How to Report AI Search Share of Voice to Your Comms Leadership: A PR Dashboard Template for 2026

Key takeaways

• AI share of voice is the share of AI answers that mention, cite, or recommend your brand across a fixed prompt set, compared with competitors. Report it per engine, never blended.
• Your dashboard needs five blocks: headline SoV, per-engine breakdown, sentiment and message accuracy, source mix, and an annotated timeline of engine changes.
• Engines change behavior without warning. Reddit citations in ChatGPT fell off a cliff in August 2026, and leadership will read that as a PR failure unless you annotate it.
• News coverage decays quickly in AI answers, so report on rolling windows and track re-citation, not one-time mentions.
• Lock the prompt library, the competitor set, and the reporting cadence before you collect any data. Changing them mid-quarter ruins trend lines.

Why traditional share of voice does not transfer

PR teams have reported share of voice for decades. You count your brand's mentions, divide by total category mentions, and show a pie chart. It works because the inputs are discrete: articles, broadcasts, social posts.

AI answers are not discrete. An engine pulls from several sources, writes a synthesized response, and may name your brand, cite your page, do both, or do neither. Shadow's guide to AI share of voice makes the same point: the measurement model built for earned media coverage does not translate cleanly to AI search.

So your leadership is going to ask a familiar question ("what's our share of voice?") and you need to give an answer that is honest about what the number is. This guide gives you a dashboard template that does that without burying the CMO or CCO in methodology.

Define AI share of voice before you report it

Use a single definition and put it in the footnote of every report.

AI share of voice = the percentage of AI responses, across a fixed prompt library and a fixed set of engines, that mention or cite your brand, relative to the same measure for named competitors.

Two variants are worth separating, because comms and SEO teams often confuse them:

• Mention share: how often the brand is named in the answer text.
• Citation share: how often your own domain (or a page you earned coverage on) appears in the source list.

A brand can be mentioned a lot and cited rarely, which usually means AI is describing you based on third-party coverage rather than your own content. That is a comms insight, not just an SEO one.

Step 1: Build the prompt library

The prompt library is the foundation. If it is sloppy, every chart downstream is meaningless.

A workable range is 15 to 50 prompts, split across three categories:

• Brand prompts: "What is [company]?" and "Is [company] any good?"
• Category prompts: "Best [category] tools for [use case]."
• Comparison prompts: "[Brand] vs [Competitor]."

Use conversational phrasing of about 6 to 10 words, the way people actually type into an assistant. Get the comms team, sales, and customer support to contribute prompts. Support tickets are an underrated source.

Two rules I would insist on. First, freeze the library for the quarter. You can add prompts, but tag them as new so they don't distort the trend. Second, agree on the competitor set up front, with leadership signing off. Nothing derails a meeting faster than "why isn't Competitor X in here?"

Step 2: Choose engines and report each one separately

Do not blend engines into a single score. The citation slots available differ too much. Promptwatch Data's sources-per-response report shows ChatGPT citing about 5 sources per web-search answer, Google AI Overviews about 10, and Perplexity almost exactly 10 every day. Microsoft Copilot swung from under 2 to nearly 17 sources per response within a few weeks before settling at the low end.

A blended number hides all of that. If you appear in 30% of Perplexity answers and 8% of ChatGPT answers, "19%" tells your leadership nothing they can act on. Show the engines side by side, and judge volatile ones like Copilot on monthly trends, not weekly snapshots.

A sensible starting set: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, and Copilot. Add Claude if your audience skews technical.

Step 3: Pick the tooling

You can run this manually for a small prompt set, but it stops being reasonable around a few dozen prompts across several engines. Tooling falls into rough groups.

For most comms teams, a dedicated tracker plus the free search consoles is enough. Promptwatch is the one I would put first for PR use, because it monitors the real interfaces of ChatGPT, AI Overviews, Perplexity and others, reports per-engine visibility and share of voice, and has sentiment, Reddit, YouTube and LinkedIn citation reports that map neatly onto a PR channel view.

If your team already lives in a media monitoring platform, Meltwater is worth checking for how it handles AI visibility next to traditional coverage.

For a broader look at options, the directories at bestgeosoftware.com and ai-rank-tools.com list GEO and AI rank trackers side by side.

Step 4: The dashboard template

One page. Five blocks. Leadership should be able to read it in two minutes and forward it without a cover note.

Block 1: Headline share of voice

A single row of tiles, one per engine, each showing your share of voice, the change versus last period, and the top competitor's figure. Add a larger tile for your average across engines only if you must, and label it clearly as an average.

Suggested layout:

The numbers above are placeholders for the format. Fill in your own.

Block 2: Breakdown by prompt type

Split share of voice by brand, category, and comparison prompts. This is where the story is. A strong brand-prompt number with a weak category-prompt number means people who already know you get accurate answers, but people shopping the category never hear about you. That is a positioning problem comms can act on.

