Metrisque
ProductAI Search MeasurementAI Recommendations
STEP 05 · SHORTLISTED

Track which brands AI recommends for each question.

After this read you can decide whether you are absent from the shortlist, or present and losing position.

We put your buyers’ real questions to each assistant, repeatedly, and record who gets named. Not mention counts: the actual shortlist, in order, with the brands that held each slot. The three assistants rarely agree with each other, and none of them answers the same way twice, which is why one screenshot of one answer tells you almost nothing.

Example result

SAMPLE · B2B SERVICES

One question, three assistants, three different shortlists.

“Best fractional CFO firm for a 40-person software company”

The same buying question does not return the same answer twice. Metrisque runs it across all three and shows you who held the slot in each.

CHATGPT
GEMINI
CLAUDE
Harrowgate Partners
#1 recommended
#2 recommended
considered
Bellweather Finance
#2 recommended
considered
#1 recommended
Ashcombe Advisory
considered
#1 recommended
not named
Kestrel & Co
not named
considered
#3 recommended
Your firm
not named
not named
not named

The corner mark is the model's own first pick. No two agree, and you are absent from all three.

Example. Real measurements, brand anonymised.
1 buyer question read on 12 August 2026. Models: ChatGPT, Gemini and Claude. Unit: rank in each model's own shortlist, positions 1 to 3, from 3 runs per model.

What this measures

INPUTS
One buyer question at a time, asked in the words your buyers use.
MODEL SET
ChatGPT, Gemini and Claude, each asked the same question in the same words.
REPEATED RUNS
3 runs per model per question, so recurrence is visible rather than a single answer.
OUTPUT
The shortlist per model with rank and order, every brand considered but not picked, and the date of the read.

What you receive

RECOMMENDED
Named in the assistant’s own picks, with its rank.
CONSIDERED
Mentioned in the answer, but not among the picks.
NOT NAMED
Absent from the answer entirely.

Next action: take the questions where you are not named into Competitor's Citations and see which sources carry the answer.

Limitations

Answers vary between identical runs; we measure across runs and models, and never promise selection.

Related measurements

Read your own site on this measure.

Your first read is free and takes about a minute. No account, no card.

Track recommendations