In a 46-prompt panel, 11 responses named no internet provider. All 11 came from three prompts about the category in general.

When an AI engine answers a prompt without naming any specific brand, that response contributes nothing to your visibility score. A panel with many such responses is a panel where the prompts are doing the wrong job: asking the engine about the category rather than asking it to choose within the category. The data from one full audit makes the pattern clear.

What a null response looks like

In an AI visibility panel, a null response is a response where the engine answers the prompt but names no specific brand. The engine generated text, but none of that text mentions a company the buyer could act on. For a brand measuring its AI visibility, a null response is a non-event: neither a win nor a loss, because the engine never reached the point of recommending or mentioning anything specific.

Null responses are distinct from low-mention-rate results. If a panel runs 40 buyer-intent prompts and the brand appears in 4 of them, the mention rate is 10%, but all 40 responses named at least some brands, just not always this one. A null response is a response where the engine named nobody. The category question was answered; the recommendation question was not.

The breakdown from a real panel

A 46-prompt full audit for one regional internet service provider across four AI engines produced 184 total responses. The prompt set covered six intent types: recommendation, comparison, best-of, alternative, review, and research. The null rates by intent group were not evenly distributed.

The 41 prompts classified as buyer-intent (recommendation, comparison, best-of, alternative, and review) produced 164 responses. None of them were null. Every response in that group named at least one internet provider, even for a small regional carrier serving a mid-size Indiana town. The engines had no difficulty committing to brand names when the prompt asked them to compare, choose, or recommend.

The 5 research prompts produced 20 responses. Eleven of those 20 were null (a 55% null rate). All 11 null responses came from three specific prompts:

• "What is internet service provider and how does it work?": 4/4 engines returned null
• "What questions should I ask when evaluating internet service provider?": 4/4 engines returned null
• "What are the key features of modern internet service provider?": 3/4 engines returned null

Two other research prompts, which included a geographic anchor, returned zero null responses:

• "How much does internet service provider typically cost in Connersville Indiana?": 0/4 engines returned null
• "What should I look for when choosing internet service provider in Connersville Indiana?": 0/4 engines returned null

Why abstract prompts produce null responses

When an AI engine gets a prompt like "What is internet service provider and how does it work?", it correctly reads that as an educational question. The most useful answer is a conceptual explanation: what an ISP does, how internet connectivity works, what the infrastructure looks like. Naming Xfinity or MetroNet would not improve that answer. The engine omits brand names because the prompt does not call for them.

When the same engine gets "What internet service provider do you recommend in Connersville Indiana?", it reads the task differently. The buyer is in a specific location, they need a specific service, and the useful answer is a list of providers that serve that market. The engine names brands because the prompt requires it.

The mechanism is the same for B2B software categories. "What are the key features of modern project management software?" produces explanations of Gantt charts and sprint planning. "What project management software do you recommend for a remote team of 20?" produces named tools. The presence or absence of brand names in an AI response is less a function of what brands are well-represented in training data and more a function of whether the prompt actually asked for a recommendation.

What null responses tell you about a panel

A well-designed buyer-intent panel should produce close to zero null responses. If a panel is returning null responses at a meaningful rate (more than one or two), the most likely explanation is that some prompts are informational rather than decisional. They are asking about the category rather than asking the engine to pick within it.

Null responses are not evidence that the engine does not know about the category, or that the category is too fragmented for AI to navigate, or that a brand needs to build more awareness before it can appear in AI answers. For the regional ISP in this panel, the engines had no trouble naming specific providers, including the brand being tracked, across every buyer-intent prompt. The engines knew the market. The abstract prompts simply did not ask them to engage with it.

The diagnostic move when you see a null response is to look at the prompt, not the category. Ask whether the prompt contains a decision context: a location, a use case, a buyer type, a constraint that forces the engine to evaluate options rather than explain concepts. If those elements are missing, the prompt is testing the engine's category knowledge, not its recommendation behavior. That is a different question, and the answer will be different too.

Low mention rate versus null: not the same problem

A brand that appears in 8% of responses and a brand that appears in panels where 30% of responses are null are facing different situations. The first brand is present in the engine's decision-making but is losing the selection. The second brand may be invisible in the actual buyer-intent responses if its panel is inflated with abstract prompts that suppress the mention rate denominator.

If you run 100 prompts and 20 return null, your effective panel for measuring buyer-intent visibility is 80 prompts, not 100. A 10% mention rate on 100 prompts looks worse than a 12.5% mention rate on the 80 meaningful ones, but the underlying brand position is the same. The null responses are not relevant data. They are noise from prompts that did not test what the panel was supposed to test.

For the same reason, counting null responses separately and excluding them from the mention rate denominator gives a more accurate read of actual visibility. A panel that produces zero null responses on buyer-intent prompts is simply a better-designed panel, one where every response reflects the engine actually engaging with the brand selection task.