For companies trying to understand how they appear in AI-generated answers, one of the biggest problems is consistency. Ask an AI assistant the same shopping question more than once and the recommendations can change, making it difficult to determine whether a marketing or content change actually improved visibility. Metrisque is taking a different approach, launching a measurement platform designed to track how AI models understand brands, categorize products and connect company language with what buyers are asking.

Search visibility has traditionally been relatively straightforward to measure.

A company can track rankings, clicks, impressions and conversions, then compare those numbers before and after making changes to its website.

AI search is different.

Ask an AI assistant for the best products in a category and the answer can change between queries. Different models can consult different websites, interpret the same question differently and surface different recommendations.

That creates a measurement problem for companies trying to understand their position in the emerging AI-driven discovery market.

Metrisque has launched a measurement platform designed to address that problem by focusing on how AI models understand a company, rather than simply counting how frequently its name appears in generated answers.

The company’s thesis is that AI visibility needs a measurement system that produces a sufficiently consistent signal to allow businesses to make a change, test it and determine whether the change actually worked.

AI visibility is harder to measure than search rankings

Traditional search optimization operates within a relatively structured environment.

AI-generated recommendations introduce more variables.

The model may rely on information stored in its learned knowledge, information retrieved from the web, or a combination of sources. Different models can also construct different candidate sets before deciding which brands to recommend.

That makes a simple metric such as “brand mentioned in AI answer” potentially misleading.

A company might appear in one answer and disappear from the next without changing anything on its own website.

Metrisque instead measures how closely a company’s language aligns with the intent expressed in buyer questions.

The objective is to create a more stable measurement that companies can use to evaluate changes to their content, positioning and product information.

Six measurements target different parts of AI discovery

Metrisque’s platform breaks AI visibility into six areas, reflecting different stages of how a model can understand and recommend a company.

Brand Recall measures what an AI model already knows or believes about a company before a new query is introduced. That can reveal whether the model recognizes the brand and whether its understanding is accurate.

Category Fit looks at how AI models classify a company and its individual products.

That distinction is important because a model can correctly recognize a brand while misclassifying one of its products.

Metrisque points to an example involving a beauty brand whose makeup collection was interpreted by an AI model as a bug-collecting kit because of language associated with “catching” creatures.

The underlying lesson is straightforward: being recognized is not the same as being placed in the right category.

Buyer intent becomes the measurement target

Another metric, Buyer Match, focuses on the relationship between company language and the questions consumers ask AI assistants.

Consumers rarely use one standardized phrase to describe what they want.

They can express the same intent in dozens of ways, and new variations appear constantly.

Instead of attempting to optimize separately for every possible wording, Metrisque measures whether the company’s language is semantically reachable from the underlying buyer intent.

That moves AI optimization closer to a concept familiar from modern search: understanding what the user means rather than matching a precise keyword.

Recommendations reveal the competitive field

AI Recommendations measures which companies are actually named when a buyer asks an AI model for a recommendation.

But Metrisque goes beyond the final answer.

Its approach also considers the broader set of companies a model may have evaluated before producing its recommendation.

That distinction could matter for marketers.

A company that is never considered has a different problem from one that consistently makes the model’s shortlist but loses out to competitors at the final recommendation stage.

The first requires improving relevance or category positioning. The second may involve differentiation, authority, product information or competitive positioning.

Competitor citations expose the AI information layer

Metrisque’s Competitor’s Citations metric looks at the websites AI models rely on when answering buyer questions and whether those sources already mention a particular company.

This highlights one of the less visible challenges of AI discovery: models do not necessarily use the same web.

According to Metrisque’s research, 46 sources were cited across one question, but only one source was cited by all three models tested. Thirty-nine sources were cited by only a single model.

That means a company can have strong visibility in the information ecosystem used by one AI model while remaining largely absent from another.

For marketers, the implication is significant.

Traditional SEO often treats Google’s index as the dominant discovery layer. AI search increasingly creates multiple discovery environments, each with potentially different source relationships.

Finding the questions worth competing for

The sixth metric, Question Finder, attempts to identify which buyer questions represent meaningful opportunities.

