Sixty prompts asked AI to recommend an internet provider. This brand appeared in none of them.

A full prompt panel for one brand in a competitive local services category ran 46 queries across four AI engines: 184 responses total. Sixteen of those queries contained the brand name. The other 30 did not. When we split the results by query type, all 52 brand appearances came from the 16 branded queries. Across the 120 responses to queries with no brand name in them, the brand appeared zero times.

Two query types, two audiences

Not all queries in a prompt panel represent the same kind of person.

Some queries already name the brand. The person typing them knows the company exists and is looking it up:

• "[brand] reviews"
• "Is [brand] good?"
• "Alternatives to [brand]"
• "[brand] vs [competitor]: which is better?"

These are branded queries. The audience is someone who has already encountered the brand and is evaluating it. They already knew who you were before they opened an AI chat window.

Other queries have no brand name at all. The person typing them is looking for a category answer, not information about a specific company:

• "What internet service provider do you recommend in [city]?"
• "Best internet service provider in [city]"
• "Top 10 internet service providers in [city] in 2026"
• "What should I look for when choosing an internet service provider in [city]?"

These are buyer-intent queries. The person asking has not yet settled on a vendor. They are reachable. If your brand shows up in the answer, they might consider it. If it does not, they will not.

The per-intent breakdown from one panel

The 46-prompt panel covered six intent categories. Below is the mention rate for the brand in each, along with what type of queries each category contains.

Why the pattern splits so cleanly

A review query like "Is [brand] good?" tells the AI which company to write about. The engine sees the brand name in the question and builds its answer around that company. High mention rate is the expected outcome. It says nothing about whether the brand is discoverable to buyers who do not already know it.

A recommendation query like "What internet service provider do you recommend in [city]?" asks the AI to choose. The engine searches for sources it treats as authoritative on that question and writes an answer based on what those sources cover. If the brand does not appear in the places those engines look, it does not appear in the answer. The 60 recommendation queries in this panel produced zero mentions because the engines were not looking for this brand. They were answering a category question, and this brand was not in the sources they found.

The same logic applies to best-of, research, and generic comparison queries. Every buyer-intent query type asks the AI to choose or inform, not to describe a specific brand. On all four of those types, the brand appeared zero times.

What this means for measuring new-customer acquisition

Branded query performance tells you about people who already know your brand exists. Strong performance there means you have enough web presence that AI can answer questions about you accurately. That is a baseline, not a growth signal.

Buyer-intent query performance tells you whether a stranger using AI to find a provider will encounter your brand. That is the new-customer discovery number. In this audit, it was zero.

The blended 28 percent mention rate across all 184 responses hid both facts. Reading only that number, the brand looked reasonably visible. Splitting by query type showed a brand with solid branded performance and no organic discoverability at all.

How to read your own panel

If you have received an AI visibility audit or are reviewing one, check whether it separates buyer-intent from branded queries before trusting any headline number. A blended rate can be inflated entirely by branded performance and still be reported as "your AI visibility score."

The number that tells you something about growth is the mention rate on queries where the buyer has not already named you. In local services, that is recommendation and best-of queries. In B2B software, it is category comparison and feature-based queries. If that rate is zero, you are invisible to the buyers who have not heard of you yet.