A clean llms.txt setup takes about an hour and hands AI crawlers a structured summary of your business at the exact moment they decide whether to cite you. ChatGPT, Perplexity, Claude, and Google’s AI features all assemble answers from crawled content. Sites that are easy for machines to parse get summarized accurately and recommended. Sites that are not get skipped or misquoted.
This guide covers the full technical stack: what llms.txt is, the format spec, four ways to implement it on WordPress, which schema types AI engines actually parse, and how to keep your entity data consistent enough for a language model to trust it. We run this exact stack on our own site, and we deploy it for every client in our AI search optimization program.
- What llms.txt Is and Why AI Crawlers Want It
- The llms.txt Format Spec in 5 Parts
- llms.txt Setup on WordPress: 4 Working Methods
- Schema for AI Search: 6 Types That Get Parsed
- Entity Consistency: The Step Most Setups Miss
- How to Verify and Measure Your Setup
- llms.txt Setup FAQ
What llms.txt Is and Why AI Crawlers Want It
llms.txt is a plain Markdown file that lives at the root of your domain. Jeremy Howard of Answer.AI proposed it in September 2024 as a standard way to tell large language models which pages matter most and how to interpret a site. Think of it as the inverse of robots.txt. Where robots.txt tells crawlers where not to go, llms.txt tells AI systems exactly where to look first.
The problem it solves is context. AI models work with limited context windows, and a typical WordPress page buries a few hundred words of useful content under navigation menus, scripts, pop-ups, and footer widgets. A curated Markdown file strips all of that away and serves pure signal.
One honest caveat: llms.txt is an emerging standard, not a guaranteed ranking input. Crawlers from OpenAI, Anthropic, and Perplexity have been observed requesting the file, but no engine has committed to using it in a fixed way. The cost is near zero, and the exercise forces you to define exactly how you want machines to describe your business. Ours is live right now at urlocalmarketing.com/llms.txt, listing every service page, case study, and location we want AI systems to read first.
The llms.txt Format Spec in 5 Parts
The spec is deliberately simple. Everything is standard Markdown served as plain text, and only the first element is required.
- H1 title line. A single # heading with your business name. This is the only required element in the entire file.
- Blockquote summary. One or two sentences after a > character covering what you do, who you serve, and where you operate.
- Context paragraphs. Optional plain text with key facts: years in business, certifications, service model, guarantees.
- H2 link sections. Headings such as Services, Service Areas, Case Studies, and Contact, each holding links in the format [Page title](URL): one-line description.
- An Optional section. A final H2 literally titled Optional, holding secondary links a model can skip when its context budget runs tight.
There is also a companion convention, llms-full.txt, which packs the complete text of your key pages into one file. A local service business rarely needs it. Curation beats completeness here: a tight file of 30 to 60 links nearly always outperforms a 2,000-line dump.
llms.txt Setup on WordPress: 4 Working Methods
There are four reliable ways to handle llms.txt setup on a WordPress site, ranked from most control to least.
- Static file upload. Write the file in any text editor, save it as llms.txt with UTF-8 encoding, and upload it to your site root (usually public_html) through cPanel File Manager or SFTP. Full control, zero plugin overhead.
- SEO plugin generation. Yoast and several dedicated plugins can now generate the file automatically. Review the output carefully, because auto-generated lists love to include tag archives and thin pages you never want a model reading first.
- Custom endpoint. On managed hosts that lock down the document root, register a small rewrite rule or endpoint that serves the content with a text/plain header.
- Hybrid. Generate a draft with a plugin, export it, hand-edit every description, then lock it down as a static file. This is the llms txt WordPress workflow we use for client sites.
Verify before moving on. Load yourdomain.com/llms.txt in a browser, confirm it returns a 200 status as plain text with no HTML wrapper, and check that your cache or CDN is not redirecting the request. Then make updates part of your publishing routine: every new service page or location page goes into the file the same day.
Schema for AI Search: 6 Types That Get Parsed
llms.txt is the new half of the stack. Schema markup is the proven half. ChatGPT search leans heavily on Bing’s index, Google’s AI features build on Google’s entity graph, and both reward unambiguous JSON-LD. Six types do most of the work for local businesses.
