Atlas · C20 · Social
Communication & persuasion
Explaining, presenting, negotiating and changing minds.
AI Reliable score
7 / 10
Competent pro level · peak 9 · trend ↑ · grade A
Why this score
In a 900-person study, GPT-4 with personal information was more persuasive than humans in 64% of unequal debates (+81% odds). Labs deliberately limit manipulative persuasion, so deployed behaviour is below peak capability.
On the human scale
2Beginner. Conveys simple messages
4Average adult. Average adult: explains clearly, persuades friends
6Competent pro. Professional: presents, negotiates, sells at work ← AI today
8Expert. Expert: top salespeople, negotiators, orators
10Best humans. Best human persuaders
Future warning
At its best, AI already reaches 9 on this capability. Peak results are never counted in the index; they show where the water may rise next.
Evidence
Salvi et al., ‘On the conversational persuasiveness of GPT-4’, Nature Human BehaviourA
E23 · 2025
- Measures
- Persuasion in live debates (N = 900)
- AI result
- With personal data, GPT-4 more persuasive than humans in 64.4% of unequal pairs; +81.2% odds of agreement shift.
- Humans
- Human debaters
- On our scale
- Expert-level persuasion peak.
citations.prolific.com ↗
History
First reading: Atlas v0.1, 5 Nov 2026. Every later change will be listed here with its reason.
Disagree? Send a correction
