AI agents are already browsing, comparing, and buying on behalf of real people. Most websites aren't ready. This guide shows you what to fix — and why it matters.
8 things AI agents look for on your site.
Traditional SEO optimizes for Google's index. Agent SEO optimizes for LLMs answering questions in real time. Google ranks pages. Agents extract facts. A porta-potty company in Tampa needs the agent to know they serve Tampa, what they charge, and that they're available — not just rank for "porta potty Tampa."
Here's what matters. Check off each item you've already handled:
GPTBot, ClaudeBot, and Applebot-Extended are not blocked. This is your front door — if it's locked, nothing else matters.
LocalBusiness, Product, FAQPage, or Organization schema tells agents who you are and what you sell — in their native language.
A plain-text summary of your business that agents read first. Think of it as a README for AI. llmstxt.org
"How much does a porta potty rental cost in Tampa?" with a direct answer on the page maps exactly to how users ask agents questions.
If your page is a blank shell until JS hydrates, agents see nothing. Server-rendered HTML with real text is what gets read.
Agents use sitemaps to discover pages. If your sitemap is stale, auto-generated junk, or missing entirely, pages get skipped.
Title tags, meta descriptions, and Open Graph data give agents a structured summary before they even read the body content.
ChatGPT's retrieval still leans heavily on Google's knowledge graph. A verified GBP with categories, service areas, and attributes feeds the data layer agents pull from.
Paste any URL. See the raw text an AI agent reads.
Your website has navigation bars, hero images, cookie banners, pop-ups, and a JavaScript bundle that takes 3.2 seconds to hydrate. A person sees a product page. An AI agent sees noise.
Agents don't render your CSS. They don't execute your JavaScript. They want the content — stripped bare. Try it yourself:
The most forward-thinking sites serve alternative representations of their content. When an agent requests a page with Accept: text/markdown, it gets a clean document instead of a bloated DOM tree. The llms.txt specification standardizes this — a machine-readable summary alongside your human-readable pages.
Copy-paste JSON-LD templates for the 4 most common site types.
JSON-LD structured data is the single biggest lever for agent visibility. Agents parse it natively. It turns a vague website into a set of structured facts an agent can cite, compare, and act on. Most sites have zero JSON-LD or only what their CMS auto-generates.
Replace the placeholder values with your real info. Paste the <script> block into the <head> of your page.
Best for: plumbers, rental companies, restaurants, clinics, any business with a physical service area. schema.org/LocalBusiness
Best for: Shopify stores, WooCommerce, any site selling products. Add to every product page. schema.org/Product
Best for: service pages, help centers, any page with Q&A content. Agents pull these answers directly. schema.org/FAQPage
The baseline. Every site should have this on their homepage. schema.org/Organization
Who's crawling your site — and why.
A new class of crawlers has arrived, and they aren't indexing your site for a search results page. They're reading it to answer questions, compare products, and make purchasing decisions on behalf of users. Here's who they are:
OpenAI — powers ChatGPT
Crawls sites to improve ChatGPT's responses and browsing. Blocking it means ChatGPT can't cite or recommend you.
Anthropic — powers Claude
Reads sites for Claude's knowledge and web search. Growing fast as Claude usage increases in enterprise and consumer.
Apple — powers Apple Intelligence
Feeds Apple's AI features across Siri, Safari, and Spotlight. Blocking it cuts you off from the entire Apple ecosystem.
Perplexity AI — answer engine
Powers Perplexity's real-time answer engine. Cites sources directly — if it can crawl you, you get attribution and traffic.
OpenAI — live browsing
The real-time browsing agent used when ChatGPT searches the web during a conversation. Different from GPTBot's batch crawling.
ByteDance — powers TikTok search
Feeds ByteDance's AI and search products. Increasingly relevant as TikTok becomes a primary search engine for younger users.
What to do: Open your robots.txt right now. Search for each of these bot names. If you see Disallow: / next to any of them, you're invisible to that agent. Our audit tool checks this automatically and generates a recommended robots.txt you can copy.
MCP, agent commerce, and what's coming next.
There's a protocol emerging that most businesses haven't heard of yet. It's called the Model Context Protocol, and it may become as important to AI commerce as HTTPS was to e-commerce.
MCP is a standardized way for AI agents to discover what a website can do — not just what it says, but what actions it supports. Can a user check out? Can an agent query inventory? Can it compare prices across variants? MCP publishes these capabilities in a machine-readable manifest at /.well-known/mcp.json.
Think of it as a secret handshake. When an AI agent lands on your domain, the first thing it looks for is that file. If it finds it, it knows what's possible. If it doesn't, it has to guess — and agents that guess move on to competitors that don't make them.
For local service businesses: AI assistants will book appointments, request quotes, and compare availability across providers — without the customer ever visiting a website. The businesses whose data is structured and accessible get routed to first.
For e-commerce: AI shopping agents are already comparing products, checking stock, and completing purchases. Sites with Product schema, clean pricing, and MCP manifests will be the ones agents recommend and transact through.
For B2B: Buyers increasingly use AI to research vendors, compare features, and build shortlists. If your positioning, pricing signals, and case studies aren't machine-readable, you don't make the list.
What to do now: You don't need to implement full MCP today. But you should understand it exists, start with structured data and clean content, and run an audit to know where you stand. The businesses that move early will set the standard everyone else follows.
Run a free audit on your site. See what AI agents see — and what they don't.
Run Your Audit See the Demo