How to measure AI visibility - and tell whether it is helping your business

Your latest report says AI visibility has increased. There are more mentions, more citations and a better-looking percentage. What should your team do with that information?

A useful AI visibility report separates appearance, engagement and business outcomes. It defines what was measured, keeps the comparison consistent and helps you decide which work deserves attention next.

It also makes the gaps visible. You will rarely be able to trace every AI appearance through to a customer decision. That is a reporting limitation to explain, not a reason to call every mention revenue.

Decide what each measure means

Brand mention

Your business is named in an answer. Record whether it is a recommendation, a comparison, a passing reference or an inaccurate description. The context matters.

Citation

A source is visibly referenced. A cited page may explain a concept without recommending the business that owns it. Count citations separately from brand recommendations.

Referral visit

A person arrives through a link and the available analytics identify its source. This is website activity, rather than an appearance in an answer. Some discovery will not produce a trackable visit.

Commercial action

A visitor submits a relevant enquiry, requests a demonstration or buys. Define the action carefully and check whether it represents genuine value to the business.

Use platform data for the question it can answer

Bing’s AI Performance reporting covers citation activity across supported Microsoft experiences and selected partners. Its documentation makes clear that citations are not clicks, rankings or proof of importance, and aggregated trends do not identify a single cause. Read Bing’s AI Performance definitions.

That makes it useful evidence of citation activity within the supported reporting scope. It is not a complete view of ChatGPT, Claude and every AI answer.

Keep platform-provided data separate from your own monitored question set. Both can be useful, but their denominators and collection methods differ. Combining them into a single unexplained “AI score” hides information your team needs.

Create a monitoring set with a commercial purpose

Group questions around the decisions buyers need to make: understanding a problem, comparing approaches, checking requirements and choosing a supplier. Record the market, language and service each question relates to.

A defined set does not need to represent every possible question. It does need to be relevant and stable enough to support comparisons.

For example, if your baseline includes 20 questions, and your business appears in eight responses, report eight appearances in those 20 observations. Do not describe that as “40% of AI search”. These numbers illustrate the calculation; they are not Re-Imagine client results.

Our guide to investigating competitor recommendations explains what to record alongside each observation.

Make the baseline inspectable

Every report should state its date range, platforms, question set, collection method and definitions. Save representative answers and cited URLs. Record changes to prompts, markets, platform settings or the monitoring method.

Compare like with like. If you add easy branded questions after a weak non-branded baseline, the percentage may rise without showing improved discovery among unfamiliar buyers.

Separate informational and supplier-selection questions. An appearance for “what does this term mean?” serves a different purpose from an appearance when someone is comparing suitable providers.

Connect the evidence with website activity

Agree meaningful actions before assessing results. In Google Analytics, recommended events include generate_lead for a submitted form or information request and purchase for a completed transaction. They require suitable implementation and context. See Google’s recommended event guidance.

Then ask practical questions: which landing pages receive identifiable AI referrals? Do those visitors explore relevant services? Are enquiries suitable? Does the sales team hear customers mention AI-assisted research?

Keep self-reported discovery separate from analytics attribution. Both are useful, but a customer saying “I saw you in ChatGPT” is not the same dataset as a recorded referral session.

What the Serviceware figures do—and do not—show

Serviceware’s ITFM content received 5,806 Microsoft Copilot citations in its reported monitoring dataset. That is evidence of citation activity. It is not a count of visitors or customers.

Separately, organic clicks across the ITFM cluster increased by approximately 127% between April and August 2026. The Google AI Overview monitoring also recorded appearance in 78% of the monitored UK results.

Each figure needs its own scope and comparison. Reporting them together gives a broader view of visibility; it does not prove that the citations caused the organic growth or produced a particular amount of revenue. Read the Serviceware programme and results.

End the report with decisions

A useful reporting review should answer:

• Which commercially relevant topics are improving or remaining absent?
• Which pages are being cited, and what questions do they help answer?
• Are there factual inaccuracies to correct?
• Which visits and enquiries can we identify?
• What should we improve next, and what evidence supports that priority?

That could lead to enriching an existing page, fixing an access issue or developing a missing comparison. Our content audit guide helps turn those gaps into editorial decisions.

If you need a connected plan for measurement and implementation, explore our AI SEO services or talk to us about your current reporting.

About the author

Molly Sishton, Founder of Re-Imagine Marketing. Molly combines marketing strategy, SEO and content expertise with hands-on experience of developing AI technology. Re-Imagine has worked with AI since 2021 and developed Sighted, its own enrichment and visibility technology. Meet the Re-Imagine team.

Recognition across our founding team and ventures: Connie Sishton featured among Wales Business Insider’s 25 Rising Stars of Welsh Business in 2025. Molly and Connie’s venture SO-ME won Best Female-Led Startup at Pitch Black Malta. Sighted was selected for Tech Nation’s Breakout 50.