How Good Am I at AI? An Honest Way to Find Out

"How good am I at AI?" It is a question that did not exist three years ago. Now it is one of the most common things professionals quietly wonder — not "do I use AI" (almost everyone does) but whether they are using it well, or leaving enormous value on the table without realising it.

The honest answer for most people: you are probably better at some dimensions than you think and worse at others. AI skill is not a single number — it is a profile. And the data from 1,800 assessed professionals tells a clear story about where that profile typically lands.

The Average AI Score Is 48 Out of 100

Across 1,800 completed AI fluency assessments, the median composite score is 48 out of 100. That places the typical professional squarely in the Developing tier — using AI regularly but with meaningful gaps in critical thinking, technical understanding, and safety awareness.

Here is how the full population breaks down:

Nearly half of all professionals land in the Developing tier. They get real value from AI but lack the critical thinking and technical understanding that separate fluent users from literate ones.

Because the sample is self-selected — these are people curious enough to take an assessment — the true workforce average is likely lower.

The Skill Spectrum Is Wider Than You Think

Most people imagine AI ability as a simple scale: beginner, intermediate, advanced. In practice, it looks more like a radar chart. You might be excellent at prompting but weak at evaluating output. You might know the tool landscape better than anyone but never consider safety implications.

AISA's research across 1,800 assessments has identified 10 distinct AI personas — and which one you are says more about your AI relationship than a single score:

The most common surprise is discovering your persona does not match your self-image. A marketing director who considers herself "pretty advanced" might discover she is a Tactician — efficient with mainstream tools but not yet orchestrating AI across her full workflow. A junior developer who feels behind might be a Builder — already creating things that don't exist yet.

Where Professionals Actually Score Worst

The five dimensions of AI fluency are not equally hard. Data from the AISA AI Fluency Index reveals which ones professionals struggle with most:

Technical Understanding is the weakest dimension across all professionals — averaging just 41 out of 100. This is the knowledge gap that leads to poor tool choices, unrealistic expectations, and the inability to diagnose why AI output is wrong. Most AI training focuses on prompting, but the data says technical understanding is the bigger blind spot.

Safety & Responsibility scores 42 — meaning most professionals have not internalised responsible AI use. They use AI, sometimes daily, but have not thought about data boundaries or when AI should not be trusted.

You Probably Overestimate Your AI Skills

Here is the uncomfortable finding: professionals overestimate their AI skills by an average of 18.5 points on a 100-point scale. Someone who rates themselves at 65 typically scores around 47.

This is not unique to AI — the Dunning-Kruger effect applies to any domain where the skills required to be good at something are the same skills required to recognise you are not good at it. But in AI, the gap is unusually wide because:

• AI tools are designed to feel easy. ChatGPT gives you a confident-sounding answer whether your prompt is excellent or terrible. The tool does not signal that you are using it poorly.
• There is no visible failure mode. When you write bad code, it crashes. When you write a bad prompt, you get a plausible-sounding answer that might be wrong — and you may never know.
• Most people compare themselves to non-users. "I use AI every day" feels advanced when your reference point is colleagues who have never tried it. But daily use and skilled use are different things.

How Do Different Roles Compare?

AI skill varies by profession, but not always in the direction you would expect:

Product managers outscore every other role at 56.4, driven by the highest Workflow score (60.2) of any profession. They are excellent at integrating AI into how they work but less confident about how the models actually work.

Engineers score highest on Technical Understanding (50.4) but have the widest Safety gap of any technical role — they focus on building, not on the downstream impact of what they build.

HR professionals face an irony: they increasingly evaluate AI skills in candidates while scoring below average on the very skills they assess. Their Technical Understanding (35.8) is the lowest of any role. For the full breakdown by role, see the individual analyses for engineers, product managers, marketers, HR, sales, and data analysts.

