The New SEO: AI Recommendations
Traditional SEO gets you ranked on Google. You show up in search results when someone types a query. That still matters -- it's not going anywhere.
But there's a new channel that most businesses are completely ignoring: AI recommendations. When someone asks ChatGPT "who does the best vehicle wraps in Utah?" or tells Claude "find me a plumber near Salt Lake City" -- those AI models are reading websites, evaluating businesses, and making recommendations. Right now. Today.
The businesses that show up in those AI answers aren't random. They're the ones whose websites are structured in a way that AI can actually read and understand. And right now, most websites are practically invisible to AI.
Here's why: AI models don't see your website the way a human does. They don't see your beautiful hero image, your animated transitions, or your pop-up newsletter signup. They see raw HTML -- a mess of divs, scripts, tracking pixels, and buried content. Most of the actual information about your business is lost in that noise.
The 3-file system fixes that. It gives AI models a clean, structured way to understand exactly who you are, what you do, and why they should recommend you.
The 3-File System
Three files. That's all it takes. Each one serves a different purpose, and together they make your business readable, citable, and recommendable by every major AI platform.
File 1: llms.txt -- The Executive Summary for AI
This is the simplest and most impactful file. llms.txt is a plain text file that sits at the root of your website (yoursite.com/llms.txt) and tells AI models everything they need to know about your business in one clean document.
Think of it as a cover letter for AI. When ChatGPT or Claude crawls your site, the first thing they look for is a structured summary. If they find llms.txt, they get your business name, what you do, where you operate, your services, your differentiators -- all in a format that's designed to be parsed, not rendered.
What goes in it:
- Your business name and one-line description
- Location and service area
- Core services with brief descriptions
- Key differentiators -- what makes you different from competitors
- Links to your most important pages
- Contact information
The format is intentionally simple. No HTML. No JavaScript. No styling. Just clean, structured text that any AI can parse in milliseconds. The llms.txt standard was proposed specifically for this purpose -- giving AI a reliable entry point to understand a business.
For the deep dive on creating your llms.txt, see the dedicated llms.txt guide.
File 2: Markdown Mirrors -- Clean Versions of Every Page
This is the file that made the video go viral. Markdown mirrors are clean, readable versions of every page on your website -- written in Markdown (.md files) and placed alongside the HTML versions.
Here's the problem they solve: when an AI reads your "About" page, it has to wade through navigation HTML, footer code, script tags, tracking pixels, image elements, CSS classes, and dozens of other things that have nothing to do with your actual content. By the time it extracts the useful information, context is lost and important details get missed.
A markdown mirror gives AI a clean alternative. Same content. Zero noise. When an AI visits yoursite.com/about/index.md, it gets pure content -- headings, paragraphs, lists -- with no HTML wrestling required.
What this looks like in practice:
# About Summit Wraps
Summit Wraps is a full-service vehicle wrap and branding company
based in Springville, Utah with a second location in Lehi.
## What We Do
- Full vehicle wraps (commercial and personal)
- Fleet branding and graphics
- Partial wraps and decals
- Paint protection film (PPF)
- Window tinting
## Service Area
Utah County, Salt Lake County, and surrounding areas.Compare that to the 200+ lines of HTML on the actual page. Same information. 10x more readable for AI.
I built a script that generates markdown mirrors for all 27 pages on the Summit Wraps website automatically. Once it's set up, any time you update a page, you regenerate the mirror in seconds. For the detailed walkthrough, see the markdown mirrors guide.
File 3: Schema Markup -- Structured Data for Google AND AI
Schema markup is JSON-LD code embedded in your website's HTML that tells search engines and AI models structured facts about your business. It's the most established of the three files -- Google has used it for years -- but most businesses either don't have it or have it set up wrong.
What schema tells AI:
- LocalBusiness schema: your business name, address, phone, hours, service area, price range
- Service schema: every service you offer, with descriptions and any associated pricing
- Review schema: your customer reviews and aggregate rating
- FAQ schema: common questions and answers about your business
- Organization schema: your brand identity, logos, social profiles
Schema is the bridge between traditional SEO and AI SEO. Google uses it to generate rich snippets (those star ratings and business info boxes in search results). AI uses it to extract structured facts. When ChatGPT says "Summit Wraps is rated 4.9 stars with 50+ reviews" -- that data came from schema markup.
The key with schema is specificity. Don't just add a generic LocalBusiness schema and call it done. Add Service schemas for each service. Add FAQ schemas for every common question. Add Review schemas with real customer testimonials. The more structured data you provide, the more confidently AI can recommend you.
Why This Works
AI models are trained on web data, but they access live websites through crawling -- the same way Google does, but with different priorities. Google cares about links, authority signals, and keyword relevance. AI models care about clarity, structure, and information density.
When an AI crawls a website that has:
- An llms.txt file summarizing the business
- Markdown mirrors providing clean content for every page
- Schema markup providing structured facts
...it can build a complete, accurate understanding of that business in seconds. It knows what you do, where you do it, what makes you different, what your customers say about you, and what services you offer. That's everything it needs to confidently recommend you.
