Skip to content

field notes · ai and agents

Agent-ready web

Agent-ready webAn isometric drawing of one web page pulled apart into five stacked layers. At the top, the page people see: a header bar, a picture, text and a book button. Below it the HTML, with a booking form of three fields (name, date, email) and an outline of its parts. Then a schema.org structured-data card, then a folded llms.txt sheet, and at the bottom a door marked api and mcp for actions. On the ground beside the stack stand a robots.txt signpost, an AI crawler with a lens, and a browsing agent. Send a crawler and it checks robots.txt, reads down through the layers and comes back with a quote. Ask the agent to fill the form: with tidy markup off the fields have no labels and the agent stalls; with it on, the agent fills the form and the booking goes through the API door.ai visitors00 actionsapi · mcprobots.txt01 site mapllms.txt02 structured dataschema.org{ json-ld }03 semantic html<div><div><div>04 the pagebook1234567

One web page, pulled apart into the layers software reads. Tidy markup is off: the form fields have no labels. Send a crawler, or let an agent try the form.

AI assistants can read your site, quote it, and use your forms.

Book a free first callEstimate the cost

In short

Agent-ready web is work on an existing website so that ChatGPT, Perplexity and AI agents can read it and use it. It's for businesses that have noticed customers asking an AI before they search. We're Zenoware, a small studio in Auckland; we've done this on products of our own, with MCP servers we built, and for clients in New Zealand and overseas. Expect structured data, tidier HTML on forms and, where it helps, an API. The first call is free.

What you get

  • Pages that ChatGPT and Perplexity can quote accurately

  • Forms an agent can fill in without guessing

  • Bookings or orders open to agents through an API

How it runs

  1. 01

    Audit

    We start by checking your markup, robots.txt and what AI crawlers can actually reach.

  2. 02

    Structure

    Next comes schema.org markup and an llms.txt file (a short Markdown map of the site).

  3. 03

    Forms

    Forms get proper labels and stable selectors, so an agent filling them in doesn't stall.

  4. 04

    Actions

    Last, key actions go behind an API or MCP server, and we watch the crawler traffic.

Who you'll work with

Brush sketch of JP Caparas

JP Caparas

Founder and lead engineer

Thirteen years of principal engineering, including development lead at Public Trust. OpenAI Codex and Devin ambassador in New Zealand. You talk to the person who writes the code.

About the team

Questions

What is agent-ready web design?

It means building a site that software can use as well as people can. AI assistants need clear text and structured data to quote you. Browsing agents need labelled forms and buttons that don't move around. And if you want agents to book or buy, they need an API they can call. Most of it overlaps with good accessibility.

What is AEO?

Answer engine optimisation. It's SEO aimed at the answer you get from ChatGPT, Perplexity or Google's AI Overviews rather than a list of links. Mostly it comes down to plain headings, a direct answer near the top of each page, schema.org markup, and not blocking the crawlers that feed those answers.

What is an llms.txt file?

A plain Markdown file at the root of your site (yoursite.com/llms.txt) that tells a language model what the site covers and which pages matter most. It's a proposed convention, and not every AI company reads it yet. We add one anyway, next to proper structured data, since it's cheap and does no harm.

Should I block AI crawlers in robots.txt?

Depends what you're selling. Blocking training crawlers keeps your content out of future models, but block the search and assistant crawlers too and you'll drop out of AI answers. OpenAI, Anthropic and Google each publish separate robots.txt names for these jobs, so you can let some in and keep others out.

How long does it take to make a website agent-ready?

Five weeks, on our standard roadmap. We'd spend the first one auditing, and that's when we'd confirm the scope. After that it's roughly a week each for structured data, the interface and the APIs, with the final week kept for testing and monitoring.

You might also need

Tell us what isn't working.

The first call is free, and we'll tell you straight whether we're the right people for it.

Start a project