Five Things That Separate Skilled AI Users

After analysing what distinguishes different proficiency levels across 1,800 assessments, five markers consistently separate skilled AI users from everyone else:

1. They Iterate, Not Just Prompt

The biggest single predictor of AI skill is not what you type first — it is what you type second. Skilled users treat AI as a conversation, not a vending machine. Anthropic's AI Fluency Index confirmed this: users who iterate are 5.6x more likely to question AI reasoning. Iteration is the gateway behaviour.

2. They Verify with a Specific Method

Everyone says they "check AI output." Skilled users can describe how. "I cross-reference factual claims against primary sources, test code in a sandbox before using it, and flag hedging language as a review trigger" is worlds apart from "I read it over." This is the Critical Thinking dimension — and the 44-point average shows most people are closer to "I read it over."

3. They Match the Tool to the Task

Using ChatGPT for everything is like using a hammer for every job. Skilled users know when to reach for Claude, when Perplexity is better for research, when a custom GPT beats a generic conversation, and when AI is not the right tool at all. This is the Tool Landscape skill — the single lowest-scoring criterion at 4.9 out of 10.

4. They Think About Structure

Beginners type a question. Intermediate users add context. Advanced users think about prompt architecture — role definitions, constraints, output format, few-shot examples. The same request can produce dramatically different results depending on structure. This is where the gap between a score of 45 and a score of 75 lives.

5. They Consider the Implications

What data am I sharing? What happens if this output is wrong? Who else will see or use this? The Safety dimension averages 42 across all assessments — and marketers score just 32.6, the lowest of any role on any dimension. This blind spot is the most common gap across all AISA assessments.

How to Actually Measure Where You Stand

You cannot measure this with a quiz. The skills above are demonstrated in context — in how you interact with AI, think through problems, and evaluate what comes back. Self-assessment is unreliable: the 18.5-point overestimation gap means your internal estimate is systematically wrong.

The AISA AI fluency assessment measures this through a 20-minute conversation with an AI facilitator that adapts to your skill level and profession. The assessment covers 11 criteria across all five dimensions, validated against Anthropic's AI Fluency Index (93% criteria overlap).

You walk away with:

• A composite score and tier classification out of 100
• Five dimension scores showing your actual profile shape
• Your AI persona — which of the 10 types you are
• Personalised recommendations for what to work on first

The assessment and full report are free. No preparation needed — just an honest conversation about how you use AI in your work.

---

Related reading: AI Literacy vs Proficiency vs Fluency — what the three levels mean and how they map to scores.

Related reading: AI Skills Gap Analysis: Real Data — the full gap analysis across roles and dimensions.

Related reading: What Is AI Fluency? — the definitive guide to what AI fluency means and how it is measured.

Frequently Asked Questions

How good is the average person at AI?

Across 1,800 AISA assessments, the median composite score is 48 out of 100 — solidly in the Developing tier. Most professionals cluster between 28 and 59. The distribution is right-skewed: genuinely advanced users (80+) represent just 8% of assessed professionals, while the Developing range (28-59) holds 47%.

Can I improve my AI skills quickly?

Yes — the fastest path is targeted practice on your weakest dimension, not generic AI training. If your gap is Critical Thinking, start verifying every AI output against a source. If it is Prompting, learn structured prompt techniques. If it is Technical Understanding — the lowest-scoring dimension across all professionals — invest time understanding how models actually work, not just how to use them.

Is there a free way to find out how good I am at AI?

AISA offers a free conversational assessment — 20 minutes, 11 criteria, a full personalised report with dimension scores, persona classification, and growth recommendations. It is the most thorough free option available. For lighter alternatives, LinkedIn has basic AI skill quizzes and Google AI Essentials includes knowledge checks — but these test factual recall, not demonstrated skill.

How do AI skills compare across different professions?

Product managers score highest at 56.4 on average, driven by strong Workflow Integration. Engineers follow at 54.8 with the best Technical Understanding. HR and Marketing trail at 43.7 and 44.3 respectively, with the largest gaps in Technical Understanding and Safety. Within any role, the spread between the highest and lowest scorer can exceed 70 points — suggesting individual variation matters more than profession.