How we measure
Engines
- ChatGPTgpt-4o
- Claudeanthropic/claude-sonnet-4-6
- Geminigemini/gemini-2.5-flash
- Perplexityperplexity/sonar-pro
Each question is asked up to 3 times per engine each time its category is measured.
ChatGPT, Claude and Gemini answer from their own knowledge through their APIs; Perplexity searches the web.
Categories
Each software category has its own public set of buying questions, listed on its ranking page. One category is measured each week, so each is re-measured every 4 weeks.
- AI video creationMeasured 26 Oct 2026
- CRMMeasured 12 Oct 2026
- Email marketingMeasured 19 Oct 2026
- Lead generationMeasured 2 Nov 2026
These are the questions a buyer asks. They are not prompts collected from real ChatGPT users. Each ranking is the result of its public set of questions, not of every way people look for that kind of software.
Scoring
A brand earns points each time an engine recommends it, more when it is named first: 1st 1.0, 2nd 0.8, 3rd 0.6, 4th and 5th 0.5, later 0.3. The score is the average over all answers of the complete rounds, out of 100. An answer that names no brand counts as zero for every brand.
Brands discovered in answers enter the ranking only when they appear in at least 2 answers. Brands in the starting catalog have no such minimum.
Rank changes compare a category with its previous measurement, and only when both were measured with the same method.
Method version 86f98828d5fe, last updated 8 Nov 2026.