Why it matters

Your next customer may send an assistant first.

People are starting to hand the job to an AI assistant: find a business, compare it, book it, buy it. When that assistant reaches your website, it either finishes the job or moves on.

Search sent you a visitor. An agent arrives with a task.

A person from search

Lands on your page, looks around and works out where to click. A confusing form slows them down, but they can usually push through.

An AI agent

Arrives with a job, like “book a cleaning for Tuesday” or “get a quote for a new roof.” It reads the page, clicks what it understands and reports back. If it can’t find the way through, it says so, and the person picks a business that works.

It is already happening.

Assistants act, not just answer

AI assistants can now open a website in a real browser and click through it for a person, the way a customer would.

Booking platforms are plugging in

As of November 2026, Fresha bookings work inside ChatGPT and Claude, and Booksy has an app in ChatGPT.

The web’s own tools are adapting

Google’s Lighthouse, the engine behind PageSpeed scores, now audits agentic browsing, and Chrome is trialling WebMCP, a way for pages to hand agents named actions.

What stops agents

The same few things turn agents away.

  1. Bot walls and CAPTCHAsSecurity that can’t tell a customer’s assistant from an attacker turns both away.
  2. Sign in before you can finishAn account or login in the way of booking or buying ends the journey for an assistant.
  3. Widgets inside another company’s frameBooking, quote and checkout tools that live on a different site are where agents most often lose their way.
  4. Fiddly inputsCustom date pickers, filters and dropdowns that only work with a mouse.
  5. Things in the wayPop-ups, cookie banners and chat bubbles covering the button the customer needs.
  6. Controls that aren’t what they seemButtons that aren’t real buttons, and boxes with no label saying what they want.

When our agent runs into one of these, your report shows where it happened and what we saw.

What good looks like

An agent ready site, in six parts.

Getting in

Can an agent reach your pages, or does something turn it away at the door?

Journeys

The ways a customer converts, and whether an agent can follow them.

Forms

Can an agent tell what each box wants and fill it in for the customer?

Understanding

Can an agent work out what you sell, what it costs and who you are?

Agent tools

Machine-readable ways in: WebMCP tools, llms.txt and discovery files.

Privacy

What your pages send to third parties when an agent visits.

Every test checks all six, with evidence. See the fixes we recommend most.

Good for agents is good for people.

Clearer for everyone

Real labels and real buttons help people using screen readers, and everyone filling in a form on a phone.

Fewer dead ends

A journey an agent can finish is one a rushed customer can finish too.

Measured, not guessed

You see the exact step where an agent stopped, so you fix the right thing first.

See what an agent sees on your site.

Two minutes, no account, one link to share with your team.

Test my site