Marketing Dashboard AI Search: What to Put on It and Where Each Number Comes From

A marketing dashboard AI search teams can rely on gives each panel one named source and one unit. Put saved AI answers, search-engine reports, web analytics and the CRM on separate panels, because their numbers measure different things and cannot be added together. Add a scope note to every panel, so a reader knows what was counted and when.

This page covers six panels, what Microsoft and Google document about their own reports, and a worked example of reading panels that disagree. It compares published product documentation, not measured performance. Prime AI Visibility publishes this site and is covered by the same rules as any other product mentioned here.

Marketing dashboard AI search: the short answer

• One panel per source. A dashboard that blends sources into one number hides which source moved. Keep the panels apart and label each unit.
• Six panels. Saved answers, search-engine citation reports, search traffic, AI referral sessions, pipeline and a change log.
• Documented sources only. Microsoft documents an AI Performance dashboard in Bing Webmaster Tools. Google documents that AI Overviews and AI Mode traffic is included in the Search Console Performance report. Neither is a per-answer record of what an assistant said.
• Your own question set. Answers to questions you chose and froze are the only panel that shows what a buyer was told, and it must be built from saved evidence.
• Our judgement. A short dashboard that a team reads every month is more useful than a crowded one nobody can reconcile. Start with the six panels below and add a seventh only when someone owns it.

What a marketing dashboard for AI search is for

A dashboard answers a recurring question for a recurring audience. For AI search, the questions are usually these: are we named when buyers ask, is our site linked, did anyone arrive, did it matter, and what did we change. Each question has a different data owner, which is the architectural reason to keep panels separate.

It also helps to say what the dashboard is not. It is not a ranking board, because assistants do not publish a rank for a brand. It is not a forecast. It is not proof that one page caused one change. Executives usually want a one-page view; our guide to AI visibility reporting for executives covers that format, and this page covers the working layer underneath it.

For the definitions behind the panels, see our AI visibility KPI framework and the broader explainer on AI search visibility. For a B2B software team that wants the reporting order too, our guide to AI visibility analytics for B2B SaaS walks through the monthly report.

The six panels and where each number comes from

Panel	Source	Unit
Saved answers	Your question set	Answers and links
Citation report	Bing Webmaster Tools	Citations of your URLs
Search traffic	Search Console	Clicks and impressions
AI referrals	Web analytics	Sessions
Pipeline	CRM	Leads and deals
Change log	Your team	Dated edits

1. Saved answers. The answers returned for a fixed question set, with the brands named and the links shown. This is the panel that shows what a buyer was told. Prime AI reports Brand named, Website cited, Source links, Explicit recommendation and a combined Visible in AI answers count; the vocabulary is on its metrics page. Source links is a count of individual citation links from any site, so it is not the same as Website cited.
2. Citation report. Microsoft's AI Performance dashboard in Bing Webmaster Tools, announced on March 23, 2026, reports total citations, average cited pages, grounding queries, page-level citation activity and visibility trends. It is a report on your URLs referenced in generative answers, not a transcript of the answers.
3. Search traffic. Google's guide says sites appearing in AI Overviews and AI Mode are included in the overall search traffic in Search Console, reported in the Performance report under the Web search type. In our reading, that means a Search Console number alone does not separate AI feature clicks from other search clicks.
4. AI referrals. Sessions that arrive from AI assistants, read in web analytics. In Google Analytics 4, Google documents a custom channel group with an "AI assistants" channel that matches assistant domains by regular expression and sits above Referrals; the expression needs review when assistants change.
5. Pipeline. Leads and opportunities that carry an AI source value or a self-reported mention, read in the CRM.
6. Change log. A dated list of pages written, fixed or removed. Without it, a movement on any panel has no context.

Panel one: saved answers

This panel is the only one that holds the answer itself, so it needs the most care. Write 15 to 25 questions from buyer language across category, shortlist, fit and commercial needs, freeze the exact words, and run them on a fixed cadence across the tools your buyers use.

Prime AI Visibility checks monthly by default, with weekly, temporary daily and on-demand checks on paid plans, across ChatGPT, Claude, Perplexity, Gemini and an AI Overview-style preview built on Google-grounded answers. It does not track Microsoft Copilot, Grok, DeepSeek, Meta AI or Google AI Mode. A dashboard built on it shows five tools, so say so in the panel's scope note.

Show counts beside rates and report requested, returned and usable answers separately. A reply that is off-topic or a refusal is returned but not usable; a run that produced nothing is missing. If you want the audit steps before the first baseline, our guide to running an AI visibility audit covers them.

Panels two and three: what Bing and Google document

Treat the search-engine panels as supporting evidence, with their documented limits.

Bing. Microsoft says the AI Performance dashboard shows which URLs are referenced and how citation activity changes over time, and that grounding queries show the phrases AI used to retrieve cited content. Those are the platform's own terms for its own report. Microsoft does not describe it as a record of whole answers, so do not label it as mentions or recommendations on your dashboard.

