Skip to main content

How it works

We record what the AI actually said, under the same conditions every time. Then, kept separate, we tell you what we think it means.

The conditions we keep the same

So that one report can be compared with the next, every question is asked the same way.

  • A fresh conversation

    Every question starts a new conversation with the AI, with no earlier history.

  • Plain buyer language

    Questions read like a real shopper's. The discovery questions never name your brand or product.

  • Live web search

    The AI answers with live web search on, the way AI shopping assistants work today.

  • Your market

    Each report is set to one country, with its language and currency, and says which.

  • The date

    Every report states the day the answers were given.

  • One answer per question

    Each question is asked once. If a request fails we retry it. We don't average runs.

AI answers change over time, so every result is one answer on one day, not a permanent ranking.

What the AI said, and our reading

Every report keeps two things apart: the record of the answer, and our interpretation of it.

What the AI said

The products it named, the order it named them in, its reasons, and every site it cited. Saved as it came back. This is the evidence.

Our reading

Why we think the winner won, drawn from the answer and the sites it cited. The report always labels this as our reading, so you can tell it apart from the evidence.

To see whether a change worked, you run the check again after making it. One report on its own can’t tell you that.

The three checks

A report asks up to three kinds of question about your product. We never publish the questions themselves, only what each one is for.

  1. 1

    Discovery

    A buyer describes what they need without naming a brand. Does the AI bring up your product at all?

  2. 2

    A specific need

    The buyer adds one requirement that matters for your kind of product. Are you still in the answer?

  3. 3

    Head-to-head

    Only when a rival comes up: the AI compares you with that rival, and we record who it picks and why.

The loop

A report is the first turn of a loop. You make a change, then check again.

What every report tells you

Each report states the market it checked and the date of the answers, so you always know what it represents.

Want us to do this for you?

This is how the free report checks one product. For clients, we run the same checks across the products you choose on every major assistant, every week, and make the fixes ourselves.

See what’s included

Run it on your product

Any product with its own page. Not your homepage or a collection.

Or read a full sample report first