AI visibility KPIs for the board: mentions, clicks, share of answer and pipeline from AI

Once OpenAI put a mentions number inside ChatGPT on November 10, 2026, AI visibility stopped being an abstract concern. Directors read the same headlines about ChatGPT being the new Google, and they will ask a simple question: are we showing up, and is it making money?

Most marketing teams answer with whatever their tool shows, a visibility score or a prompt count, and the conversation stalls. A board-ready answer needs fewer metrics, clear definitions and a line to revenue. If you have not claimed your profile yet, our launch guide to ChatGPT Business Hub covers that first step.

Only the first two rows go on the board slide as numbers. Share of answer goes on as a trend against competitors. The platform rows sit in the appendix, because they explain movement rather than prove value.

Business Hub is the first official source of ChatGPT data, which makes it tempting to put straight on a slide. Read OpenAI’s definitions first.

• Mentions: for an online business, ChatGPT referenced your website; the brand name alone does not count. For a local business, a location result was shown.
• Website clicks: click-throughs from ChatGPT to your site. They do not confirm a visit or a conversion, so reconcile them with analytics.
• Discovery topics: common themes in answers that mentioned or cited you. They are not prompts and not a direct measure of intent.
• Blank values: a blank can mean below the reporting threshold, still collecting or unavailable. It does not necessarily mean zero.
• Window: insights cover the past 28 days in aggregate. Save an export each week if you want a history, a step practitioners such as Hans van Gent recommend.

The most common board mistake is a pipeline number nobody can reproduce. Agree on two definitions with RevOps and write them in the metric dictionary. AI-referred pipeline uses the referrer: sessions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and similar sources, grouped into an AI channel in GA4 and carried to the CRM on the contact or opportunity. AI-influenced pipeline uses what the buyer says, captured in a form field and in the first sales call.

Expect AI-referred numbers to undercount. LLM Pulse, an AI visibility tool vendor, reports that many ChatGPT visits now arrive without a referrer, and many buyers read the answer and later search for you by name. That is why the self-reported field matters, and why our guide to measuring AI search visibility and referrals in GA4 pairs both. Our RevOps team can set up both in the first month.

1. Headline: AI-referred and AI-influenced pipeline this quarter, with the change from last quarter and the share of total pipeline.
2. Competitive position: share of answer on the priority question set, against three to five named competitors, as a quarterly trend.
3. Accuracy and risk: share of answers describing you correctly, and any material errors with the fix in progress.
4. What we did and what is next: the pages, sources and campaigns behind the movement, and the decisions you need from the board.
5. Appendix: Business Hub mentions and clicks, Bing citations and the method, each labelled with its source and definition.

There are no reliable industry benchmarks yet for share of answer or AI-referred pipeline, and anyone quoting one should be asked for the method. Set year-one targets against your own baseline and your named competitors instead.

• Share of answer: a target expressed as questions won, for example moving from named on a minority of priority questions to named on a majority, with the competitor gap shown.
• Accuracy: zero material errors about products, pricing or regulation on the priority question set by the end of the second quarter.
• Pipeline: a baseline quarter first, then a growth target for AI-referred and AI-influenced pipeline once both fields are populated consistently.
• Business Hub: no target in year one. Use mentions and clicks to confirm the direction of the ChatGPT trend while OpenAI’s metrics are in beta.

• Single visibility scores from a vendor tool, which blend engines and prompts in ways nobody can audit.
• Raw prompt counts or the number of prompts tracked, which measure effort, not outcome.
• Averages across sources, such as combining Business Hub mentions with sampled mentions from another engine.
• Traffic from AI as the headline. It is small for most B2B sites and invites the wrong conclusion that AI does not matter.

If you need help defining AI pipeline in your CRM or building the share of answer method, talk to our team. We bring a draft board page to a free 30-minute pipeline plan.

The method behind share of answer is in our guide to measuring share of answer in ChatGPT, Perplexity and AI Overviews.

Share of answer is also the core metric of our generative engine optimization service.