What's a Good AI Score? [2026 Data]

What's a Good AI Score? Here's What 1,219 Assessments Tell Us

The average AI fluency score across 1,219 completed AISA assessments is 48 out of 100. The median is also 48. That places the typical person squarely in the Developing tier (28–59), well below where most people predict they'll land. A good AI score — one that puts you into the Proficient tier — starts at 60. Advanced begins at 80. Expert at 92.

If you expected higher, you're not alone. Among the 321 people who predicted their score before taking the assessment, the average prediction was 63.6 against an average actual of 44.8 — an overestimation gap of 18.8 points. Engineers predicted 71.6 and scored 56. The gap is real, it's consistent, and it's worth understanding before you decide what "good" means for you.

This post breaks down how the 0–100 composite score works, what each tier actually represents, why your persona matters as much as your number, and what it concretely takes to move up.

How the 0–100 AI Fluency Score Is Built

The AISA composite score isn't a single test result — it's a weighted aggregate across five dimensions, each assessed through a conversational exchange with an AI facilitator and scored independently by a separate AI evaluator. Understanding the architecture matters because two people with the same composite can have very different skill profiles.

The Five Dimensions and Their Weights

AI fluency isn't one skill. AISA measures it across 11 criteria grouped into five dimensions:

Workflow & Application carries the most weight because it reflects whether you can actually do things with AI, not just talk about them. Technical Understanding and Safety score lowest across the population — these are the dimensions where most people have the biggest blind spots.

The Tier Bands

Each criterion is scored 1–10, then aggregated into a 0–100 composite. The tiers:

You can see the full scoring methodology in the AISA rubric.

Where the Population Actually Sits

With an average and median of 48, the center of mass is in the lower half of Developing. The persona distribution (covered below) gives a more granular picture, but the headline: most people are not yet Proficient. The bar for "good" is lower than the internet discourse would suggest.

What a Good AI Score Looks Like by Role

Context matters. A score of 48 means different things for a student exploring AI for the first time versus an engineering lead who's supposed to be driving AI adoption across a team.

Role-Level Benchmarks

Engineers score highest on Technical Understanding (51.5) relative to the population average (41.0) — no surprise there. But even engineers average only 55, which is still Developing. Product managers and founders cluster around 56–57, also Developing. Students sit at 37, early Developing.

Safety & Responsibility is the weakest dimension for every single role. This is worth flagging if you're an engineering manager evaluating your team's readiness: the gap isn't just in prompt engineering — it's in understanding what not to do.

The Prediction Gap by Role

Engineers who predicted their score before the assessment (n=66) expected to land at 71.6 — solidly Proficient. They actually scored 56.0. That 15.6-point gap is the smallest role-specific gap in the data. The overall gap is 18.8 points. People consistently overestimate by roughly one full tier.

For a deeper look at this overestimation pattern, see AI Skills Gap: You Overestimate by 18.5 Points.

The Persona Lens: Why the Same Score Means Different Things

AISA doesn't just give you a number — it assigns one of 10 personas based on your score profile across dimensions. This is where interpretation gets interesting, because a composite of 55 as a Sceptic signals a fundamentally different skill shape than 55 as an Enthusiast.

Current Persona Distribution

Note: Oracle (the Expert-tier persona) doesn't appear in the distribution yet — fewer than 30 assessments have reached that band.

What the Personas Reveal That Scores Don't

The largest single group is Dabblers at 27.6% — people who've tried AI tools but haven't built consistent habits or critical evaluation skills. Combined with Bystanders and Copy-Pasters, that's nearly 38% of all assessed individuals sitting at or below 30.

The second-largest group is Enthusiasts at 22.1%. These are people who are genuinely excited about AI and use it regularly, but their average score of 52.2 reveals a gap between enthusiasm and effectiveness. They tend to score well on Prompting but fall short on Critical Thinking and Technical Understanding. Enthusiasm without evaluation is a pattern, not a strategy.

Sceptics (5.9%, avg 42.8) are an interesting counterpoint. Their Critical Thinking scores tend to be relatively stronger — they question AI outputs — but they underinvest in learning how to use the tools well. A Sceptic at 55 might have strong evaluation instincts but weak workflow integration. An Enthusiast at 55 might have the opposite profile. Same number, different growth paths.

Builders (17.3%, avg 71.1) and Conductors (4.6%, avg 70.7) represent the Proficient tier — people who've moved past casual use into deliberate, integrated AI workflows. Builders tend toward technical depth; Conductors toward orchestrating AI across team processes.

Architects (4.4%, avg 87.6) are the rarest group with enough data to report. They demonstrate strong scores across all five dimensions and typically design AI systems or strategies for organizations.

To find out which persona you are, you can take the AI fluency assessment — the persona assignment is part of the results.

Why the Bar Is Lower Than You Think

If you scored 55 and felt disappointed, recalibrate. You're above the median. If you scored 60, you've crossed into Proficient — a tier that only about 34% of assessed people reach (Tacticians + Conductors + Builders + Architects combined).

The Clustering Effect

The bulk of the population clusters between 27 and 59 — the Developing tier. This isn't because the assessment is hard for the sake of being hard. It's because most people's AI usage follows a predictable pattern:

1. They prompt, but don't iterate. A single prompt, accept the first output, move on. Effective prompting involves structured iteration, constraint-setting, and role definition.
2. They use AI, but don't evaluate it. They can't reliably distinguish a confident-sounding wrong answer from a correct one. This is the Critical Thinking gap.
3. They know tools, but not mechanisms. They use ChatGPT or Claude daily but can't explain what a context window is, why it matters, or how token limits affect output quality. With models like GPT-5.6 Sol now offering over 1 million tokens of context and DeepSeek V4 Flash at 1 million, understanding these constraints is increasingly practical, not theoretical.
4. They skip safety entirely. Data handling, bias, hallucination risk — these register as abstract concerns rather than daily practice.

