Agent Readiness
Published · 4 min read
In 1986 the BBC recorded a national survey of Britain onto LaserDisc for the Domesday Project. Within roughly fifteen years the discs were physically fine and the data was intact, but almost nobody could open it. The hardware and software that made the format legible had disappeared. A university preservation effort, CAMiLEON, eventually rebuilt the means to read it. Jeff Rothenberg had already named the problem in Scientific American in January 1995: digital documents outlive the machines that can interpret them.
Agent extractability is whether an AI agent, a shopping assistant, a research tool, a crawler summarizing the page for a person, can pull the key facts off a page cleanly: pricing, what the product does, who it's for, proof it works. A page can read perfectly to a human and still score low on extractability if the substance is trapped in an image, a JavaScript-only render, or a carousel an agent never triggers.
Rothenberg's point transfers directly. Readability is a property of the container, not the text. A price that exists only inside an interactive calculator is in the same position as that survey data: present, correct, and unreadable to the thing trying to read it.
Pricing is where extractability breaks most often. Rendered as static text or an HTML table, a price is trivially readable. Rendered only inside a calculator that requires a seat count and a plan toggle before it produces a number, the price does not exist in the page an agent parses. The same pattern applies to feature claims shown as icon rows with no text label, and value propositions that live entirely inside a hero image.
Find out why your page loses people
Drop a URL. Get the first audit free when Nudgent opens.
Move the core facts into static, visible text: pricing, what the product does, who it's for, and your proof points. Give text alternatives to anything communicated only through icons or infographics. Check that content behind tabs and accordions also exists in the underlying markup, not just after a click event. And check what you are serving to crawlers at all, which is a separate failure from how you render.
That last one caught us. Our own robots.txt blocked the path serving the markdown export that our report page advertises, and the site shipped with no llms.txt. The substance was sitting right there behind a closed door.
Research assistants, comparison tools, and browsers acting on someone's behalf increasingly form purchase-relevant recommendations without any human reading your raw page. This is not the same failure as poor crawlability. A page can be indexed, ranked, and completely opaque to an agent asked what the product costs. Extractability is one of the seven agent-plane dimensions in Nudgent's scoring method, and an audit flags the specific places where substance is trapped.
Open your own pricing page with JavaScript turned off. Is the price in there?
Agent extractability is whether an AI agent can pull the key facts off a page cleanly: pricing, what the product does, who it's for, and proof it works. It measures the container, not the quality of the writing.
Yes, and this is the common case. An interactive pricing calculator can be excellent for a human visitor while producing no parseable price for an agent. The information is present in the experience and absent from the document.
No. Traditional SEO asks whether a crawler can reach and index the page. Extractability asks whether the facts on that page can be lifted out and used. A page can pass the first test and fail the second, which is why the two get audited separately.
Find out why your page loses people
Drop a URL. Get the first audit free when Nudgent opens.
Get early access