Traditional assignment of voice (SOV) is obsolete, yet many organizations have replaced it with an equally flawed successor: AI assignment of voice.
Software vendors now say they measure brand visibility across ChatGPT, Gemini, Claude, Perplexity, and other AI platforms using a single percentage point. The problem is that these metrics depend on a hidden denominator.
Unlike traditional search, where visibility can be measured against a set of known keywords, the universe of possible AI commands is endless.
Traditional SOV had limitations, but at least its assumptions were clear. Advertisers define a fixed keyword set, track visibility against competitors, and use that list as a stable minimum price. Everyone understood the limits of the measure.
That model no longer exists. Search results are dynamic and personalized, and are increasingly being replaced by conversations. Yet many AI visualization platforms continue to present seemingly accurate percentages that cannot be audited or verified.
To stop presenting mythical metrics to leadership teams, we must rethink how we define and measure visibility in search AI.
Why traditional SOV metrics are now failing
The basic assumptions of search engine optimization and digital product tracking are being broken by two major shifts: the disappearance of the static results page and the rapid rise of personalized, interactive responses.
Search engines have become highly dynamic, personalized environments that continuously change shape based on real-time data.
Between AI-generated summaries, localized results, continuous scrolling, interactive vendor grids, and real-time social feeds, no two users will encounter the same interface, even when entering the exact same query at the same time.
Because the search area is constantly changing, trying to calculate an accurate “share” of that screen has been a mathematical impossibility.
Your customers are searching everywhere. Make sure it’s your product he appears.
The SEO toolkit you know, and the AI visibility data you need.
Start a Free Trial
Start with
A new dynamic adaptation of the standards
Ranking high in the old marketing model meant taking a very predictable percentage of user click-through rates.
In the modern search environment, however, material prioritization may underline a few sponsored listings, AI-generated overviews, interactive query accordions, and featured discussions from public forums.
Because search engines now dynamically build rankings in response to a user’s immediate intent and past search history, rankings change by the hour.
Estimating the volume of a voice based on a static position is not as productive as trying to measure the volume of an ocean wave with a wooden ruler.
Voice assignment for modern AI
When marketing teams realized that traditional position tracking was losing its utility, software vendors quickly introduced other metrics, called LLM Visibility or AI voice sharing.
These dashboards present highly polished, authoritative scores that suggest the product pipeline has been successfully mapped across platforms such as ChatGPT, Claude, Gemini, and Perplexity.
These tools fail to deliver on this promise, exposing a methodological problem that we need to address directly.
| Legacy tracking (obvious) | Appearance of LLM (black box) |
| – Define a list of fixed (known) keywords. – Measure the rate in SERPAuditable constant denominator. | – Unlimited information possible for the user. – The seller uses a small set, which is not allowed. – Subject denominator. |
An endless tail
Legacy SEO tools rely on a user-defined keyword list that serves as a transparent value, while modern search engines present a completely different statistical reality where the universe of possible user commands is effectively infinite.
Consumers are no longer looking for solutions using simple, two-word phrases. Instead, they field specific, conversational questions that describe their specific organizational context, integration requirements, and feature requirements.
Because no marketing tool can realistically sample the infinite natural language landscape, software vendors instead select a small, arbitrary set of static data, run it through AI models behind the scenes, and combine those limited results into percentages that represent the world.
This process creates a metric that only measures voice sharing within the built and artificial environment, presenting a closed sandbox as if it were an open web.
Get the newsletter search marketers rely on.
The problem with black box metrics
Advertisers retained full visibility into the data they were analyzing with legacy tracking tools, meaning that if the system reported a certain percentage of visibility, the list of underlying keywords could be researched and adjusted. Modern LLM visualization tools hide their value among nearly incomplete vendor-defined systems.
This structural flaw became incredibly clear in September 2025, when OpenAI upgraded to its ChatGPT 5.0 model. After this release, the field-wide volume of outbound citations and source links decreased.
For marketing teams that rely on LLM tracking dashboards, this model change has led to a sudden, sharp drop in their reported visibility metrics. The decline had nothing to do with a loss of product relevance or a marketing strategy failure. ChatGPT has recently changed the way it presents source data to users.
This review shows that modern AI metrics are highly variable and out of your control. Although software vendors are really trying to solve an incredibly complex engineering problem, the basic approach just can’t support the high-confidence dashboards they deliver, meaning these metrics should be treated as guiding signals rather than hard numbers.
Beyond AI share of voice: 3 most important metrics
We must shift our focus from measuring pure search volume to measuring how effectively a brand is integrated into the broader context of digital conversations.
As search queries evolve into conversational discovery, product visibility is no longer defined by the keywords it carries, but by how deep it is in AI-powered models.
1. Sharing mentions
AI models incorporate relationships between concepts instead of simply indexing pages, meaning that the product must exist within the model’s training data, fine-tuning datasets, or real-time retrieval sources to be displayed at all.
Mention sharing tracks how often the name of your brand, products, or key executives are included in the responses generated in the broader information ecosystem.
This metric shifts the focus of performance from positions to wording, ensuring that the product is seen by the model even if it is not clearly stated in the seller’s listing.
To influence this metric, organizations should focus on getting mentions across the most trusted forums, engineering communities, and authoritative industry publications where AI models gather and review their knowledge.
2. Sharing recommendations
When consumers use search engines to make purchasing decisions, they often ask for direct comparisons, short lists, and product recommendations to simplify their research process.
Recommendation sharing measures how often your product or service is featured prominently when a user asks the AI engine to act as an advisor on a specific business challenge.
This approach shifts our focus from getting raw traffic to winning a buyer’s consideration set, which is important because search engines filter the noise of the web to deliver a highly curated list of options.
If your product positioning is too generic, the model will have a hard time differentiating the offering and it will evolve into admirable competitors who have developed a very clear, well-documented use case.
3. Sharing a story
Simply getting what’s said in an AI response isn’t enough if the context of what’s said reflects your brand negatively, as high visibility within a bad frame can quickly become a liability.
Narrative sharing measures the quality attributes, attributes, and associations linked to your brand name in conversation results, allowing you to understand how your business is doing independently.
| Narration | What follows | It is a core strategic question |
| A “best of” narrative. | How often are you framed as the ultimate, gold-standard market leader. | Does the model position our brand as the most effective solution available? |
| A “popular” narrative. | It is often cited as the default, widely accepted industry standard. | Is the model that identifies our product as the most used option? |
| It’s a “budget” issue. | How often are you classified as an alternative that has cost, value, or entry level. | Does the model frame our product primarily as a low-cost, entry-level option? |
If an AI engine consistently covers your product but consistently describes your product as a complex, legacy system, your high share of voice may be hurting your sales.
See the complete picture of your search visibility.
Track, optimize, and win in Google search and AI from one platform.
Start a Free Trial
Start with
Redefining your success metrics
Leadership teams need competitive benchmarks to analyze market performance, which means you can’t simply stop reporting on share of voice without offering a viable alternative.
Delivering your senior report smoothly requires a systematic, three-step plan.
Reshaping your management narrative involves educating your leadership team on the limitations of modern AI dashboards.
This means explaining the hidden denominator problem and showing why treating these figures as absolute metrics presents an unnecessary risk.
Contributing writers are invited to create content for Search Engine Land and are selected for their expertise and contribution to the search community. Our contributors work under the supervision of editorial staff and contributions are evaluated for quality and relevance to our students. Search Engine Land is owned by Semrush. The contributor has not been asked to speak directly or indirectly about Semrush. The opinions they express are their own.