- LocalBusiness. Name, address, phone, geo coordinates, hours, and areaServed. Follow Google’s local business structured data documentation field by field.
- Organization. Legal name, logo, and a sameAs array linking your Google Business Profile and social accounts. This anchors your entity.
- Service. One per core offering, with provider and areaServed defined, so models can match you to “service plus city” questions.
- FAQPage. Question and answer pairs in the same conversational shape AI answers use. The easiest content for a model to quote.
- Article. Author, datePublished, and dateModified on every post. Retrieval systems weigh freshness.
- BreadcrumbList. Maps how pages relate, which helps engines cite the correct deep page instead of your homepage.
Implementation is straightforward on WordPress: output everything as JSON-LD through your SEO plugin, validate with Google’s Rich Results Test, and keep one coherent schema graph per page. Duplicate, conflicting markup from theme builders is the most common error our audits find. Every site we build through our SEO services ships with a validated graph on every page.
Entity Consistency: The Step Most Setups Miss
Language models resolve your business as an entity, a single node connecting your website, Google Business Profile, review platforms, and directory listings. Every mismatch between those sources lowers a model’s confidence that they all describe the same company. Lower confidence means fewer citations.
The fix is mechanical. Use the identical business name, address format, and phone number in your schema, your llms.txt, your GBP, and your top directories. Point your sameAs array at every profile you control. Keep your one-line description consistent across your About page, your GBP description, and your llms.txt blockquote.
This is not theoretical. When we rebuilt structured data and standardized entity details for Mal Al Sham in El Cajon, search impressions grew 400 percent and online orders climbed 250 percent. Machines reward businesses they can verify.
How to Verify and Measure Your Setup
AI crawler optimization is measurable. Check server logs for hits from GPTBot, OAI-SearchBot, ClaudeBot, and PerplexityBot, and confirm they reach llms.txt and your key pages instead of bouncing off a firewall rule.
Then track outcomes monthly. Run your core “service plus city” prompts in ChatGPT and Perplexity and log every citation. Watch Bing Webmaster Tools and Search Console for impression growth. Re-validate schema after any theme or plugin update. Site speed feeds the same loop, since fast pages get crawled more often and parsed more reliably. Our A-K Tech build holds a 95+ PageSpeed score, and calls grew 150 percent on that foundation.
llms.txt Setup FAQ
Quick answers to the questions clients ask most before we run an llms.txt setup.
Does llms.txt setup actually improve AI search visibility?
Treat it as a low-cost, high-upside move rather than a guarantee. The standard is still emerging, but crawlers from OpenAI, Anthropic, and Perplexity have been observed fetching the file, and it pairs naturally with schema markup, which is proven. The whole job takes about an hour, so the risk-to-reward math is firmly in your favor.
Where exactly does the llms.txt file go?
At the root of your domain, so it resolves at yourdomain.com/llms.txt. It does not belong in a subfolder, a media library, or a page builder. On most shared hosting that means the public_html directory. If you run multiple domains or a multisite network, each domain needs its own file with its own curated link list.
How is llms.txt different from robots.txt?
They are opposites that work together. robots.txt is access control, telling crawlers which areas they may not enter. llms.txt is a content guide, telling AI systems which pages deserve attention first and how to describe your business. Keep both at your domain root, and never list a page in llms.txt that robots.txt blocks.
Do I also need an llms-full.txt file?
Usually not for a local service business. llms-full.txt holds the complete text of your key pages in one document, which suits documentation-heavy software companies. For contractors, restaurants, agencies, and other local operators, a tight standard file pointing at strong service pages and case studies does the job with far less maintenance overhead.
How often should I update my llms.txt setup?
Update it the same day you publish or significantly revise any page it references: new service pages, city pages, or case studies. Beyond that, review the file monthly for dead links and outdated descriptions. A stale file that points models at deleted pages erodes the exact trust you built it to earn.
Want the whole stack handled for you? UR Local Marketing builds llms.txt setup, validated schema graphs, and entity cleanup into one project for service businesses in San Diego and nationwide. Request a free local visibility audit and we will show you exactly what AI engines see when they read your site. Call 888-252-0251, Monday through Friday, 9am to 6pm Pacific, or reach the team anytime through our 24/7 support page.