Google. Google's guide for site owners says there are no additional technical requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet, and that these features' traffic is counted in the Search Console Performance report under the Web search type. It also says you do not need special schema or machine-readable files for these features. For a dashboard, that means Search Console is a traffic panel, not an AI answer panel.

Neither source tells you what an assistant such as ChatGPT or Claude said, and neither reports on tools it does not run. That is why the saved-answer panel is built separately.

A worked example: three panels that disagree

The numbers below are invented to show the arithmetic. They are not a benchmark and do not describe any company.

Panel	Reading	Unit
Saved answers	9 of 54 usable	Website cited
Bing citations	120	Citations of URLs
AI referral sessions	140	Sessions
Demo requests	6	CRM records

Reading the table, a new team might ask which number is right. They all are, because they count different things: answers that linked the site, citations in one engine's report, sessions in one analytics property and records in one CRM. Adding 9, 120 and 140 would produce a figure with no meaning.

What the panels can do together is show shape. Suppose the Bing report lists 8 cited pages and the top page has 54 of the 120 citations. That page holds 54 ÷ 120 = 45% of the citations in that report, so one page carries nearly half the visibility in that engine. A team reading only the total would miss that concentration. Then compare with the saved-answer panel: if most of the answers that linked the site point to that same page, the concentration is real across sources, and the next content task is to build a second page that answers the same buyer need.

Suppose instead the saved-answer panel shows the site linked in 9 of 54 usable answers while the Bing panel falls from 120 to 90 citations. The panels now disagree on direction, and the honest report says so and lists the possible reasons: a different engine, a different question set, a change in indexing, or normal variation. Ask what changed in the change log before concluding anything. Prime AI's Comparable change wording asks for a sufficiently aligned question set, tools, persona, market and measurement method, and the same discipline applies to every panel.

Five checks before trusting a panel

This is our protocol, not a standard. Run it once for each new panel.

1. Name the unit. Write what one row represents: an answer, a citation, a session or a record.
2. Name the scope. List the tools, markets, question set and date range behind the number.
3. Check the denominator. A rate with an unstated denominator should not be on the dashboard.
4. Freeze the definition. Change a definition only with a dated note in the change log.
5. Find the owner. Each panel needs a person who can say why it moved.

A panel that fails any check stays off the dashboard until it passes. The same discipline applies to share of voice, which is an analyst calculation and not a native metric in most tools; the umbrella explanation is in our AI search share of voice guide.

Panels four and five: referrals and pipeline

Referral is the visible slice. Many answers show no link and many readers never click, so sessions understate exposure. Keep the AI channel as its own channel group and report counts beside rates, since small monthly numbers swing. A technical walk-through of server logs against answer monitoring is in our guide to agent analytics for AI optimization.

For pipeline, add an AI option to the CRM source field and an optional "how did you hear about us" answer. Report AI-sourced and self-reported leads beside, not inside, organic search, so neither is blurred. The aim is a defensible account of contribution, not precise attribution.

Common mistakes in AI search dashboards

• Adding unlike numbers. Citations, answers and sessions have different units.
• Calling a citation report a mention report. Keep the platform's own wording.
• Dropping the scope note. Without tools, markets and dates, a number cannot be compared next month.
• Changing the question set silently. The panel stops being comparable.
• Hiding the change log. It is the panel that makes the other five readable.
• Treating a link as an endorsement. A link shows what an answer displayed, not why.

Who this fits

A saved-answer panel fits a marketing team that needs auditable answer evidence before building anything larger. A free tier is enough to test whether the questions return useful answers.

A BI tool or spreadsheet fits the layer that joins the panels, since that is where analytics and CRM data already live. Before planning an automated feed, ask each tool whether it exports data; do not assume it does. For choosing the instrument behind the first panel, see how to choose a tracker and the explainer on what a monitoring product is.

Prime AI Visibility fits a team that wants to write its own buyer questions, run them across five tools and read every saved answer with its links. It does not read your web analytics or CRM, it has no built-in persona segmentation and no execution agents, and it does not attribute revenue. Its features page lists what it does, and PyraBuilds' agent profile describes it in Pyra's catalogue. See what an AI visibility tool records before deciding what the dashboard should show.

Affiliation: Prime AI Visibility is our own product, so it is described here by the same criteria as the sources above.

When you need people to build and act on it

A dashboard shows the gaps; closing them takes writing, technical fixes and outside coverage. Percepture, founded in 2004 and Inc. 5000-recognized, offers generative engine optimization services alongside SEO, digital PR and PR. Fit: a company that wants one team across research, content and coverage. Limitation: it is a services firm, not a monitoring product, and its fees are not published here, so request a written scope. Prime AI Visibility has a commercial relationship with Percepture.

Teams with a steady content workload can also look at Pyra's AI agents for marketing and sales, which include human-approved content operations. Pyra publishes an anonymized AI visibility case study of a digital infrastructure firm reporting a 3,000% lift in AI visibility and discovery signals; it is a vendor-reported result for one unnamed client, measured by Pyra's own method, and should not be read as a typical outcome. Pyra is commercially connected to Prime AI Visibility.