7 AIs ranked software in 104 categories. They agreed on #1 exactly zero times.

Everyone assumes the AIs basically agree on the best tools. They don't. SurfBloom has every major model's ranking for the same categories, so we measured it. The disagreement is the story.

0/104

categories where all AIs agreed on the #1 pick

4.61

different "best" picks named per category, on average

16%

average pairwise agreement on the #1 pick across the panel

35%

average overlap between any two AIs' top 5

Across 104 software categories, a panel of 7 leading AIs never once unanimously agreed on the best tool. Across the full panel, any two AIs chose the same #1 only 16% of the time, and their top-5 lists overlapped just 35% on average. — SurfBloom, Q3 2026

One AI's #1 is another's afterthought

The tools the panel split hardest on — best rank vs. worst rank across the AIs.

Julesas high as #2as low as #25OpenAI Agents SDKas high as #3as low as #25Gleanas high as #3as low as #25Optimizelyas high as #1as low as #20Zoho Booksas high as #1as low as #20impact.comas high as #1as low as #20UpPromoteas high as #1as low as #20Kademias high as #1as low as #20

Where AIs disagree most

Most distinct #1 picks across the panel.

Where AIs agree most

Highest average top-5 overlap.

How this was measured

Computed from every panel model's filed rankings across the 104 categories with the full panel present. "Agreement on #1" means naming the same product first; "top-5 overlap" is the average shared members between any two AIs' top five. The headline numbers use all 7 filed panel rankings. Rankings are AI opinions, refreshed quarterly, with no paid placements. Full methodology →

Get the data as JSON