Research
State of AI recommendations
We asked ChatGPT, Gemini, Perplexity and Claude the same buyer questions across 10 software categories and counted who was named. Latest check: 2026-10-15. This is a dated snapshot, not a ranking of quality.
Findings
- Answers are concentrated. In 10 of 10 categories, one brand was named in at least four of every five answers.
- Engines agree more than you might expect, and differ at the edges. 28 of the 148 brand names that appeared in at least two answers (19%) were named by only one engine.
- The answer set is small. Across 200 answers, each category repeats a short list of names. A brand outside that list is easy to miss, whichever assistant a buyer uses.
By category
| Category | Named most | Share of answers | Names seen twice or more |
|---|---|---|---|
| HR software in India | greytHR | 90% | 18 |
| cloud hosting and managed databases for developers | AWS | 80% | 20 |
| email platforms for SaaS | Postmark | 85% | 16 |
| project management tools for startups | ClickUp | 100% | 15 |
| CRM for small businesses | HubSpot | 100% | 15 |
| customer support software for startups | Freshdesk | 90% | 10 |
| product analytics tools | Mixpanel | 100% | 16 |
| payroll software in India | greytHR | 85% | 12 |
| website builders for small businesses | Wix | 95% | 14 |
| accounting software for small businesses in India | Zoho Books | 100% | 12 |
Method and limits
- Five buyer-style questions per category, each asked once of every engine, without a brand name in the question.
- Brand names are taken from the answers as written, and variants of one brand are merged. A name counts once per answer.
- ChatGPT, Gemini and Claude answered from their own knowledge, with no live search. Perplexity searches the web. Answers differ by engine and change over time.
- Small sample. Treat each figure as a snapshot of the date shown, not a market share.
Want your own brand measured? Run the free check. All categories: Who AI recommends.