A query may appear competitive from one measurement approach but relatively open from another.

Metrisque argues that this happens because AI systems can draw on different combinations of remembered knowledge, retrieved sources and information considered during answer generation.

The platform attempts to distinguish between questions that are already dominated by a competitor, contested between several companies or still relatively open.

That could help companies prioritize their AI visibility efforts instead of trying to optimize for every conceivable question.

The company’s bigger argument: measurement needs reproducibility

The underlying problem Metrisque is trying to solve is not simply visibility.

It is experimental reliability.

If the result of an AI query changes substantially from one run to another, marketers cannot confidently connect an intervention to an outcome.

Metrisque says its own research demonstrates this instability.

In one test involving three leading AI models, the company found that the models relied on substantially different websites when answering the same buyer question. It also reports that roughly 60% of the websites used by the same model changed when the identical question was asked days apart.

That does not necessarily mean AI visibility cannot be measured. It means the methodology needs to account for the variability inherent in AI systems.

A pre-registered test adds an evidence layer

Metrisque says it tested its predictive methodology in a pre-registered study covering roughly 1,100 real recommendations from two leading AI models.

The company reports that 92% of the recommendations aligned with predictions made before the models were queried.

The study has been published with a permanent DOI citation, 10.5281/zenodo.21417361.

Metrisque also highlights that it published a prediction it got wrong.

That transparency is relevant because AI visibility measurement is still an emerging category. Claims about performance can be difficult to evaluate if vendors only publish successful examples.

Independent replication and transparent methodology will become increasingly important as businesses begin allocating marketing budgets specifically toward AI-generated discovery.

From SEO to AI visibility

Metrisque’s launch points toward a broader change in digital marketing.

For years, businesses optimized around search-engine rankings because search engines provided relatively stable interfaces and measurable signals.

AI assistants are becoming another gateway between consumers and products.

The challenge is that the ranking system is less visible.

There may be no conventional results page, no fixed position and no guarantee that the same sources will be used every time.

That changes the optimization problem.

Companies increasingly need to understand not only where they rank, but how AI systems understand their brand, which category they associate with their products, which sources they trust and which competitors appear alongside them.

Metrisque is betting that these questions require a new measurement layer.

If that market develops as expected, AI visibility could become a distinct discipline alongside traditional SEO — with its own metrics, experimentation methods and competitive intelligence.

The immediate challenge, however, is proving that those measurements remain useful as AI models, retrieval systems and consumer behavior continue to change.

Market Landscape

The emergence of AI assistants is creating a new layer of digital discovery sometimes described as AI search, generative search, answer-engine optimization (AEO) or generative engine optimization (GEO).

Several trends are converging:

• AI recommendations: Consumers increasingly ask conversational systems to compare products and recommend brands.
• Semantic discovery: AI systems can interpret intent rather than relying solely on exact keywords.
• Retrieval variability: Different models can draw from substantially different sources.
• AI brand perception: Companies need to understand what models already “know” about their brands.
• Competitive AI visibility: Being considered by a model may be as important as appearing in its final answer.
• Measurement infrastructure: Marketing teams need repeatable methods for testing whether content and positioning changes affect AI visibility.

The emerging opportunity is therefore broader than traditional SEO. It is about understanding the information pathways through which AI models discover, classify and recommend companies.

Top Insights

• Metrisque is targeting AI visibility’s biggest measurement problem: inconsistent model outputs make it difficult to prove whether optimization changes actually work.
• The platform measures more than mentions, examining brand recall, category classification, buyer intent, recommendations, citations and question-level opportunities.
• Different AI models can see different webs, making visibility across one assistant insufficient evidence that a brand is broadly discoverable.
• Reproducibility could become a competitive advantage as marketers demand measurement systems that support controlled experimentation rather than anecdotal AI-search wins.
• AI visibility is emerging as a new marketing discipline, potentially extending SEO into brand perception, semantic relevance, retrieval sources and model recommendations.

Power Tomorrow’s Intelligence — Build It with TechEdgeAI