Agent Experience for WordPress — make your site agent-readyYour website has visitors
that don’t use a browser.

What an agent-ready website actually is

An agent-ready website is one that an AI agent can read, understand and act on without a human in the loop. That sounds like a restatement of accessibility, but the constraints are different. A human visitor arrives with context, scans a page visually, tolerates ambiguity and clicks around until something makes sense. An agent arrives with a task, fetches whatever the server returns, and has one or two chances to find what it needs before it moves on to a competitor that was easier to parse.

In practice this comes down to three things. The site has to be discoverable — there must be a known path where an agent can find a description of what the business does and which resources exist. It has to be legible — the content that matters must survive without JavaScript execution, because most crawlers fetch raw HTML. And it has to be actionable — if there is something an agent could reasonably do on your behalf, such as request a quote or check availability, there needs to be a defined way to do it rather than a form that only works under a mouse pointer.

Most WordPress sites fail on all three, not because of bad engineering but because nobody ever specified the requirement. The site was built for people, and it does that job well. AX Protocol adds the machine-facing layer on top without changing anything a human sees.

AX, AEO and SEO — how they differ

These three terms get used interchangeably and they should not be. SEO optimises for a search engine that ranks pages and sends a human to click one of them. The unit of success is a session. AEO, answer engine optimisation, optimises for a system that reads several sources and synthesises a single answer. The unit of success is a citation — your name appearing in an answer that the user may never click through from.

AX, agent experience, is a step further out. It optimises for an agent that has been given a task and needs to complete it. The unit of success is task completion: the agent found your site, understood what you offer, and either handed the user a correct recommendation or performed an action against your endpoint. SEO and AEO are about being seen. AX is about being usable.

The practical consequence is that the three do not compete for the same fixes. Better titles help SEO. Clear, self-contained paragraphs help AEO. A machine-readable manifest and defined endpoints help AX, and nothing else does. A site can rank on page one and still be unusable to an agent — that combination is currently the norm rather than the exception.

What agents actually get when they fetch your site

The gap between what a site looks like and what an agent receives is usually much larger than owners expect. Page builders emit deeply nested markup where the visible text is a small fraction of the payload. Content loaded after the first paint — tabs, accordions, product listings, anything behind a filter — is frequently absent from the raw HTML entirely. Consent banners and bot-protection layers return interstitials to clients without a recognised browser fingerprint, so the agent receives a challenge page rather than the content.

None of these are visible in analytics, because JavaScript-based analytics cannot record a client that never executes JavaScript. This is why so many site owners are convinced they have no agent traffic at all: they are measuring with an instrument that is structurally incapable of seeing it. Server logs tell a very different story, and on most business sites the share of non-browser requests has been climbing steadily.

The fix is rarely a rebuild. It is usually a handful of targeted changes: make the key content present in the initial response, expose a manifest at a predictable path, and stop blocking clients that identify themselves honestly.

What the AX Score measures

The AX Score is a single number that summarises how well a site performs against the checks above. It is deliberately not a marketing metric. It combines discovery — whether the manifest, llms.txt and sitemap exist and are internally consistent — with legibility, which tests how much of the meaningful content survives a fetch without script execution, and with access, which checks whether well-behaved agents are being served content or blocked.

The score is most useful as a change detector rather than a badge. A site that drops fifteen points overnight has almost always had a caching layer, a firewall rule or a plugin update change its behaviour, and catching that within a day is worth far more than the absolute number. Compare your site to itself over time, not to someone else.

Who this is for

AX Protocol is aimed at sites where being found and correctly described by an AI system has commercial consequence: professional services, B2B software, ecommerce catalogues, publishers, and any business whose buyers have started asking a model for recommendations before they open a search engine. If your customers are researching with an assistant, the assistant is now part of your funnel whether or not you have accounted for it.

It is less relevant to sites with no public commercial surface — internal tools, private portals, or brochure sites that exist only to satisfy an existing customer who already knows the URL. If nobody is asking a model about your category, there is nothing yet for the protocol to improve.