The AI Fluency Report
AI fluency measured from 1,800 conversational assessments — not surveys, not quizzes. The only benchmark built on a published rubric with evidence-linked scoring.
Key Findings from 1,800 AI Fluency Assessments
The least AI-fluent professionals overestimated their score by 40 points. The most fluent underestimated by 27. The Dunning-Kruger effect is a 67-point perception gap.
Two out of three professionals fail to reach AI Proficiency when measured through conversation — not self-reported on a survey.
Same company, same tools, same training budget. The gap between the least and most AI-fluent employee in one team was 82 points (15 to 97).
Only 15 out of 1,800 professionals reached Expert-level AI fluency. Your company probably doesn't have one.
Product managers outscore engineers on AI fluency (59.2 vs 53.7). Applied judgement beats technical knowledge.
The average professional overestimates their AI skills by 19 points. Self-assessment is inversely correlated with actual ability.
The average AI fluency score across 1,800 professionals is 48 — squarely in the Developing tier. Most people use AI regularly but without systematic practice.
The most common AI persona is The Dabbler (340 people). The rarest is The Oracle (2 people). The distribution is a long tail with almost nobody at the top.
The weakest AI skill isn't prompting — it's Tool Landscape. Most professionals use one AI tool and have stopped looking for alternatives.
Company-wide AI training assumes everyone starts from the same place. The data says starting points vary by 5x within the same team.
Self-reported AI literacy correlates with actual AI fluency at r = 0.07 to 0.24. That's barely above random. Surveys are not measurement.
The skills that transfer from non-AI work rank highest (5.4/10). The AI-specific skills that require deliberate learning rank lowest (4.8/10). Nobody is investing in what actually differentiates.
AI Fluency Score Distribution
How 1,800 professionals scored across the full 0–100 range. The distribution peaks in the 30–60 band — a broad middle of people who use AI regularly but lack the systematic practices that separate competent from proficient.
The AI Skills Prediction Gap
Before starting, 307 candidates predicted their score. The overall gap is +19 points. But the real story is how the gap changes by tier — a textbook Dunning-Kruger pattern measured at scale.
The least fluent professionals overestimate by 40 points. The most fluent underestimate by 27. Self-assessment is inversely correlated with actual ability.
| Tier | Predicted | Actual | Gap | n |
|---|---|---|---|---|
| Emerging | 57.5 | 17.5 | +40 | 85 |
| Developing | 60.5 | 43 | +17.5 | 149 |
| Proficient | 72.2 | 68.7 | +3.5 | 80 |
| Advanced | 83.4 | 83.8 | -0.5 | 11 |
| Expert | 67.2 | 94.6 | -27.4 | 5 |
Consistent with Zhang et al. (LAK26): self-reported AI literacy correlates with objective measures at r = 0.07–0.24.
AI Fluency Scores by Criterion
All 11 criteria cluster between 4.8 and 5.4 — a narrow band around the Competent threshold. The rank order reveals which skills transfer from non-AI work (top) vs which are AI-specific and need deliberate investment (bottom).
Strongest: Task Decomposition (5.4) and Context Management (5.3) — transferable skills. Weakest: Tool Landscape (4.8) and AI Fundamentals (5.0) — AI-specific knowledge requiring deliberate learning.
Curious about your AI Fluency?
AISA helps you measure, prove and improve your AI skills — free report in a 20-minute chat.
AI Fluency by Role
Product managers outscore engineers — not because they know more about AI, but because they decompose tasks better, evaluate outputs more critically, and think at the workflow level.
Technical knowledge matters less than applied judgement. A quiz that tests AI knowledge would reverse this ranking.
The 10 AI Fluency Personas
AISA classifies each candidate into one of 10 personas based on the shape of their dimension scores — not the composite number. The most common persona is The Dabbler (29%). Only 2 professionals earned The Oracle classification.
Full persona profiles — what each persona means for hiring and development.
