Updated our benchmark from 412 to 1,800 assessments — the averages barely moved

We run aisa.to -- conversational AI skills assessment. Been collecting data since launch and just hit 1,800 completed assessments, up from 412 when we first published the report.

The surprising thing: average score moved from 52/100 to 48/100. More data, slightly worse average. 67% still score below Proficient (was 62%).

What DID change: Safety awareness dropped further, now the weakest area. Tool Landscape awareness is 4.8/10 -- most people use one tool for everything. The overestimation gap held steady at +19 points. People consistently think they're about 19 points better than they actually are.

The uncomfortable takeaway: the market for AI skills assessment is growing but the average user still doesn't know they need it. You can't sell people a mirror they don't think they need.

Building in public because this kind of data should be shared. Full report: aisa.to/ai-fluency-index