Metrics to Track with Waikay
Waikay tracks six core metrics that together tell you how AI models understand, describe, and recommend your brand. Each metric measures a different layer of AI visibility – from whether AI knows who you are, to which topics it associates you with, to whether the facts it states about you are actually true.
This page explains what each metric measures, how Waikay calculates it, and what to do when the numbers are not where you want them to be.
Understanding Score
This is the first score you receive when creating a Waikay project – a number out of 100 representing how well an AI model understands your brand.
What the number means
Above 70 means the vast majority of what your brand conveys online is also conveyed in LLM answers. Under 70, key concepts are likely missing from how AI models describe you.
Waikay reads your website’s top pages and uses NLP to extract meaningful entities – everything you talk about, sell, and stand for. It builds a mini knowledge graph, then separately queries four AI models (Sonar, ChatGPT, Gemini, and Claude) with a direct prompt: what do you know about mybrand.com?
The score compares entities in those AI responses against your knowledge graph. Gaps in the score are gaps in how AI understands your brand.
Head to the action plan in the top right hand corner of your report, or go to the Action Plans tab in the menu, to see which entities competitors are associated with that you are not and get a step-by-step plan to strengthen your brand’s associations.
Share of Model
Share of Model tells you how often your brand is mentioned in AI responses relative to every competitor tracked for the same prompt. It is the AI equivalent of share of voice – but instead of measuring ad impressions or search rankings, it measures how prominently your brand features in the answers AI models give to buyers in your category.
In the prompt tracking feature, you design commercial prompts to track daily across multiple LLMs simultaneously. For example: What are the best SEO tools for enterprise teams? Waikay queries that prompt across ChatGPT, Gemini, Perplexity, and Claude, tracks every brand mentioned in every response, and calculates your share across all of them.
Brand A mentioned 100 times across all LLM responses in a month, Brand B 75 times, Brand C 25 times. Total: 200 mentions.
Brand A = 50% · Brand B = 37.5% · Brand C = 12.5%
You can track share across all LLMs combined or break it down by individual model to see where you are strongest and where you have the most ground to make up. View share for all prompts at once or isolate a single prompt to understand how you perform for a specific buyer question.
Waikay automatically detects competitor brands appearing in LLM responses, even ones you did not think to add. You can benchmark against up to 20 competitors per prompt, giving you a complete picture of how the AI landscape is distributing visibility in your category.
Entity Mentions
When a prompt is tracked, all brands mentioned are captured. Waikay goes further – it breaks each brand down into the specific entities it is described by, and tracks those associations over time as a visual heatmap.
| Brand | Entity SEO | Content | Internal Links | Brand | Data |
|---|---|---|---|---|---|
| InLinks | 96 | 88 | 61 | 18 | 22 |
| Competitor A | 54 | 79 | – | 82 | 45 |
| Competitor B | 21 | 48 | – | 55 | 91 |
In the example above, InLinks is strongly associated with entity SEO and content, but weakly with brand-related features – a clear signal for where to focus content investment. Entity associations are tracked over time so you can measure progress.
AI Topical Presence
Share of Model tells you whether your brand appears in AI recommendations. AI Topical Presence tells you what for – and that is the question that determines whether your SEO investment is building the right associations in AI models.
This metric measures how strongly and how broadly an AI model associates your brand with the commercial topics that matter in your market. It produces a 0-100 score built from three components:
Depth
How strongly the AI associates your brand with the highest-value topics in your market, weighted by commercial importance.
Breadth
How many core commercial topics in your market your brand is associated with – measured on a soft scoring curve, not a hard threshold.
Concentration
How evenly distributed those associations are. A brand concentrated in one topic is fragile – one model update can cause a significant score drop.
How to read your score
Scores are normalised against the strongest brand in your competitive set. 75+ = category leader. 45-74 = visible but patchy. 20-44 = narrow or emerging. Under 20 = largely absent from AI recommendations.
Two brands can have identical Share of Model scores for completely different reasons. One may be broadly associated with many topics but weakly; another deeply associated with two or three but invisible elsewhere. AI Topical Presence separates these profiles and pinpoints exactly which topics represent actionable gaps.
Run your AI Topical Presence analysis in Waikay
Identify which topics your brand is strongly associated with, which are weak, and which are entirely absent.
Diagnose the gap type
Product gap, documentation gap, or content depth gap – each requires a different fix and a different team conversation.
Build content and strengthen internal linking
Create semantically rich content and properly link topic clusters so associations flow across your whole site.
Retest after 8-12 weeks
Track directional movement in Depth and Breadth scores. Flat scores after a full quarter signal a need to revisit your gap diagnosis, not produce more of the same content.
Citation Tracking
Every time Waikay receives a response from an LLM, it records the sources that model used to ground its answer. What makes Waikay’s citation tracking distinctive is that it keeps two separate citation pools – and that separation is what makes the data actionable.
Knowledge citations
Sources retrieved when AI is asked directly about your brand – for example, “what do you know about Waikay?” These citations shape your brand narrative: what AI believes to be factually true about who you are, what you do, and how you are described. If misinformation about your brand is circulating, it will show up here first.
Commercial citations
Sources cited when AI answers buyer intent queries – for example, “what is the best SEO tool?” These citations influence whether your brand gets recommended. A domain that appears consistently in commercial citations is actively shaping which brands AI puts in front of buyers. These are your highest-value PR and content placement targets.
By tracking both pools separately, you can diagnose two distinct problems independently. Weak knowledge citations point to gaps in how your brand is documented across the web. Weak commercial citations point to gaps in the third-party content, reviews, and comparisons that drive AI recommendations. Both matter, but they require completely different responses.
In knowledge citations: Are your own pages appearing? Is misinformation coming from a specific domain – a review site, an outdated article, a competitor’s blog? Trace it, understand why it has influence, and work to correct it at the source.
In commercial citations: Which domains are cited most heavily when buyers ask questions in your category? These are the sites AI trusts to answer commercial queries. Getting your brand featured on them is more valuable for AI visibility than almost any on-site optimisation.
Fact Checker
Waikay takes the raw responses that AI models generate about your brand and synthesises them into clean, individual sentences – one claim per line. Each sentence is labelled with the model it came from (for example, gemini-grounded or chatgpt) so you always know which AI produced which statement.
Every entity in each sentence is highlighted and colour-coded, making it immediately clear what the AI is associating with your brand and how. You can filter by entity across all sentences to see every claim that mentions a specific concept – useful for spotting patterns in how a particular aspect of your brand is being described across different models.
Each sentence has two actions: Check marks it as accurate. Flag marks it as a hallucination or misinformation. When you flag a sentence, you can open it to see exactly which model produced it and follow any citations the LLM provided – giving you a direct path to tracing where the incorrect information is coming from and what is reinforcing it.
Every time your topic reports update, the sentences reset with fresh AI responses. This means your team is always reviewing current claims, not stale ones – making it a practical tool for ongoing brand accuracy monitoring rather than a one-time audit.
