For years, PR measurement has run on a familiar set of numbers and metrics. Reporting has relied on impressions, reach, share of voice, and sentiment across print, broadcast, digital, and social channels to demonstrate value and gauge brand visibility. 

However, the evolving media landscape warrants a shift in how we track visibility. As AI answer engines become a growing source for buyer activity, your brand’s visibility across AI platforms directly impacts the bottom line. And PR and comms teams realize that. Agility’s industry research, ‘Closing the AI Visibility Gap’, highlights how almost 50% of PR professionals consider AI visibility as a critical priority.  

Measuring AI visibility means learning a new set of metrics, understanding why each one matters, and knowing how to calculate it. This guide breaks down the metrics that count, grouped into the two questions every PR team needs to answer: is our brand present, and is it portrayed correctly? 

Why AI visibility can’t be measured like search rankings

AI visibility requires a different measurement approach because AI answer engines are non-deterministic: the same prompt returns different brands on different attempts. Traditional metrics assume stable, repeatable results. AI answers do not behave that way.

Conventional PR and SEO metrics were built for a world of clips and rankings. When a buyer asks ChatGPT, Gemini, Perplexity, or Claude for a recommendation, they do not get a ranked list of ten links. They get a single synthesized answer naming a handful of brands. That is a fundamentally different output, and it calls for a different measurement framework.

The bigger difference is stability. PR teams track brand mentions across broadcast, print, and social, and those mentions stay fixed once published. Search rankings shift with localization and personalization but remain relatively consistent over time. AI visibility does neither. Research found that AI engines cite 40 to 60 percent different domains for identical prompts one month apart. Over six months, that variation climbs to 70 to 90 percent.

The cause is architectural. Large language models are non-deterministic, meaning identical inputs can produce different outputs on repeat runs. Different outputs mean different brands named for the same query.

The practical consequence: no single query result is reliable on its own. Measuring AI visibility is a continuous practice, not a one-off check.

9 AI Visibility metrics for PR

Every metric below relies on running a stable, representative set of prompts repeatedly (three to ten runs per prompt per engine), then aggregating the results.

With that consideration, let’s go over the top nine AI visibility metrics that are crucial for PR teams: 

Presence metrics: Is your brand in the answer?

Presence metrics help establish a baseline and show AI visibility at a surface level. If a brand does not appear in the answer at all, nuanced insight doesn’t matter. 

1. Brand mention frequency (visibility rate)

This is the AI equivalent of reach: how often your brand appears across a defined set of relevant prompts. It is the single clearest indicator of whether you are in the consideration set. To calculate it, run your prompt set across each engine and divide the number of responses that mention your brand by the total number of responses: 

Visibility rate = (responses mentioning your brand ÷ total responses) × 100 

For instance, if your brand is mentioned 60 times out of the 100 responses to the prompts you are tracking, your visibility rate would be 60%. 

Research across 175 brands highlighted that 89% of brands aren’t mentioned in category prompts. This highlights a search landscape that favours early movers, rewarding them for optimizing their content for AI visibility before the industry catches up. 

2. Platform coverage

Agility’s ‘Closing the AI Visibility Gap’ report reveals every AI engine cites and recommends brands differently, so a total visibility score can hide a complete absence on the one platform your audience actually uses. 

For instance, the platform breakdown of a 30% overall visibility score can look something like: 

• ChatGPT: 33% 
• Perplexity: 32% 
• Gemini: 35% 
• Claude:  20% 

The breakdown highlights room for improvement in AI visibility for Claude. If that is an AI platform your audience prefers, this quickly becomes a priority to look into. This is why PR teams need to track visibility rate separately for each engine rather than blending them. This helps PR and comms teams truly understand how they’re performing across different AI platforms and optimize for the ones their audience uses. 

3. Prompt-cluster and query coverage

A brand can appear reliably for a broad knowledge-based question like “what is [category]” yet vanish for a high-intent, conversion-based one like “best [category] tools.” 

Many AI visibility platforms enable users to create prompt topics, aligning similar prompts under one group. Grouping prompts into clusters by topic, funnel stage, or use case shows exactly where visibility breaks down. Because individual prompts are volatile, clustering also produces more stable, meaningful data than tracking any single query. 

You can calculate coverage as: 

Prompt coverage = (prompts that mention your brand ÷ total relevant prompts) × 100 

If you are tracking 50 prompts for your brand and your business appears in the answers for 20 of those, your prompt coverage is 40%. 

