Run the prompt yourself.

How we capture

What makes proof worth anything.

• Where the engine allows it, we link to the original answer on the engine's own domain. You are not trusting an image we produced, you are reading the answer at the source.
• We publish the exact prompt, character for character. If a prompt needed a market added to it, that is part of the prompt we show you.
• We prefer logged-out sessions. A logged-in session can be shaped by chat memory and past conversations, which makes the answer personal rather than representative.
• We state the market. AI shopping answers change completely by country, so a capture without a stated region tells you very little.
• We date every capture. AI answers move week to week, and a screenshot with no date is a claim with no shelf life.
• Screenshots are unedited. No compositing, no retyping, no cropping that hides the brands we lost to.

Off-page work

An answer is only as good as its sources.

Engines do not invent recommendations. They assemble them from pages they already trust, which is why most of this work happens off the client’s own site. These are placements we earned. The clearest example is Fortune: it reviewed Create in June, ChatGPT cites Fortune, and ChatGPT now names Create first.