This pattern is consistent with external research. Anthropic's AI Fluency Index — against which AISA is validated with 93% overlap — identifies similar clustering in the middle tiers. The World Economic Forum's Future of Jobs Report 2025 found that 63% of employers cite skills gaps as the primary barrier to business transformation, and AI literacy is among the top 10 fastest-growing skill demands globally.

What Moving Up One Tier Actually Takes

Moving from Developing (28–59) to Proficient (60–79) isn't about learning more prompts. It's about building three specific capabilities:

1. Structured evaluation habits. Can you look at an AI output and articulate why it might be wrong? Can you identify when a model is confabulating versus when it's reasoning from insufficient context? This is the single biggest differentiator between Developing and Proficient.

2. Workflow integration. Proficient users don't just "use AI sometimes." They have repeatable patterns: specific tools for specific tasks, clear handoff points between AI-generated and human-verified work, and an understanding of when AI is the wrong tool. AISA's Workflow & Application dimension (25% of the composite) is the highest-scoring dimension on average at 49.4, which means it's where people are closest to the Proficient threshold — and where focused effort pays off fastest.

3. Technical grounding. You don't need to train models. But you do need to understand the basics: what models can and can't do, how context length affects quality, why temperature and sampling matter, and what the difference is between retrieval-augmented generation and fine-tuning. Technical Understanding averages 41.0 across the population — the lowest of all five dimensions. Even modest improvement here moves the needle.

Stanford's 2025 AI Index Report found that while AI adoption in organizations has more than doubled since 2017, individual skill development has not kept pace — a finding that maps directly to the clustering we see in the Developing tier.

The Motivation Signal

AISA data also shows score variation by why people take the assessment:

People motivated by leadership score highest, which makes sense — they're typically further along in their careers and have more context for where AI fits. Career transitioners score lowest, which also makes sense — they're starting from a different baseline. Neither group's score is inherently better or worse; the context determines what "good" means.

How to Use Your AI Score

A score is only useful if it changes what you do next. Here's a practical framework:

If You're Emerging (0–27)

You're early. That's fine. Focus on basic tool exposure: pick one AI assistant, use it daily for a week on real tasks, and pay attention to where it helps and where it doesn't. Don't try to learn everything — build one habit.

If You're Developing (28–59)

You're in the majority. Your growth edge is almost certainly Critical Thinking or Technical Understanding — the two lowest-scoring dimensions in the population. Try this: for your next 10 AI interactions, before accepting the output, write one sentence about why you think the output is or isn't reliable. That single practice builds the evaluation muscle that separates Developing from Proficient.

For a self-diagnostic on specific prompting habits, see How Good Am I at Prompting?.

If You're Proficient (60–79)

You're ahead of roughly two-thirds of assessed people. Your growth path is about depth and breadth: deepening technical understanding, broadening to new use cases, and starting to think about how AI works at the team or organizational level. The AI Coach can help identify specific dimension gaps.

If You're Advanced or Expert (80+)

You're in the top ~9% of the assessed population. At this level, the question shifts from "how do I get better at using AI" to "how do I help others get better" and "how do I shape how my organization adopts AI." The team assessment is designed for this use case.

Why "Average AI Score" Is a Moving Target

One note on durability: the numbers in this post reflect 1,219 completed assessments as of publication. As more people take the assessment and as AI tools evolve, these averages will shift. The tier bands (Emerging, Developing, Proficient, Advanced, Expert) are fixed by design — they represent absolute capability levels, not percentiles. A score of 60 means the same thing whether the average is 48 or 58.

This is a deliberate design choice. If the average rises to 60 over time, that's great — it means the population is getting more capable. It doesn't mean the bar for Proficient should move. The rubric measures what you can do, not how you compare to everyone else.

For context on how AISA's measurement approach compares to alternatives like multiple-choice quizzes and self-assessments, see Beyond Multiple Choice.

---

Related reading: AI Skills Gap: You Overestimate by 18.5 Points — why people predict 63.6 and score 44.8.

Related reading: How Good Are People at AI? 1,103 Tested — the full population-level breakdown.

Related reading: Am I Good at Using AI? 5 Signs [2026] — a practical self-check before you take the assessment.

Frequently Asked Questions

What is a good AI fluency score?

A good AI fluency score starts at 60 out of 100, which places you in the Proficient tier on the AISA scale. This means you can reliably prompt AI tools, evaluate their outputs critically, and integrate AI into real workflows. Across 1,219 assessments, only about a third of people reach this level. A score of 80+ puts you in the Advanced tier, representing roughly the top 9% of assessed individuals.

What is the average AI score?

The average AI score across 1,219 AISA assessments is 48 out of 100, with a median also at 48. This places the typical person in the Developing tier (28–59). Scores vary by role: engineers average 55, product managers 56.8, founders 56.6, and students 37. Most people overestimate their score by nearly 19 points before taking the assessment.

What score do I need to be considered advanced at AI?

You need a composite score of 80 or higher to reach the Advanced tier on the AISA scale. Currently, only the Architect persona (4.4% of assessed individuals, average score 87.6) consistently lands in this range. Reaching Advanced requires strong performance across all five dimensions — Prompting, Critical Thinking, Technical Understanding, Workflow, and Safety — not just depth in one area.