4. Position and placement

Not all mentions carry equal weight. A brand named first in a recommendation gets far more attention than one buried mid-list or relegated to a footnote citation.  

For instance, high-volume prompts such as ‘X best tools for media monitoring’  may get high visibility (>50%) and recommend your brand consistently. However, if you’re ranked 7, it may not do much good for traffic and lead acquisition.

In contrast, a highly specific prompt like ‘how to set up media monitoring’ might not be very high in volume, but ranking in the top three for such high-intent prompts can lead to higher brand visibility, better engagement, and improved leads. 

Track where your brand lands within the answer and aggregate weekly, since position can swing day to day. Being present is the baseline, but the goal should be to rank prominently for your industry prompts. 

Portrayal metrics: what does the answer say about you?

These are the metrics most teams overlook, and the ones that separate genuine visibility from raw presence. A mention is not automatically a good thing. An engine can name your brand while describing it inaccurately or unfavourably, and that kind of visibility can hurt more than absence.  

Here are some AI portrayal metrics PR teams should track: 

5. Sentiment

Sentiment captures the tone and context in which an engine describes your brand, distinguishing a genuine recommendation from a neutral reference or a side-by-side comparison. It is one of the most actionable areas in AI visibility because the sources shaping sentiment can often be corrected quickly. 

Most platforms offer a 100-scale sentiment score, where a 70+ sentiment score denotes a largely positive user experience that reflects how people are talking about your brand across various mentions. Track the share of your mentions that are positive, neutral, or negative, and pair it with visibility rate to help distinguish a successful PR campaign from a crisis. 

6. Portrayal accuracy

Because models synthesize from fragmented and sometimes conflicting sources, they can state confident but wrong claims about a brand’s products, pricing, or positioning.

Accuracy tracking, checking whether the engine’s description matches the intended messaging, is how a team catches AI hallucinations or an outdated claim before it becomes part of an AI-generated response. In a channel the brand does not control, this metric directly helps protect brand reputation. 

Competitive and trend metrics: how do you stack up against the competitors?

7. Share of voice

Since every AI answer is a closed list of names, relative position matters more than absolute mention count. Share of voice turns a visibility score into a competitive benchmark, pitching your proportion of mentions against competitors within the same prompt clusters. The standard calculation is straightforward: 

AI share of voice = (your brand mentions ÷ total brand mentions across tracked prompts) × 100 

For example, if engines produce 200 brand mentions across your tracked prompts and 50 are yours, your share of voice is 25%. There is no universal target: in a two-player category, 50% is parity, while in a fragmented market 15% may be category leadership. Judge it against competitors and against your own trend, and measure it per engine, since your ChatGPT share will differ from your Perplexity share. 

8. Citation and source analysis

This is the metric most directly actionable for PR. It identifies which outlets and source types an engine references alongside your brand, and which are lifting your competitors instead. 

Since AI answers are overwhelmingly built on earned media, knowing the specific publications and journalists an engine treats as authoritative in your category tells you exactly where to focus outreach. 

9. Historical trajectory

A one-off audit shows your brand visibility today. Since AI-generated answers can change daily, analyzing the trajectory and performance over time is what shows whether your strategy is working. 

Tracking visibility, sentiment, and share of voice over time is what separates a snapshot from a signal. Movement is often the most telling sign, since an increase in share of voice over a quarter signals a comprehensive AI visibility strategy that combines concrete messaging, prompt-centric content, and successful media outreach. This is something that a single reading would never reveal. 

Turning metrics into a measurement practice

The value of these metrics comes from tracking them consistently against a stable prompt set and connecting them to the earned-media activity that moves them. A workable strategy is tracking the prompts every month to spot real trends, conducting a deeper competitive review each quarter, and an ad hoc check after any major brand announcement or model update. 

Doing this by hand across four engines, dozens of prompts, and multiple runs each is what’s most challenging. Agility’s proprietary research, Closing the AI Visibility Gap, which analyzed 24,480 prompt-and-answer pairs across ChatGPT, Perplexity, Gemini, and Claude, was built on this kind of structured measurement at scale. 

And this is exactly the problem Visibility Intelligence solves for PR teams, helping them monitor presence, portrayal, source influence, and competitive share across every major AI engine. 

Ready to see how your brand fares across AI engines? Book a demo.