AI Search Brand Metrics Framework

You may have noticed these new KPI widgets appearing in the AI Search platform recently. Here's a quick explanation of what they mean and what they are telling you.

But first, let's discuss our recommended approach to tracking.

Market Research Waves

Prompt tracking is not the same as SEO rank tracking.
​
LLMs are more volatile and the responses change considerably.

That's why you won't see useless metrics like ranking or average rank in our AI search product. 
​
Just because LLMs are volatile, it doesn't mean you can't track them. Our research shows that the volatility is at the fringes rather than the core, and that around 45% of brands (that are displayed for unbranded prompt sets) may only appear once in a 7-day window.
​
This means you need to measure the consisentcy of these KPIs to get the best picture of how you are performing.
​
We recommend tracking prompts daily in '7-day Research Waves' that you repeat monthly or quarterly. If budget allows, you can check in once a week or once every couple of weeks to see if there's been any major shifts due to algorithm updates or model updates. Only if there is a significant shift would you start another 7-day research wave.

This is so important because LLMs are volatile. You could just happen to pick the wrong day, and you may or may not be showing, and you'll get a completely misleading view of the world. It's also really helpful for citations, because then you can focus on the citations that actually are consistently appearing, not just a long list of citations that might appear sporadically. 
​
So use and report on these metrics - but take a market research mindset to prompt tracking and leave the traditional SEO mindset behind.

Brand Mentions Metrics

Foundation Metrics (Understanding the Landscape)

1. Total Prompts Sent

Definition: The total number of queries sent across all AI models being tracked.

Formula:

Total Prompts Sent = Number of Prompts × Number of Models

Example: 1,000 prompts × 5 models = 5,000 total prompts sent

Typical Goal: Maintain consistent daily tracking across target models.

How to Measure: Track prompt volume per model per day; ensure 100% completion rate

2. Total Responses Received

Definition: The actual number of responses returned by AI models (may be less than prompts sent as some models may not respond).

Formula:

Total Responses Received = Sum of all responses across all models

Example: 4,250 responses received out of 5,000 prompts sent (85% response rate)

Typical Goal: Understand baseline opportunity size; monitor model response rates

How to Measure: Track responses received per model; calculate response rate percentage by model

3. Responses with Brand Mentions

Definition: The subset of total responses that mention at least one brand (your realistic opportunity for brand visibility).

Formula:

Responses Mentioning Any Brand = Count of responses containing ≥1 brand mention

Example: 2,500 responses out of 4,250 total responses mentioned at least one brand (59%)

Typical Goal: Understand the size of brand-relevant conversations

How to Measure: Track percentage of responses that are "brand-relevant" over time; identify which topics/queries drive brand discussions

Performance Metrics (How You're Doing)

4. Share of Responses %

Definition: The percentage of all AI responses (across all models) that mention your brand at least once.

Formula:

Share of Responses = (Responses mentioning your brand / Total Responses Received) × 100

Example: Your brand appears in 340 of 4,250 responses = 8%

Typical Goal: "Increase Share of Responses from 8% to 12% within 90 days"

How to Measure:

• Track weekly/monthly trend

• Monitor absolute change (+4 percentage points)

• Calculate responses gained (+170 additional responses)

• Measure by model to identify best/worst performers

5. Share of Unique Brand Mentions %

Definition: Your brand's share of brand-relevant conversations, counting each brand only once per response (deduplicated).

Formula:

Share of Unique Brand Mentions = (Unique mentions of your brand / Total unique mentions of all brands) × 100

Example: 340 unique mentions of your brand out of 2,500 total unique brand mentions = 13.6%

Typical Goal: "Increase competitive share from 13.6% to 20% within 6 months"

How to Measure:

• Track against specific competitors

• Monitor category leadership position

• Measure by topic area to identify strengths/weaknesses

• Calculate share gain vs. competitors

6. Share of Brand Mentions %

Definition: Your brand's share of total brand conversation volume, counting all mentions including multiple mentions within the same response.

Formula:

Share of Brand Mentions = (Total mentions of your brand / Total mentions of all brands) × 100

Example: 450 total mentions of your brand out of 8,000 total brand mentions = 5.6%

Typical Goal: "Increase brand mention volume from 5.6% to 10% within 6 months"

How to Measure:

• Track mention density (Metric 6 ÷ Metric 5) to understand depth

• Compare against Metric 5 to identify "wide reach vs. deep engagement" pattern

• Monitor which content types drive multiple mentions

Insight:

• If Metric 6 > Metric 5: You dominate fewer conversations with multiple mentions ("narrow reach, high depth")

• If Metric 5 > Metric 6: You appear briefly in many conversations ("wide reach, low depth")
  ​
  ​
  More metrics are coming - feedback and suggestions are also very welcome. ;)