Agent Experience Optimization (AX)

Agent experience optimization is the practice of designing a product, website or API so that AI agents can discover what it offers, understand how to use it, get access and complete tasks on a person's behalf. It applies the idea of agent experience, usually abbreviated AX, to marketing and product work. If user experience is about people using a product and developer experience is about developers building on it, agent experience is about AI agents doing either.

The term AX was introduced by Mathias Biilmann, CEO and co-founder of Netlify, in a blog post titled Introducing AX: Why Agent Experience Matters on January 28, 2025. He placed it alongside 'user experience', which Don Norman coined in 1993, and 'developer experience', coined by Jeremiah Lee in 2011, and defined AX as 'the holistic experience AI agents will have as the user of a product or platform'. His example was Netlify's ChatGPT integration, which lets an agent deploy a site that the user can later claim, and which he said was creating more than 1,000 sites a day.

In marketing writing the practice is sometimes abbreviated AEO, which collides with answer engine optimization. AX avoids the confusion and is the term Biilmann and Netlify use.

Agent experience covers four areas:

• Discovery: an agent can find out what the product does and how to use it, from clear documentation, an llms.txt summary for tools that read one, an AGENTS.md file in code repositories, or a Model Context Protocol server.
• Access: an agent can get started without a person completing every step, for example by creating a trial resource that the user claims later, with a person approving anything sensitive.
• Operation: APIs, site tools and forms behave predictably, are well described and return error messages an agent can act on.
• Trust: prices, limits, policies and terms are stated plainly, so an agent can recommend or choose the product with confidence.

AX is broader than agentic search optimization, which focuses on agents that search the web and complete tasks on a website. AX also covers APIs, SDKs, developer platforms and services that agents call directly. It feeds agentic commerce when the task is a purchase.

The idea matters well beyond developer tools. Coding agents such as Claude Code and Codex choose which SDKs and services to use when they write code, which turns AX into a distribution channel for any company with an API. Our article on how AI coding agents choose SDKs looks at what influences those choices, and Promptwatch can track how often coding agents recommend a product.

Measure AX with task-based tests: give agents realistic goals, such as integrating your API, comparing your plans or booking a service, and record success rates and failure points. Add server logs of agent traffic and, for developer products, how often coding agents suggest your SDK unprompted.