# llms.txt for WordPress: What It Is, What It Is Not, and How to Curate One > A practical guide to the /llms.txt convention for WordPress sites. Learn how it differs from robots.txt and sitemap.xml, how to curate links, and what publishing one can and cannot do. Canonical HTML: https://seoboosterpro.com/seo/llms-txt-wordpress-guide/ **Quick answer:** `/llms.txt` is a proposed Markdown file that introduces your site and points AI systems at a curated set of pages. On WordPress, treat it as editorial housekeeping: useful, inexpensive, and honest about its limits. It does not replace `robots.txt`, `sitemap.xml`, or good on-page SEO. This article sits inside our [AI search readiness](/seo/ai-search-readiness-wordpress/) cluster. When you need structured organization context beyond a link list, continue to [Entity Map vs llms.txt, schema.org, and sitemap.xml](/seo/entity-map-vs-llms-txt-schema-sitemap/). #### Where the idea comes from The [llms.txt proposal](https://llmstxt.org/) (Jeremy Howard / Answer.AI, 2024) argues that full HTML sites are awkward for language models: navigation, ads, and scripts get in the way, and context windows are finite. A predictable root file can offer: 1. A short description of the site 2. Guidance for how to use the content 3. Curated links to the pages that matter most 4. Optional pointers to cleaner Markdown versions of docs The proposal expects the file to help mainly at inference time (when a user or agent is looking something up), not as a training-index mandate. That distinction matters for marketing copy: `llms.txt` is closer to a receptionist than to a ranking algorithm. #### What llms.txt is not Be blunt with stakeholders: | Claim you may hear | Reality | | --- | --- | | “It is the robots.txt for AI” | False. `robots.txt` governs crawl permission. `llms.txt` offers curated context. | | “Google ranks sites with llms.txt higher” | Unsupported. Google Search staff have stated Google does not support the file for Search. | | “Publishing it guarantees ChatGPT citations” | Unsupported. Fetching a file is not the same as citing a brand. | | “You should list every post” | Usually harmful. Curation is the point. | If a vendor promises “AI SEO wins” solely from dropping a file in the root, ask for proof beyond their own product marketing. #### llms.txt vs robots.txt vs sitemap.xml ![Comparison of robots.txt, sitemap.xml, and llms.txt](/media/2026/07/discovery-files-comparison.jpg "frame: robots.txt controls access, sitemap lists URLs, llms.txt curates starting pages") * **`robots.txt`**: Says what automated clients may or may not crawl. Still the primary public control plane for many AI crawlers. * **`sitemap.xml`**: Exhaustive (or near-exhaustive) list of indexable URLs for search engines. * **`llms.txt`**: Short Markdown overview plus hand-picked links for LLM-oriented discovery. They complement each other. A healthy site often has all three. None of them replace clear page content or [structured data](/seo/structured-data-for-ai-search/). #### What a good WordPress llms.txt looks like A useful file is boring on purpose: 1. **Site title** as a Markdown H1 2. **One-paragraph summary** (blockquote is common in the proposal) 3. **Sections** with H2 headings (`## Product`, `## Docs`, `## Blog`) 4. **Links** as `- [Title](URL): short note` 5. **Optional** section for secondary material Curation rules that work well for WordPress owners: * Prefer 10 to 30 links over 300 * Include homepage, pricing or contact, core feature or service pages, and a few evergreen guides * Exclude thank-you pages, tag archives, faceted filters, and login URLs * Update when a pillar page changes slug or drops out of your strategy * Keep descriptions factual; do not stuff keywords If you also publish an Entity Map, you can link `/entitymap.json` and `/entitymap.html` from a “Structured knowledge” section. That is how this marketing site does it today. #### Common WordPress implementation options **1. Static file in the web root** Upload `llms.txt` next to `wp-config.php` / the public document root. Simple, but easy to forget when content changes, and some hosts serve the physical file in ways that fight with rewrite rules. **2. Theme or custom plugin rewrite** Register a rewrite for `/llms.txt` and output Markdown. Flexible, but you own maintenance. **3. SEO Booster Tools (free)** SEO Booster’s [llms.txt generator](/docs/tools/llms-txt/) lets you: * Enable dynamic serving of `/llms.txt` * Choose post types and directory include/exclude rules * Cap the number of links * Preview and download the Markdown * Optionally add discovery signals (`LLMS:` line in robots.txt, alternate link, HTTP Link header) * When Pro Entity Map is live, link Structured knowledge endpoints from the same file Open **SEO Booster → Tools → llms.txt**. Product context: [Tools](/feature/tools/). Physical-file warning: if a real `llms.txt` already exists in the web root, it may shadow the virtual route depending on the server. Remove or rename the static file if the dynamic generator should win. #### How to choose which URLs to include Use evidence, not vibes: 1. **Revenue or lead pages**: pricing, product, contact, booking 2. **Top Search Console pages** by clicks (not vanity impressions alone) 3. **Pages AI bots already fetch** (from [AI Bot Tracking](/feature/ai-traffic/)), especially if they are thin in GSC 4. **Evergreen explainers** that define your category 5. **Policy pages** that answer trust questions (about, privacy) when they help agents attribute you correctly Avoid using `llms.txt` as a dumping ground for every blog post from 2019. That recreates the sitemap problem the proposal tried to escape. For the measurement angle, see [Bridge the Gap](/seo/bridge-the-gap/). #### Discovery signals: useful, still optional Some implementations advertise the file beyond the well-known URL: * `LLMS: https://example.com/llms.txt` in `robots.txt` * `` on the homepage * HTTP `Link` headers These can help agents that look for hints. They do not create an obligation for any AI product to fetch or trust the file. Prefer signals you can turn off cleanly if a host or security plugin conflicts. #### Quality checks before you call it done * Open `/llms.txt` in a private window. Confirm Markdown, not an HTML 404 theme page. * Click every link. Broken URLs hurt trust faster than a missing file. * Keep the intro under a few sentences. * Align Organization naming with your About page and schema. * If you publish Entity Map, validate JSON separately at [entitymap.org/validate](https://entitymap.org/validate). * Re-check after migrations, HTTPS changes, or directory moves. #### Where llms.txt sits in the readiness stack ``` robots.txt → may I crawl? sitemap.xml → which URLs exist? llms.txt → where should an LLM start? Entity Map → who are you, and how do concepts relate? Page HTML → what exact answer can be extracted? ``` `llms.txt` is the curated starting index. It does not explain entity relationships or attach publisher-attributed evidence chunks. That is why [Entity Map](/seo/entity-map-vs-llms-txt-schema-sitemap/) exists as a complementary layer, and why [AI search readiness](/seo/ai-search-readiness-wordpress/) treats both as optional but coherent. #### The takeaway Publish `/llms.txt` when you want a clear, maintained shortlist of the pages that represent your WordPress site to AI systems. Curate ruthlessly, keep expectations modest, and pair the file with solid rendered pages, Search Console work, and bot measurement. For setup inside SEO Booster, follow the [llms.txt docs](/docs/tools/llms-txt/). For the broader playbook, return to [AI search readiness for WordPress](/seo/ai-search-readiness-wordpress/).