What is the Prolific HUMAINE benchmark for AI models?

The HUMAINE benchmark is a vendor-neutral evaluation by Prolific that uses blind testing with real users to assess AI models on attributes like trust, performance, and adaptability in real-world scenarios.

How did Google's Gemini 3 perform in Prolific's latest test?

Gemini 3 achieved its highest-ever trust score of 69% and ranked first in performance, interaction, and trust/safety categories during Prolific's HUMAINE benchmark.

Why is blind testing important for evaluating AI models like Gemini 3?

Blind testing removes brand bias, allowing users to judge AI model outputs solely on their quality and relevance, providing a more accurate measure of performance across diverse demographics.

Home / Technology / AI Trust Soars: Gemini 3 Tops Real-World Tests

AI Trust Soars: Gemini 3 Tops Real-World Tests

4 Dec

•

Summary

Gemini 3 achieved a trust score surge from 16% to 69% in blind user testing.
The AI model ranked first in performance, interaction, and trust/safety categories.
Prolific's HUMAINE benchmark tests real-world user preferences, not just technical specs.

AI Trust Soars: Gemini 3 Tops Real-World Tests

A recent vendor-neutral evaluation by Prolific has placed Google's Gemini 3 model at the forefront of AI performance, particularly in user trust and real-world applicability. The HUMAINE benchmark, developed by researchers from the University of Oxford, utilized blind testing with 26,000 users to assess AI models across practical attributes, moving beyond traditional academic benchmarks. Gemini 3 demonstrated a remarkable increase in its trust score, rising from 16% to 69%, marking the highest score Prolific has ever recorded.

In the comprehensive evaluation, Gemini 3 secured the top position in three out of four key categories: performance and reasoning, interaction and adaptiveness, and trust and safety. While it slightly lagged in communication style, its overall consistency across 22 diverse demographic user groups, including variations in age, sex, and ethnicity, highlights its broad appeal. Users were found to be five times more likely to select Gemini 3 in head-to-head blind comparisons, underscoring its significant advancement over its predecessor.