The same brand, three different prompts, three completely different visibility scores

Across 270 live B2B SaaS panels, we found that a brand's visibility score can swing from near zero to near-certain depending on which prompt type you measure. Not because the brand is better or worse known, but because each prompt type causes AI engines to retrieve a different set of pages, and each page set contains a different set of brands.

The same brand, asked three ways

Take a mid-size customer support software company. Ask an AI three different questions:

First: "Is [brand] good for support teams?" That question names the brand. The AI engine retrieves review pages about that company and answers based on what it finds. The brand is mentioned in the response. Mention rate for this query type across our panels: 70-90%.

Second: "What customer support software do you recommend for a 50-person team?" That question names no brand. The AI retrieves category ranking pages and recommendations lists. Only the brands that appear consistently on those pages make it into the answer. In our 270 panels, 87.4% of brands scored zero on this prompt type.

Third: "How does customer support software help reduce ticket volume?" That question asks about a category concept, not a specific vendor. The AI retrieves educational content. It may name a few prominent brands as examples, or none at all. Across our panels, this prompt type produced the lowest brand mention rates of any intent category.

Same brand. Same AI engines. Three different prompt types. Three different answers about the brand's visibility.

Why prompt type changes the page set

Search-grounded AI engines answer questions by retrieving pages that rank well for that query and synthesizing what those pages say. The question determines which pages rank, and the pages determine which brands appear.

A recommendation query like "best customer support software" retrieves pages that rank for that phrase: G2 category leaders lists, editorial roundups from software review sites, and comparison guides from digital publications. A small number of brands appear consistently across those pages, and they are the ones that appear in the AI response.

A research query like "how does customer support software reduce ticket volume" retrieves educational content: vendor blog posts, how-to guides, and analyst explainers. These pages may mention categories and generic workflows without naming specific vendors, or they may name a few as examples. The brands that appear here are not necessarily the same ones that dominate recommendation queries.

A branded comparison query like "Is [brand] good?" retrieves review pages specifically about that company. Those pages exist and rank because the brand has some market presence. The brand appears in the response almost by definition.

What the 270-panel data shows

We ran live 10-prompt panels across 270 B2B SaaS brands in three categories: customer support software, project management software, and HR software. Each panel included a mix of prompt types: branded review and comparison queries where the target brand was named, and organic recommendation, best-of, and research queries where no brand was named in the question.

The split is stark.

Across the 270 brands: 247 were mentioned at least once somewhere in their panel. But only 31 of those 247 (12.6%) received a recommendation in any organic prompt. The other 87.4% showed up only when the prompt named them.

The blending problem

A panel that mixes branded and organic prompts produces a blended mention rate. A brand with a 75% branded mention rate and 0% organic mention rate might score a 30% blended rate. That 30% is technically accurate, but it tells you almost nothing about new-customer discoverability.

The 30% organic mention rate threshold is roughly the 75th percentile across the 270 brands we measured. Reaching it means appearing in organic recommendation and best-of queries with some consistency. Reaching it on a blended score while sitting at 0% organic means nothing has changed for new-customer discovery.

The number worth tracking is the organic-only mention rate: prompts where no brand was named in the question, and your brand still appeared in the answer. That is the rate that tells you whether buyers who have never heard of you will encounter you.

Which prompt type to optimize for

The buyer journey has three distinct moments where AI gets consulted, and each maps to a prompt type.

Early-stage discovery: the buyer does not know which vendors exist. They ask "what customer support software should I use?" This is a recommendation or best-of prompt. Brands that are not on the editorial ranking pages that answer this question are invisible at this stage.

Mid-stage evaluation: the buyer has a shortlist and is researching options. They ask "Is [brand] good?" or "How does [brand] compare?" This is a branded review or comparison prompt. Brands that have review coverage on major review platforms and editorial sites do fine here. Most mid-size brands with any market presence perform reasonably well on these.

Late-stage research: the buyer is trying to understand how the category works. They ask "how do support tickets get routed automatically?" This is a research prompt. Brands that produce educational content indexed at scale can appear here, but it is the least predictable of the three stages.

If your organic mention rate is low, the bottleneck is almost always the early-stage discovery prompts. The fix is not better product pages or more branded review responses. It is placement on the editorial and comparison sources that rank for category recommendation queries: the pages the engine retrieves when no brand is named in the question.