This stage turns customer situations into questions you can ask AI. You will record whether the business appears and how it is described, then compare the answers with your business profile.
Prepare two types of question
Branded questions name your business. They test recognition and the description AI gives. Non-branded questions describe a customer need without naming you. They test whether AI introduces the business as a relevant option.
What does [business name] offer, who is it for and where does it operate? Explain the practical details and why someone might choose it. Include the sources you use.
I am [customer situation] in [location]. I need [offering or outcome] with [important requirements]. Which providers should I consider, and why? Include the sources you use.
The workbook's AI answers tab can compose these prompts from your inputs. Use a fresh conversation for each test and retain the full answer. Record the system, date, whether web search was available, and any context you supplied.
Look for a pattern
Repeat your core questions in fresh conversations and across the systems in your scope. Use the same wording for comparisons. Record all results, including answers that omit the business. Keep the question set manageable enough to repeat.
In the workbook, record the prompt, answer, sources and conditions in the observation record. Use P1–P8 to examine Presence and I1–I11 to examine Interpretation.
Choose the outcome scores
Presence moves from absent or unreliable recognition at 1, through occasional consideration at 3, to consistent, prominent consideration when relevant at 5. Interpretation moves from materially wrong at 1, through broadly accurate but incomplete at 3, to accurate and appropriately differentiated across situations and time at 5.
Use the full anchors shown in the workbook. A mention is different from a recommendation. Judge Interpretation against substantiated facts, including suitability and limitations. If no description appears, leave Interpretation unassessed.
Record evidence strength: low for limited or incomplete observations; medium for repeated observations across most priority situations; high for repeated coverage across the agreed scope, systems and time. Level 5 needs a sustained pattern.
Capture the important gap
For example: “AI recognises our kayaking tour but repeatedly omits the minimum age when discussing family activities.” Name the customer consequence and retain the answer that demonstrates it.
Record your answers and outcome scores, then assess the information foundations.