Working out whether AI assistants would ever recommend a brand, before spending a quarter on content

Buyers increasingly get an answer instead of a results page. Teams were still measuring rankings, which no longer described whether they were present in the moment a decision was formed.

Decide where to invest attention next quarter with some evidence that the investment would show up in AI-mediated discovery.

Most AI visibility audits stop at rankings or surface-level brand mentions. They do not show what a brand is associated with, how competitors own the conversation, or whether an assistant is likely to recommend it at all. Doing that manually took four to eight hours per brand, so it rarely happened before decisions were made.

Visibility in AI answers is an entity and coverage problem, not a keyword problem. A model recommends what it can reliably associate with a job, a category and a set of attributes. If those associations are missing, no amount of publishing volume fixes it.

• Reframed the audit question from 'where do we rank' to 'what are we known for, by whom, and against whom'.
• Modelled the brand and its competitors as entity sets rather than page sets.
• Generated and clustered underlying search demand so coverage could be measured against real jobs, not a keyword list.
• Simulated assistant citations to test whether the association actually surfaces under pressure.

• Built a crawler that profiles a brand and named competitors and extracts the entities each is associated with.
• Built demand generation and clustering to produce 100+ clusters per audit, then scored coverage per cluster.
• Built AI citation simulation and a gap view ranking where to focus first.
• Turned the whole thing into a repeatable diagnostic run before strategy, not after.

• Competitive AI visibility audits went from 4–8 hours to roughly 60–90 seconds.
• Produced a repeatable framework for diagnosing AI search visibility before execution begins.
• Made 'where should we focus' an evidence-backed answer rather than an opinion.

See it live→

Diagnosis is the scarce good. Teams do not lack the ability to publish; they lack a defensible reason to publish one thing rather than another. Compressing the diagnosis changes what gets decided.

• Search is becoming an answer, and that is a product problem
• Most growth problems are positioning problems wearing a costume