How the AI Shortlist works

Positions are not for sale

No brand can pay to be listed, to move up or to be left out. Every ranking comes from the answers AI assistants gave, counted the same way for every brand, and customers of Beket are ranked by the same method as everyone else.

The method

• Each category in each place is measured with 3 unbranded prompts: questions a buyer would ask that name no company.
• Each prompt is run once on each AI engine listed on the edition.
• Brands are ranked by how many of those answers named them. Brands named in the same number of answers share a rank.
• Countries are measured every month; cities every quarter.
• Movement compares a brand's rank with the previous edition. A change of fewer than 2 places shows as steady, so small run-to-run noise does not read as a trend.
• A brand we could not match to a website is marked unverified: it is listed, but has no brand page and no movement.

How sentiment is scored

Sentiment is an automated reading of the sentences that name a brand. It is not a verdict on the brand, and it does not measure how well the brand performs.

• Each sentence that names the brand is scored by a word-list model, then the scores are averaged for the brand.
• The average is placed in one of three bands, shown on a scale from 0 (most negative) to 1 (most positive).
• A brand needs at least 3 sentences that name it. Below that, the table shows "Not enough data".
• Sentiment is scored for the ten brands at the top of each edition. Other brands show no sentiment.

Corrections

If an edition names the wrong brand, links the wrong website or misses a merger, tell us through the contact form. A corrected edition shows the date it was corrected and what changed.

Your own AI visibility

The Shortlist shows a few prompts per category. To see how AI assistants describe your brand across every prompt your buyers ask, see what Beket measures.