Block 3: Sentiment and message accuracy

Share of voice says whether you show up. This block says how. Track two things:

• Sentiment: positive, neutral, or negative framing in answers.
• Message accuracy: whether the answer repeats your key messages or gets facts wrong, such as old pricing, a discontinued product, or a stale executive.

Pull three to five verbatim answer excerpts each period. Leaders react to quotes more than to charts. A screenshot of ChatGPT describing your company incorrectly will drive more action than any percentage.

Block 4: Source and channel mix

Show which kinds of sources feed the answers that mention you: your own site, news outlets, review sites, Reddit, YouTube, LinkedIn. This tells comms where earned coverage is actually landing.

Be careful here, because the mix differs sharply by engine. In Promptwatch Data's social media citations report for Sep 28 to Oct 4, 2026, social platforms made up only 0.46% of ChatGPT's citations but 12.2% of Google AI Overviews citations, roughly 26 times higher. Within AI Overviews, YouTube was 4.28%, Facebook 2.38%, and Reddit 2.15%. A social-heavy PR strategy will therefore pay off in one engine and almost vanish in another.

Block 5: Annotated timeline

A line chart of share of voice over time with vertical markers for two kinds of events: things you did (launches, press releases, executive interviews) and things the engines did (model updates, citation behavior changes).

The second kind is not optional. Promptwatch Data's Reddit citation report shows reddit.com's share of ChatGPT citations falling from roughly 4% to 0.5% on August 14, 2026. If your share of voice dipped that week and nobody annotated the chart, someone in the room will assume the comms team did something wrong. Add the marker, add one line of explanation, and move on.

Step 5: Report on rolling windows, not one-off wins

PR people are used to celebrating a placement. In AI search, the placement matters only while it keeps getting cited.

Promptwatch Data's content half-life study tracked pages ChatGPT cited on July 6, 2026 for 90 days. News was the fastest to decay: almost a quarter of news pages were never cited again after day 0, half were gone after 36 days, and 41% were still cited at day 60. Landing pages and docs lasted far longer, with 78% and 76% still cited after 30 days.

What that means for the report:

• Use a rolling 30-day window for share of voice, not a single-day snapshot.
• Add a re-citation metric: of the pages cited last period, how many are still cited now?
• Frame news coverage as a short burst of visibility, and frame owned landing, docs, comparison and pricing pages as the long-term citation assets.

This is also a useful way to justify budget. If earned news decays in AI answers, a comms team that also owns the content that stays cited has a stronger case than one that only delivers placements.

Step 6: Add the business link

Leadership will eventually ask whether any of this drives revenue. Two sources are worth connecting.

Google's Search Console added generative AI performance reports in June 2026, reporting impressions for AI Overviews, AI Mode and Discover by page, country, device and date. Per Search Engine Land's coverage, it carries no click, click-through rate or query data, so treat it as a visibility signal only. Bing Webmaster Tools launched an AI Performance report in public preview in February 2026.

Pair those with AI referral traffic from your analytics (GA4 or a visitor analytics feature in your tracker). Supermetrics has a useful walkthrough of combining these sources into one report if you want to pipe everything into a BI tool.

Keep this block to two or three numbers: AI-referred sessions, conversions from those sessions, and the pages receiving them. Resist the urge to build an attribution model. A simple trend line is more credible than a clever one.

Step 7: Set the cadence

The monthly dashboard is the one that matters. Weekly numbers on volatile engines will whipsaw, and nobody wants to explain a two-point swing every Monday.

Common mistakes to avoid

• Blending engines into one number.
• Changing the prompt library mid-quarter and then comparing to last quarter.
• Reporting mentions without checking whether the answer was accurate or positive.
• Leaving engine-side changes unannotated.
• Presenting a one-time news hit as a durable gain.
• Showing 50 charts when five blocks would do.

What to say in the meeting

A short script helps. Something like: "Across 40 tracked prompts, we appear in roughly one in five answers on Perplexity and one in seven on ChatGPT. Our main competitor leads both. We are strong on brand prompts and weak on category prompts, so the work this quarter is getting our comparison and landing pages cited. The dip in ChatGPT in mid-August was a platform change, not a campaign issue."

That is four sentences, one number per engine, a diagnosis, and an explanation for the dip. It sounds like a person who understands the data.

If you would rather hand the reporting setup, GEO strategy and PR-led citation work to specialists, 1001 SEO Media runs this kind of engagement for brands, using Promptwatch as its measurement platform. Otherwise, the template above is enough to get a credible first dashboard live this month.

Quick checklist

• Prompt library of 15 to 50 prompts, frozen for the quarter
• Competitor set agreed with leadership
• Engines reported separately
• Five-block one-page dashboard
• Annotated timeline for engine changes
• Rolling 30-day window plus re-citation rate
• Search Console, Bing and AI referral data for business impact
• Monthly cadence for leadership, weekly for the working team