How to measure and improve your visibility in AI search.

“How do we rank in AI?” sounds like a simple question. It is actually several questions: Which platform? Which audience? Which prompt? Which location? Which date? Are we measuring a mention, a citation, a recommendation, a linked source, or a visit that became a lead?

Start by defining AI visibility

AI discovery can include Google AI features, Microsoft Copilot experiences, answer engines, assistant tools, and AI-assisted search interfaces. Each can retrieve, combine, and present information differently. Visibility may mean a brand mention in one experience, a linked citation in another, or inclusion among recommended providers after a multi-part question.

That means a responsible benchmark must declare its scope. It should identify the platforms checked, the prompts used, the markets and user context represented, the frequency of testing, and what qualifies as a successful appearance.

Why one AI score can mislead

A blended number is attractive because it is easy to compare. But it can hide more than it reveals. A company could improve on informational questions while disappearing from high-value local queries. It could earn mentions without citations. It could be visible for broad topics that never produce a qualified prospect.

Scores also move because the environment moves. Models, indexes, interface rules, query interpretation, source availability, freshness, and personalization can change. The score should be treated as a directional indicator—not a fixed grade on the business.

A good score is:

• Repeatable enough to compare across time.
• Segmented by audience, market, service, and intent.
• Transparent about platforms and prompts.
• Connected to specific cited pages and source evidence.
• Reviewed alongside search, analytics, lead, and revenue data.

A better AI visibility measurement framework

Microsoft’s Bing Webmaster Tools now exposes AI citation activity across supported experiences, including total citations, cited pages, sample grounding queries, and trends. Google provides Search Console reporting for visibility within its supported generative search features. These platform-owned measurements are valuable because they are closer to the source than a third-party estimate—but they still need business context.

The CAG reporting standard

CAG’s recommended scorecard begins with a baseline and a competitive benchmark. It then separates technical eligibility, traditional search visibility, AI citation activity where available, content performance, conversion behavior, and qualified business outcomes.

1. Baseline: indexed pages, technical barriers, current rankings, citations, profiles, and authority gaps.
2. Priority question set: the real questions prospects ask at awareness, comparison, and decision stages.
3. Source map: the pages, profiles, articles, publications, and third-party signals supporting each question.
4. Monthly movement: changes in citations, visibility, cited URLs, competitors, traffic, engagement, and leads.
5. Action log: what was changed, why it was changed, and what happened afterward.

The most important feature is the action log. Reporting without action is decoration. If a priority page is technically eligible but never appears, improve the answer, evidence, internal connections, third-party support, and distribution. If a page is cited but does not convert, improve the on-page experience and offer. If visibility rises but qualified demand does not, revisit the audience and topic strategy.

Primary sources

1. Google Search Central: AI features and your website
2. Google Search Central: Generative AI performance reports
3. Bing Webmaster Blog: AI Performance reporting