AX vs AEO: What's the Difference?

AX, agent experience, is how well an AI agent can use your site: reach pages, read them and, where it is allowed, complete a task. AEO, answer engine optimisation, is the work of being cited or recommended in the answers AI assistants give. AX asks whether an agent can get in. AEO asks whether the answer names you once it has.

You need both, in that order. An assistant cannot cite a page it cannot read, and a readable page still earns nothing if the answer goes to someone else.

Key Takeaways

• AX is about access and usability for agents. AEO is about outcomes in AI answers.
• AX failures are usually technical and fast to fix: a robots.txt rule, a firewall setting, a page that only renders in a browser.
• AEO work is slower: the content, the independent sources that mention you and how consistent they are.
• A block is an AX problem that shows up as an AEO symptom. Fix the block before writing more content.
• Neither guarantees a citation or a recommendation.

What Each One Covers

	AX (agent experience)	AEO (answer engine optimisation)
Question	Can an agent reach and use my site?	Do AI answers mention or cite me?
Who it serves	Agents acting for a person	The person reading the answer
Typical work	robots.txt, firewall and bot rules, readable pages, clear navigation	Answer-first content, independent sources, consistent facts about the brand
How you check it	Fetch the site as each agent and see what comes back	Ask a fixed set of buyer questions and record who is named
Typical failure	A CDN refuses ChatGPT-User	A competitor is named in every answer and you are not

Why They Are Different

An assistant answering a question can use two sources. It can use what it already learned in training, and it can fetch pages live. The fetch is where AX matters. Each vendor documents separate agents for it: OpenAI's OAI-SearchBot and ChatGPT-User, Anthropic's Claude-SearchBot and Claude-User, and Perplexity's PerplexityBot and Perplexity-User. Which crawlers do what is the first thing to settle.

AEO covers everything after the fetch: whether your page answers the question, whether other sources agree with it, and whether the assistant picks you over alternatives. A site can pass every AX check and still be absent from answers, because nothing makes its claims stand out.

How They Relate

• AX is a precondition for AEO. It gates the fetch, not the choice.
• Blocks show up late. A refused agent does not produce an error anyone sees. The page is just missing from the answer, and the cause looks like an AEO problem.
• AX can be checked in minutes. AEO needs repeated readings, because a single answer is an example, not a trend.
• Agents also act. Beyond reading, some agents sign up, fill forms or call tools. That is the part of AX that has no AEO counterpart.

Which to Fix First

1. Run an AX check. The free agent access check reports which agents robots.txt and your firewall let in, and names the rule behind each block.
2. Fix any block on a search or user-triggered agent.
3. Measure AEO: pick buyer questions, run them across assistants on a schedule, and record who is named. AI share of voice explains how to read the result.
4. Work on the gaps the measurement shows.

The score view of the AX side is covered in agent readiness score.

What We Do Not Claim

Making a site readable to agents does not guarantee that any assistant will cite it. Lectern's free check also shows what a request from our servers got, not what the real assistant gets. See pricing for the plans that track AI answers over time.