I asked the same buying question across different AI assistants. The product rarely came up, even though every assistant could describe it accurately when named.
WORK / 01
When AI becomes the shelf,
who gets picked up?
THE QUESTION
AI can describe a product accurately when you name it, yet never bring it up when you ask for a recommendation.
THE FINDING
I tested the same question across different models, again and again. The recommendations kept changing, but the sources behind them didn't.
SO WHAT
Visibility is built upstream: be credibly present in the sources the answer is made of.
THE SAME FIVE SOURCES, EVERY TIME
- 01Brand-owned information
- 02Retail and e-commerce content
- 03Independent reviews
- 04Creator coverage
- 05Community discussion
HOW I TESTED
Repeated category-entry questions across multiple AI assistants and model versions, using fresh sessions and public information only.
Then the same question, again and again, in fresh sessions. The recommendations changed every time. The sources they were built on didn't.
Finally I checked the AI's decision criteria against real consumer discussions. They matched.
Being known by AI is not the same as being chosen.
THE PRACTICAL QUESTION BECOMES:
Are we credibly present in the sources AI draws from?
WHAT CAME OUT OF IT
Two things: a repeatable way to check how a brand shows up in AI answers, and a map of the five sources those answers are built on. It won't rank the market. It shows where to be present.
An independent experiment using public information. Brand, category and product details have been generalised to protect confidentiality.