Contrast that with a typical website: AI has to parse through navigation menus, JavaScript-rendered content, cookie banners, chat widgets, and hundreds of lines of irrelevant HTML to maybe extract some useful information. Most of the time, it gives up or gets it wrong.
The 3-file system doesn't replace your website. It augments it with AI-specific entry points. Your human visitors still see the beautiful, well-designed website. AI visitors get the clean, structured version they need to understand and recommend you.
The Summit Wraps Results
This isn't theory. Here's what happened after implementing all three files on the Summit Wraps website:
- ChatGPT started recommending Summit Wraps when users asked for vehicle wrap companies in Utah. Not because we paid for placement -- because the AI could actually understand our website.
- AI crawlers are actively reading the site. We can see them in our server logs -- OpenAI, Anthropic, and other AI companies regularly crawling our llms.txt and markdown mirrors.
- Organic traffic increased significantly -- not just from Google, but from people who found us through AI recommendations and then visited the website directly.
- The video explaining this system went viral. Over 1.5 million views across TikTok and Instagram. Thousands of business owners watching because this solves a real problem they didn't know they had.
The most telling metric: before implementing the 3-file system, asking ChatGPT "who does vehicle wraps in Utah?" would return generic results from national directories. After implementing it, Summit Wraps appears in the recommendations consistently -- with accurate information pulled directly from our structured content.
How Traditional SEO and AI SEO Work Together
This is important: traditional SEO and AI SEO are not competing strategies. They're complementary. Doing one makes the other more effective.
Schema helps both. Google uses schema for rich snippets and search features. AI uses schema for structured facts. Adding comprehensive schema to your website improves your Google rankings AND your AI recommendations simultaneously.
Good content ranks everywhere. A well-written service page with clear descriptions, real examples, and specific details ranks on Google AND gets quoted by ChatGPT. The markdown mirror just makes it easier for AI to extract that content without the HTML noise.
Authority compounds. When Google ranks your website highly, AI models treat that as a trust signal. When AI recommends your business, users visit your website, which increases traffic, which improves Google rankings. It's a flywheel.
The businesses that will dominate the next 5 years are the ones doing both. Google SEO gets you found when people search. AI SEO gets you recommended when people ask. Together, they mean you show up everywhere your customers are looking.
Who Should Do This First
The 3-file system works for any business with a website, but some businesses will see outsized results:
- Local service businesses -- plumbers, roofers, HVAC, landscaping, auto shops. When someone asks AI "who should I call for [service] in [city]?" you want to be the answer.
- Professional services -- agencies, consultants, lawyers, accountants. AI recommendations carry enormous weight for high-consideration purchases where people want a trusted referral.
- E-commerce with niche products -- if you sell something specific that people ask AI about ("best handmade leather wallets" or "organic baby formula"), structured content makes the difference between being recommended and being invisible.
- Any business competing in a local market -- most of your competitors haven't heard of AI SEO. Being the first in your market to implement it gives you a window of competitive advantage that's hard to close once established.
The businesses that DON'T need this: if you only get customers through word-of-mouth or a closed network, AI SEO won't change much. But if any of your customers might ask an AI for recommendations in your category -- and increasingly, they do -- this is your opportunity.
The Implementation Order
If you're implementing the 3-file system, here's the order that gets you results fastest:
1. Schema Markup First
Most businesses already have some schema markup (your web developer or site builder may have added basic LocalBusiness schema). Start by auditing what you have and filling the gaps. Add Service schemas, FAQ schemas, and Review schemas. This improves both your Google presence and your AI readability immediately.
Test your existing schema at Google's Rich Results Test. It will show you what Google (and by extension, AI) currently knows about your pages.
2. llms.txt Second
This takes 15-30 minutes. Write a clean summary of your business in the llms.txt format, upload it to your website root, and you're done. Immediate impact because AI crawlers check for this file specifically.
The format is simple enough to write by hand, but if you want it done precisely, describe your business to Claude Code and it will generate a perfect llms.txt file in under a minute. See the llms.txt guide for the exact format.
3. Markdown Mirrors Third
This is the most work upfront but the highest long-term value. Every page on your site gets a clean .md version. You can write these by hand (copy your page content, strip the HTML, format as Markdown) or automate it with a script.
I built a Python script that reads every HTML page on the Summit Wraps site and generates a clean markdown mirror automatically. It runs in under a minute and produces mirrors for all 27 pages. Once it's set up, regenerating mirrors after a content update takes seconds. See the markdown mirrors guide for the full walkthrough.
Where to Go From Here
The 3-file system is your foundation for AI visibility. Here's how to go deeper:
- llms.txt deep dive -- read the complete llms.txt guide for the exact format, examples, and best practices
- Markdown mirrors deep dive -- read the complete markdown mirrors guide for the technical walkthrough and automation script
- Learn to build these systems yourself -- read Vibe Coding for Business to understand how to use Claude Code to implement everything
- See the full automation picture -- read How to Automate Your Business to understand where AI SEO fits in the broader system
- Get the templates -- join the free Skool community for the llms.txt template, mirror generator script, and schema examples
Every guide on this site gives you the full what and why. The community gives you the how -- the exact files, the prompts, the scripts. No paywalls. No upsells. Just the playbook that's already working.