Am I Recommended? / Case study / Our own AI visibility
Public self-audit · September 2026

We measure ourselves before we sell the fix

Am I Recommended? began with no measured recommendation visibility in its original benchmark. We preserved that starting point, improved the public representation of the business, then ran a controlled comparison. The first movement is now measurable.

The starting point stays frozen

0/40OpenAI visibility in the historical benchmark
0/40Gemini visibility in the historical benchmark
Baseline preservedthe original result is not overwritten when later comparisons improve
The baseline is a point-in-time measurement, not an external certification or permanent score.

Then the first measurable movement appeared

In a later clean comparison, OpenAI surfaced Am I Recommended? in four of five Discovery & shortlist questions.

4/5OpenAI Discovery & shortlist appearances
#1–#4positions recorded across those four appearances
First liftfrom no measured OpenAI visibility to repeated discovery appearances
Important boundary: the comparison records what changed in the observed outputs. It does not reveal a secret model penalty and does not prove that any single intervention caused the movement.

What we are willing to publish

We improved the clarity, consistency and public corroboration around the business, then compared recommendation visibility again. That is enough to show that measurable movement is possible.

We do not publish the exact source correlations, detailed question taxonomy, provider configuration, internal weighting, intervention sequence or other research that could be copied as a competing implementation playbook.

What the comparison taught us

Discovery can moveOpenAI began placing Am I Recommended? into Australian AI-visibility shortlists.
Different buyer contexts behave differentlyWinning broad discovery did not automatically make us the answer for every type of buyer question.
External corroboration is worth investigatingSuccessful observations contained stronger public corroboration around the business and category. We treat that as a research signal, not a hidden ranking formula.
Brand recognition is a longer gameEstablished entities can retain an advantage, which is why we continue building a broader, truthful public footprint.

What customers get — and what remains our IP

Customers receive their measurement record, the relevant competitive evidence, reviewed gaps and prioritised next steps. They do not need our private research dataset or the exact rules we use to turn recurring evidence into an investigation priority.

Publish proof. Sell the conclusion. Keep the mechanism private.

Why the baseline still matters

Later comparisons only mean something if the starting point is preserved. We keep enough internal provenance to make controlled comparisons without publishing the full configuration publicly.

A later result still does not prove that one particular intervention caused the change: provider variability and changes elsewhere on the public web can also affect results.