How to easily assess a site’s AI Search Visibility?

To do that, I, first, apply pattern-based rules or Regular Expressions to isolate interrogative and comparative or evaluation queries. Likewise, these may contain “How?”, “Who?”, “Why?”, “Which?”, “When”, etc words.

Then, — manually qualify the resulting cohorts according to their relevance to the business. For example, this can be a division into 3 classes: 

1. A — strictly relevant, including where there is a direct provider evaluation or excellent comparative query, 
2. B — relevant, but might be broad or brand-led, etc
3. C — not relevant, for example, only applies to a sub-set of a proposition or totally irrelevant.

This must produce a smaller set of AIO-relevant queries that are commercially meaningful.

I then reverse-engineered these queries (limiting myself to class A first) into a structured vocabulary based on the principles I used throughout the SEO work: predicate, object, industry/sector, business qualifier, geography and other meaningful contextual modifiers. The objective is to strip away incidental wording while preserving enough context for the resulting entity to represent a meaningful commercial question. 

Example of a resulting entity: courier service provider in France for e-commerce, which obeys the following grammar: predicate + geo qualifier + industry.

The resulting entities’ invites the idea of expansion — at the expense of strategic entities that are important to an organization but have not yet been observed in the current AIO query corpus. This is to create a reference universe that combines both demonstrated search demand and strategic areas where an organization wants to establish visibility.

The resulting entity universe is then converted into a set of standardised reference questions and selected LLMs are asked to recommend the most prominent brands for each entity on a weekly basis. By the way, if you are already using Hubspot (or Semrush's AI Visibility Toolkit), you may simply use each entity as a “prompt”.