Within One Organisation
An anonymised 8-person team. Same company, same access to tools, same training budget. Scores ranged from 15 to 97. The gap between the least and most fluent employee is 82 points.
Company-wide AI training assumes uniform starting points. The data suggests starting points vary by 5x.
How AI Fluency Is Measured
Every number in this report comes from AISA — a conversational AI fluency assessment that uses a dual-track architecture, a published 11-criterion rubric, and a full-transcript calibration pass.
What makes this data different
- Measured from demonstrated behaviour, not self-report
- Every score tied to a verbatim quote from the conversation
- Dual-track: separate conversation and evaluation models
- Calibration pass by Claude Opus reviews the full transcript
- Published rubric with behavioural anchors at every score level
Cross-referenced against
- Anthropic AI Fluency Index — 93% marker coverage (9,830 conversations)
- U.S. Department of Labor AI Literacy Framework — 100% sub-competency coverage
- Self-audit: predictive validity 3.5/5, reliability 4/5
- Formal validity study in progress with university research partner
Limitations
- Self-selected sample. Professionals who seek out an AI fluency assessment are more AI-curious than the average knowledge worker. The true workforce distribution likely skews lower.
- Single-session measurement. A 20–40 minute conversation samples behaviour. It does not observe it longitudinally. A strong session may not reflect daily habits.
- Communication style matters. The conversational format advantages articulate communicators. Scoring separates AI proficiency from language fluency, but the format is not neutral on expressiveness.
- External validation pending. Self-audit ratings (predictive validity 3.5/5, reliability 4/5) reflect internal assessment. A formal validity study with a university research partner is in progress.
AI Fluency Report FAQ
What is the average AI fluency score?
Across 1,800 completed AISA conversational assessments (August 2026 edition), the average AI fluency score is 48 out of 100, with a median of 48. The sample is self-selected professionals, so the true workforce average is likely lower.
What is a good AI fluency score?
A score of 60+ reaches the Proficient tier (top 33% of professionals). 80+ reaches Advanced (top 8.4%). Only 1.3% of professionals reach the Expert tier (92+). The Proficient threshold requires demonstrated, consistent AI practices — not just regular use.
How is AI fluency measured?
AISA measures AI fluency through a conversational assessment: candidates demonstrate how they actually work with AI in a 20-40 minute adaptive dialogue while a separate AI model silently scores 11 criteria across 5 dimensions (Prompting, Critical Thinking, Technical Understanding, Workflow, and Safety). A calibration pass reviews the full transcript. Every score is tied to a verbatim quote. The framework covers 93% of Anthropic's AI Fluency Index markers.
How accurate are self-reported AI skills?
Not very. AISA data shows professionals overestimate their AI fluency by an average of 19 points (predicting 64, scoring 45). The gap is largest for the least skilled: Emerging-tier professionals overestimate by 40 points. This is consistent with Zhang et al. (LAK26), who found self-reported AI literacy correlates with objective measures at r = 0.07-0.24.
How often is this report updated?
The AI Fluency Report is updated quarterly. Each edition includes all prior assessments — the dataset is cumulative. The August 2026 edition covers 1,800 completed assessments.
See where you stand
The assessment is free. The report is free. 20–40 minutes of conversation — scored against the same rubric that produced every number on this page.
Edition history: Q2 2026 Baseline (N=412) · AI Literacy 2026 (N=1,017) · August 2026 (N=1,800) — current
The Science Behind AISA
In 2026, Anthropic published the AI Fluency Index — the largest empirical study of AI fluency to date, analysing nearly 10,000 conversations. AISA covers 93% of the behaviours Anthropic identified as markers of AI fluency and goes even deeper with 4 additional dimensions. The U.S. Department of Labor's AI Literacy Framework (TEN 07-25) defines what every worker needs to know about AI — AISA covers 100% of its 25 sub-competencies.Read our analysis: Anthropic's AI Fluency Study & AISA · DOL AI Literacy Framework & AISA
