Which AI Writing Tools does ChatGPT recommend?

Five fixed buyer-intent questions, asked of ChatGPT with web search enabled, from a US location, twice each — ten answers per category per week. We read every answer, extract the product and company names it contains, and count the share of answers that named each brand. A brand named in nine of ten answers scores 90 percent; a brand named once scores 10 percent.

The share is a mention rate, not a quality score. It says how reliably one engine reaches for a name when a buyer asks an open question. It says nothing about whether a product is good, whether it fits you, or what it costs — and we deliberately publish no rating, star or quality ranking of anyone else’s product. A brand can be excellent and absent from this table, which is the whole problem the table exists to make visible.

Because the questions never name a brand, nothing here is prompted. Extracted names are filtered so that section headings, article authors and image alt-text are not counted as products. Standings are stored one row per week rather than overwritten, which is what makes the change block above possible. The underlying numbers are public: the same JSON this page renders is served from /api/public/leaderboard on the app.