Define the business question
Start with the decision your team needs to make and the finding that would change it. Agree the buyer situation, relevant service or product, geography or segment, competitors, AI interfaces and success criteria.
How it works
A clear line from your question
to a justified decision.
We study observed AI answers in representative buying situations. Every investigation has a defined scope, an evidence trail and an honest account of what it can tell you.
The method
Design first.
Interpret carefully.
Build representative buyer scenarios
Design questions around realistic needs, constraints and decision stages. Discovery, requirement filtering, comparison and trust questions serve different purposes. Fresh-session tests and multi-turn conversations are treated as different research instruments.
Collect and preserve observed answers
Record the exact questions and relevant conversation history, collection date, interface or model label where visible, observed answers and cited sources. Repeat observations where agreed and record unavailable data as unavailable.
Inspect the consequential evidence
Distinguish a passing mention from an actual recommendation. Check material facts and what the visible sources support. Compare alternatives within the tested conditions and separate observed patterns from plausible explanations.
Deliver a decision-ready handoff
Explain what the evidence supports, what remains uncertain and which priorities deserve action. Agree who can act on the findings. Implementation briefs, stakeholder readouts and matched rechecks are specified in the scope.
What we examine
Useful measures.
With their meaning intact.
We use the measures that serve the question. A study does not need every metric, and a single composite score cannot replace the evidence.
Mention vs recommendation
Being named in an answer is different from being recommended. We classify the role a business plays and state the eligible observations behind any rate.
Competitive consideration
Which alternatives appear, when they replace or accompany your business, and how they are positioned under the same buyer requirements.
Portrayal and accuracy
What AI associates with the business, including capabilities, limitations, service areas and important factual errors.
Source and citation evidence
Which visible sources are referenced, what their contents support and where evidence is missing. A citation does not prove causation.
Context and coverage
The buyer situations, markets, services, interfaces and collection dates actually tested. Consumer interfaces and API conditions are kept distinct.
Stability and change
Variation across agreed repeats or dates. A matched recheck can describe change without proving that an intervention caused it.
Research standards
What the research
can tell you.
And what it can’t.
A defined set of AI answers can inform a business decision. It is not a census of your buyers or an inside view of an AI model.
- Standard studies use representative scenarios. We do not access real prospects’ private AI conversations.
- Recommendations in tested answers are not consumer market share, lost sales or direct revenue attribution.
- Answers can vary with question wording, context, interface, model, date and account conditions.
- Visible citations can be investigated, but they do not reveal every influence on an answer or establish its exact cause.
- Platform coverage, repeats, evidence retention, follow-up and delivery timing are agreed in the written scope.
- A recheck can detect change under matched conditions. It does not alone prove causal improvement or business impact.
Start with the question
What is your team
trying to figure out?
Tell us the campaign, product, service or market decision you’re facing. We’ll help define whether AI recommendation research can provide useful evidence.