1,000 AI skill assessments in — the gap nobody's measuring

crossed 1,000 assessments on aisa.to last month. the data pattern that keeps surprising me:

the biggest AI skill gap isn't prompting or tool knowledge. it's verification — whether people actually check if AI output is correct before using it. Safety and critical evaluation score lowest at 45/100 across all users. average overall: 52/100.

62% of people fall below Proficient. and the highest scorers aren't the most technical — they're the ones with a repeatable process for validating output.

this is showing up in hiring conversations too. companies want "AI skills" but measure it by asking "do you use ChatGPT?" which tells you approximately nothing.

full data from 1,017 assessments: aisa.to/blog/ai-skills-report-2026-what-1017-assessments-reveal