Hugo and Docusaurus). On an older project, `seoagent okf scaffold` adds the skeleton without overwriting files.
Fill the Bundle From Your Context
Your coding agent maps `.seoagent/` into OKF files: the business overview into `index.md`, each keyword cluster into a concept, each answered question into an FAQ, each published article into an entry pointing at its live URL.
Generate and Validate
`seoagent llms` writes `.seoagent/llms.md`. `seoagent okf validate` reports any file missing a `type`, or any cross-link that does not resolve inside the bundle.
Sync, Then Commit and Deploy
`seoagent sync` copies both into your static directory. You review the diff, commit, and deploy through your normal pipeline. That last step stays yours.
AI Search Readiness Use Cases
Answer Engine Optimization
Give ChatGPT, Claude and Perplexity a curated account of the product instead of leaving them to infer one from landing-page copy.
Documentation Sites
Publish a machine-readable map of your docs so an assistant answering a setup question reaches the current page, not a cached old one.
Pricing and Plan Accuracy
Put your plans in an OKF concept file so an assistant quoting your pricing quotes the version you shipped.
Post-Launch Positioning
After a repositioning, regenerate the bundle so the description models read matches the one on the site.
Content Libraries
Keep a large blog legible to answer engines by listing published articles with their live URLs and factual summaries.
Agencies Onboarding a Client
Scaffold, fill and validate a client’s knowledge bundle in one session, then hand back a diff they can review before deploying.
Frequently asked questions
What is llms.txt and does it do anything yet?
llms.txt is a markdown file at your site root that lists your important pages and what the business does, written for language models rather than browsers. Adoption is still early and no engine guarantees it reads the file, so treat it as cheap insurance: SEOAgent generates and republishes it on every sync, which costs you nothing to keep current.
What is the Open Knowledge Format?
OKF is an open standard Google published in 2026 for packaging an organization’s knowledge as a directory of markdown files with YAML frontmatter, designed to be read directly by AI agents. It is vendor-neutral and version-controlled like code. The spec and reference implementation are at github.com/GoogleCloudPlatform/knowledge-catalog/okf.
Does SEOAgent write the bundle content for me?
The deterministic parts are CLI commands: scaffolding the directory, validating it, and publishing it. The knowledge itself is written by your coding agent from your `.seoagent/` context, strategy clusters and published articles, and you review it before it ships. Nothing is invented on your behalf.
Do I need a paid plan?
No. Generating and publishing both files runs locally and happens before the login check, so it works on the free Skill with no account.
Will this overwrite an llms.txt I already serve?
No. If your app already serves `/llms.txt` from a route, sync detects the conflict and publishes nothing rather than shadowing your route with a static file. Set `llms_owner: route` in `.seoagent/project.md` to keep ownership and silence the notice.
Does SEOAgent deploy the files?
No. SEOAgent writes them into your repo and tells you they are ready. You commit and deploy them through your own pipeline, which is also why nothing reaches your live site without your review.
More of the SEO engine
Every feature runs from the same Skill in your repo — install once, get all of it.
Make Your Site Readable to Answer Engines
Install the free SEOAgent Skill, run `seoagent llms` and `seoagent okf validate`, and ship the two files AI assistants look for.