Skip to content

For AI agents: a documentation index is available at /llms.txt

AI search, explainedPart 4 of 6

What is AXO? Agent experience optimization explained

AXO, or agent experience optimization, means making your website and products usable by AI agents that act for a person: reading your pages, filling in your forms, calling your tools and completing purchases. Where AEO and GEO get your brand named in AI answers, AXO makes sure that when an agent visits to get something done, it can.

6 min read

What is AXO in marketing?

AXO is the part of marketing that starts after the answer. A buyer can now ask an assistant to shortlist three vendors, check their pricing and book demos; the agent reads your site, compares you, and either completes the task or moves on to a rival.

AXO prepares your site for that visit. It's the marketing team's concern because a failed agent visit costs a shortlist place or a demo request, and the web team's because most fixes live in HTML, forms and site files. Our AXO services describe the work we do for clients.

Two other uses of the letters exist. "Agentic experience orchestration" is a customer-experience term about coordinating support bots, and some vendors use AXO for AI brand visibility, which is closer to GEO. This guide means agent experience optimization.

Where does "agent experience" come from?

The idea comes from Mathias Biilmann, co-founder of Netlify, who introduced "AX" in January 2025 as "the holistic experience AI agents will have as the user of a product or platform". It borrows from UX and DX. If agents are going to be users, someone has to design for them.

The tools for agents to act arrived in the same period. Google added "auto browse" to Gemini in Chrome in January 2026, Claude in Chrome became generally available on paid plans in August 2026, and Chrome opened a WebMCP origin trial, a draft way for a page to declare tools a browser agent can call, from version 149.

What are the four layers of AXO?

AXO has four layers: readable, operable, callable and observed. Each is a question you can test on your own site today.

LayerThe questionWhat to check
ReadableCan an agent read your pitch and facts?Copy and pricing in the HTML without JavaScript; structured data; robots.txt rules for AI crawlers; an llms.txt file and Markdown versions of key pages
OperableCan an agent finish a task?Forms with labelled fields; buttons with names; WebMCP tool markup on key forms such as demo requests
CallableCan an agent call your product?An MCP server or app, listed in the official MCP Registry, Claude's connector directory, ChatGPT's apps directory or Google's connected apps
ObservedCan you see agent visits?Server or CDN logs checked against published bot IP ranges, since GA4 misses bots that skip JavaScript

For online stores, agentic commerce adds a fifth concern: product feeds and checkout protocols that let an agent buy. That's the subject of our ACO work.

How is AXO different from AEO and GEO?

AEO and GEO are about being named and cited in AI answers; AXO is about what an agent can do once it arrives. You can win the first and lose the second: an assistant recommends you, then fails to book the demo because your form has no labels.

The work overlaps in the readable layer, since clean HTML and structured data help both. It splits after that. AEO and GEO work happens in content and on third-party sites; AXO work happens in forms, site files, tools and logs. Our guides to what AEO is and AEO vs GEO cover the answer side.

Is AXO the same as agentic engine optimization?

Not quite. Addy Osmani described "agentic engine optimization" in April 2026, also abbreviated AEO, as making content and docs usable by AI agents, especially coding agents. That matches AXO's readable layer. AXO also covers forms, tools, checkout and measuring agent visits, so it's the wider term.

How ready are websites for AI agents today?

Mostly not ready, beyond the readable basics. Cloudflare checked the top 200,000 domains and found Content Signals on 4% and MCP server cards on fewer than 15 sites.

Our own measurements agree.

  • We tested 33 agencies that sell AI search services. None exposed a form as a WebMCP tool, none published an AI catalog, and the median agent-readiness score was 51 out of 100. Full results are in our agent-readiness study.
  • Of 60 large US retailers we checked, 42% refused or blocked our self-identified agent, and only 27% served a product's price in structured data without JavaScript.
  • Of 250 company sites in SaaS, healthcare, finance and insurance, 43% published a real llms.txt, from 63% in SaaS to 20% in healthcare.

Copy is usually readable, but forms often load by script, few sites declare tools, and bot filters sometimes block agents by accident.

How do you measure AXO?

Measure AXO with checks that each have a clear outcome. We score each check Pass, Partial, Fail or Not applicable, and a missing optional file like llms.txt counts as Not applicable.

For operable tasks, pick 3 to 5 key tasks from your money pages, such as finding pricing or booking a demo, and run them with a browser agent. Each ends Completed, Reached the confirmation step, Partial or Blocked, with the step where it stopped. A careful test never submits the form, so it doesn't create fake leads.

For callable, check whether you're listed in each agent directory. For observed, read your logs and mark each bot visit as verified, mismatched or unverified against the operator's published IP ranges.

How do I start with AXO?

Start with five questions for your web team:

  1. With JavaScript off, does the text of your homepage, product pages and pricing appear?
  2. Does robots.txt set rules for AI crawlers, or block them by default?
  3. Do you publish an llms.txt and Markdown versions of key pages?
  4. Do your demo, trial and contact forms have labelled fields an agent can read?
  5. Are you reading your server or CDN logs for agent visits that GA4 misses?

A "no" or "not sure" on any of them is your first fix. Our free Agent Ready extension answers the first four for any page you open, and agent-ready websites explains the fixes in more detail.

Is AXO worth doing now?

The readable and operable layers are worth doing now, because they're cheap and they also help search crawlers and people using assistive tech. Labelled forms and HTML that loads without scripts are basic hygiene that many sites skip.

The callable layer depends on your product. If you have an API or a tool buyers would use from an assistant, an MCP server and directory listings put you inside ChatGPT and Claude; our MCP series covers how. WebMCP is still experimental, so add it to key forms as a low-cost bet rather than a requirement.

Questions

Questions buyers ask

What does AXO stand for?

AXO stands for agent experience optimization: making your website and products usable by AI agents that read, fill in forms, call tools and buy on a person's behalf.

What does "agent experience" mean?

Agent experience, or AX, is the experience an AI agent has as the user of a website or product, a term Netlify's Mathias Biilmann introduced in January 2025.

What is agent engine optimization?

"Agentic engine optimization" usually means making content and docs usable by AI agents, especially coding agents. It overlaps with AXO's readable layer; AXO also covers forms, tools, checkout and agent visits.

Does AXO replace SEO or GEO?

No. SEO gets your pages indexed, AEO and GEO get you named in AI answers, and AXO makes sure agents can act on your site.

Free AEO, GEO and AXO audit

See what AI says about you, and whether agents can use your site.

Enter your website. An operator checks the audit and sends it within 3 working days, and it is yours to keep.