Is Your Website Agent-Ready? A Practical Checklist
A growing share of web traffic isn't a person clicking through search results anymore — it's an AI agent doing it on their behalf. ChatGPT browses. Claude fetches pages and calls tools. Perplexity answers questions by reading and citing sites directly. Shopping assistants, research assistants, and internal company agents are all, in effect, a new category of visitor to your website.
Most business websites were never built with that visitor in mind. They're built for a human with eyes and a mouse — heavy client-side rendering, content that only appears after a click, no machine-readable summary anywhere, no documented way for software to actually do something on the site rather than just look at it. To an AI agent, a lot of the modern web is either unreadable, ambiguous, or effectively invisible.
"Agent-ready" just means: can an AI agent reliably read your site's content, understand what your business does, and — ideally — take a defined action on it? Here's the checklist we actually use.
1. A machine-readable summary (llms.txt)
A plain-text file at /llms.txt summarizing who you are, what you offer, and where the important pages are. It's the fastest way for an agent to understand your site without having to parse a fully rendered page. Ours is here if you want to see the shape of it.
2. Content an agent can actually parse
If your key content only exists as text baked into an image, or only appears after JavaScript runs a fetch call an agent won't trigger, it may as well not exist. Content should be present in the page's HTML — even content that's visually hidden behind a tab, an accordion, or (in our case) a window that isn't "open" yet — rather than something a script has to construct from scratch client-side. This site is a good stress test of that: everything inside these little desktop windows is real HTML from page load, not something JavaScript writes in afterward.
3. A robots.txt that doesn't quietly block the agents you want
A lot of default robots.txt files block crawlers by name without anyone realizing GPTBot, ClaudeBot, or PerplexityBot were in that list. If you want to be found through AI answer engines, your robots.txt needs to say so explicitly, not just avoid saying no. Worth checking what yours actually allows — here's ours.
4. A documented way for an agent to take action, not just read
Reading is the first layer. The next is letting an agent actually do something — check pricing, request a consultation, run a tool — without scraping a form and guessing at field names. That's what an MCP server is for. We wrote a separate post on that, since it deserves its own explanation.
5. Clear, unambiguous contact and action endpoints
A real email address in the HTML (not just behind a "Contact us" button that opens a JavaScript modal an agent won't trigger), a real mailto: link, a documented API or MCP endpoint if you have one. Agents work better with explicit, addressable actions than with UI they have to reverse-engineer.
6. Sitemap and structure that reflect what's actually on the site
A sitemap.xml that's stale or lists only the homepage tells an agent (and a search engine) less than the site actually contains. If you add pages, add them to the sitemap — it sounds obvious and gets skipped constantly.
Curious where your own site actually stands? We built a free scanner that checks a URL against this list — and a few more — and gives you a Level 0–4 verdict with specific fixes.
Run the Agent-Ready Scanner →None of this is exotic. It's mostly the same discipline as basic technical SEO, pointed at a newer kind of visitor. The sites that get this right now are going to be the ones agents actually recommend, cite, and transact with over the next few years — the same way ranking well on Google mattered for the last twenty.