AI Visibility Starts With the Right Buyer Questions
· Signal Harbor · Originally published in Signal Harbor Weekly
The Signal Harbor Journal
Research and perspectives from Signal Harbor
By Sebastian Miller
Co-Founder, Signal Harbor
AI Visibility Starts With the Right Buyer Questions
An AI visibility tool can track hundreds of questions. The report may look thorough, with scores for every platform and competitor.
Before trusting the numbers, ask a simpler question: Would the buyers you want actually face the decisions these questions describe?
A large prompt list can measure the wrong market with impressive precision. Signal Harbor starts with the customer and the buying process, then builds the tests around them.
Branded Prompts Measure Recognition
A branded prompt names the company. “What does this supplier make?” asks AI to recognize and describe a business already in the question.
An unbranded prompt asks for options without naming a provider. “Which Midwest manufacturers can supply custom components at the volumes our team needs?” asks AI to decide who belongs on the shortlist.
Both tests matter, but they answer different questions. A company can be well understood when named and rarely introduced when a buyer is still discovering options. A score that mixes those results without showing the difference can hide a discovery problem.
In a 2026 preprint studying 112 Product Hunt startups, Amit Prakash Sharma found that ChatGPT and Perplexity recognized most products when asked about them by name, but included them in fewer than 9% of the unbranded discovery tests. That is one sample of startups and AI systems, not a benchmark for every industry. It shows why branded and unbranded results should be measured separately.
Buyer Intent Changes as the Decision Gets Closer
Consider a hypothetical Chicago-area manufacturer. An early buyer might ask which suppliers serve its industry. Later, the questions become more demanding:
Which suppliers handle a particular material and production volume?
Which have the necessary quality documentation?
What do they say about lead times and implementation?
How do the finalists compare on the requirements that matter most?
This is buyer intent: the decision the person is trying to make. A useful prompt library, sometimes called a prompt bank or prompt index, organizes researched test questions around those decisions.
Signal Harbor uses product information, sales questions, buyer objections, customer segments, and available search and CRM context to choose what to test. These are researched questions, not a complete record of what people type into AI or how often they ask it.
Each Stage Reveals a Different Gap
We group tests by the work a buyer is doing.
Discovery prompts show whether the company enters consideration before its name is known.
Comparison prompts show how AI describes it alongside alternatives.
Validation prompts check capabilities, limitations, proof, and tradeoffs.
Purchase-stage prompts examine requirements such as availability, implementation, service coverage, or contract terms when those facts can be verified.
A missing mention early in research and a false statement in a final comparison create different risks. We measure them separately, then identify which matter most to the business.
Repeated Testing Makes the Result Useful
AI answers can vary across platforms, runs, and time. A 2026 preprint by Julius Schulte, Malte Bleeker, and Philipp Kaufmann explains why a single AI answer is an unreliable measure of visibility. Prompt tracking means rerunning a defined set of questions so a team can see patterns and changes. Repeatedly testing irrelevant questions, however, only gives a more stable answer to the wrong problem.
We establish a baseline using relevant questions, examine competitor recommendations and cited sources, and identify where the company is absent or misunderstood. After the marketing team or agency publishes a focused change, we retest comparable questions.
For agencies with many client brands, this prevents one generic dashboard from substituting for client-specific buyer research. For executives, it makes the score easier to interpret: which customer and which decision does it represent?
The prompt library becomes a useful business asset when it keeps the team focused on qualified buyers. Choose the questions that shape consideration. Then measure whether the answers improve.
Sebastian Miller
Co-Founder, Signal Harbor
signalharborai.com
sebastian.miller@signalharborconsulting.com
402-306-2213
This article first appeared in Signal Harbor Weekly, the Signal Harbor newsletter on AI visibility.