Buyers ask AI which brand to choose. Your dashboards never see the answer.

Your measurement stack was built around clicks. Web analytics reports sessions on your site. Search console data reports impressions and clicks on a results page. Rank trackers report positions on that page. Each does its job well, and you should keep them. They all start from the same event: a buyer saw a link or clicked one. An AI answer is a written recommendation, and none of them records how often it is given or what it says.

Two shortcuts don't close the gap either.

Referral tracking shows visits that arrived from an assistant through a link. It can't show the answers where you were left out, the answers that named you without a link, or what those answers said about you.

Manual spot checks give you one answer to one question on one day. AI answers vary from run to run, so a single check can't tell you whether you are ahead of a competitor or within the normal variation.

There is a deeper issue. Much of what an assistant says comes from what the model learned in training: old reviews, forum opinions, and facts that have since gone out of date. Our read, from two decades inside the SEO and content playbook, is that content written to rank is in that mix too. That knowledge doesn't change when your rankings improve, and it still shapes answers when the assistant searches the web. To know what buyers are told, you have to measure what the models already hold about your market, separately from what a search returned today.

AI's version of your market is what buyers are told. Where it differs from what is true, or from your own research, the AI version wins over time.

Keep your SEO partner and your tools. The AI Readiness Check measures the part of the buyer journey they were not built to see. If you fund AEO work elsewhere, the check gives it a baseline.

A fixed panel of leading consumer AI models is asked a large set of realistic buyer questions about your market, each need in several phrasings, and a sample of questions is re-asked. Results are built on 7,000+ analyzed data points.

Whether AI recommends you unprompted. Asking AI what it knows about you measures recall. The check asks the questions customers ask without your name, and records whether you come up. You see where your brand shows up in AI answers, topic by topic, and where it doesn't. You also see whether this differs from one model to the next, because buyers don't all use the same assistant.

Who AI recommends instead of you. When you don't come up, you see which brands AI names instead. Share of voice covers the brands that own the answers you're missing, including named competitors you add. When the difference between two brands is too small to be real, the report says "Too close to call." It won't invent a ranking the data doesn't support.

Perception. What AI models say about your brand, set out so you can judge whether it's right. Results can differ from your own customer research. That difference matters: where AI's version of your brand differs from reality, AI's version is what buyers are being told.

Machine access. Whether AI models can reach, read, and extract your website and product information: structured data, machine-readable specs, crawlability, and llms.txt (reported for information). Every failing check comes with fix text.

Small differences are not shown as leads. The check uses statistics so that a difference too small to be real is marked "too close to call" instead of being shown as a lead. The model panel is pinned to dated model snapshots and a slice of questions is deliberately re-asked, so a re-scan measures movement in the models rather than drift in the method. Machine access is a check list where each item is marked Pass, Opportunity, or Gap. When you report on this internally, you can show which differences hold up.

A report where each section answers one question a marketing lead would ask.

1. Verdict: Where do we stand, in one score and a plain verdict?
2. Machine access: Can AI models read our website, and what do we fix first?
3. Visibility map: Where do we show up in AI answers, and where don't we?
4. Model split: Does this differ model by model?
5. Competitive picture: Who owns the answers we're missing?
6. Whitespace: Which buyer topics does no brand own yet?
7. Perception: What does AI say about us when it mentions us?
8. Monday plan: What do we do first?
9. Methodology footprint: How was the scan run, and can a re-scan be compared with this one?

The Visibility map, Competitive picture, and Perception sections hold the numbers your dashboards are missing. The Methodology footprint shows how the scan was run, so a re-scan measures movement in the models rather than drift in the method.

The Monday plan is one ranked to-do list. Each task has an owner (whoever runs your website, owns your content, or owns your positioning) and links to its evidence. Give it to your own team or any agency. Pure Neo measures; we don't sell content, backlinks, or GEO packages.

Your results are never published. They belong to your brand.