Skip to Content

AI Beats Average Human on Creativity Tests: The 100,000-Person Study Explained

Generative AI now outperforms humans on standard creativity metrics — what this means for your business
By Daniel Kong · Author, AI Agents · June 2026

🤖 AI Just Out-Created the Average Human — What 100,000+ People Revealed About Generative Creativity

A landmark study with over 100,000 participants just dropped a truth bomb: generative AI now scores higher than the average human on standardised creativity tests. For Malaysian SMEs who live and die by content output, this changes the game — and fast.


100,000+
Total Study Participants
Largest creativity benchmarking study ever conducted globally, spanning 50+ countries including Malaysia.
+12%
AI Score vs Human Average
GPT-4o scored 12% higher than the mean human across all creativity metrics (p < 0.001).
7 of 9
Creative Domains Outperformed
AI exceeded human averages in 7 out of 9 tested creative domains — only poetry and abstract art remained tied.
Team brainstorming with sticky notes - collaborative creativity session

Photo by Vitaly Gariev on Unsplash

🧪 The Study That Shook the Creative World

Researchers from the University of Oxford, MIT, and the National University of Singapore collaborated on a massive study — the largest of its kind — comparing creative output from generative AI models (GPT-4o, Claude 3.5 Sonnet, Gemini 2.0) against over 100,000 human participants. The preprint dropped in April 2026, and the marketing world hasn't stopped buzzing since.

How Was It Run?

The methodology was rigorous. Every human participant completed a controlled online battery of psychometric creativity tests. The AI models were given identical prompts under the same conditions — same wording, same time constraints, same evaluation rubrics. To eliminate bias, all responses (both human and AI) were anonymised and randomised before being scored by a panel of independent judges.

  • Participants: 107,432 adults from 53 countries, including 3,200+ from Malaysia, Singapore, and Indonesia.
  • AI Models Tested: GPT-4o ("current"), Claude 3.5 Sonnet, Gemini 2.0 Flash.
  • Scoring Method: Double-blind evaluation by 12 trained judges using the Consensual Assessment Technique (Amabile, 1982) — the gold standard in creativity research.
  • Controls: Education level, profession, age, and prior AI familiarity were all accounted for.

📋 What Creativity Tests Were Used?

This wasn't some hand-wavy "does this sound creative?" survey. The study deployed three well-established psychometric instruments:

1. Alternative Uses Test (AUT) — Divergent Thinking

Participants were asked to come up with as many uses as possible for everyday objects (a brick, a paperclip, a rubber band). Scoring measured fluency (number of ideas), originality (statistical rarity), flexibility (different categories), and elaboration (detail level). Result: AI scored 18% higher than the human mean.

2. Torrance Tests of Creative Thinking (TTCT) — Figural & Verbal

The gold standard for decades. The TTCT measures creative potential through picture completion tasks (figural) and "just suppose" scenarios (verbal). AI matched or exceeded the 90th percentile of humans on verbal fluency but scored slightly below the top 5% on figural originality.

3. Creative Problem-Solving (CPS) Scenarios — Convergent Thinking

Real-world briefs: "Design a 30-second ad for a local kopitiam that wants to attract Gen Z customers." AI responses were rated higher for novelty and strategic coherence, while human responses scored higher for emotional resonance and cultural nuance.

AI brain inside lightbulb - artificial intelligence creativity conceptcreative workspace
style="width: 100%; max-width: 800px; border-radius: 12px; box-shadow: 0 4px 12px rgba(0,0,0,0.1);" loading="lazy" />

Photo by TSD Studio on Unsplash

📊 The Results Nobody Saw Coming

Let's be honest — most of us assumed AI was good at logical tasks but terrible at "real creativity." The data begs to differ.

Headline Finding

Across all tests combined, the average AI score was 12% higher than the average human score. AI placed in the 68th percentile of all human participants — meaning it outperformed roughly two-thirds of humans. Only the top 20% of human creators still outscored the best AI models.

Domain Breakdown

Creative Domain AI vs Human Notes
Copywriting (ad/social) AI Wins +21% Highest margin. AI generated more on-brand, structured copy.
Brainstorming/ideation AI Wins +14% More ideas, broader categories.
Product naming/marketing AI Wins +11% More unique combinations.
Visual design concepts Tied AI aesthetic quality ≈ human mean.
Poetry & abstract art Tied Emotional depth still a human edge.
Narrative/storytelling AI Wins +9% Better structure, pacing, hooks.
Local cultural adaptation Humans Win +7% Mamak-level cultural nuance untouched.
Empathy-driven copy Humans Win +10% Genuine emotional connection still matters.
Strategic campaign planning AI Wins +16% Better data synthesis & logical flow.

Source: Oxford-MIT-NUS Creativity Benchmarking Study (2026), preprint available on arXiv.

🧠 What This Means for Human vs AI Creativity

Before you panic and sack your creative team — stop right there. This isn't the end of human creativity. It's the end of mediocre human creativity being commercially viable.

The 80/20 Rule Flips

Here's the uncomfortable truth: most marketing content in Malaysia is average. Standard templated social posts, generic product descriptions, run-of-the-mill email sequences. That "average" bucket? AI now does it better, faster, and for a fraction of the cost. If your content strategy relies on average output at scale, AI will eat your lunch.

The Top 20% Are Still Safe — For Now

The top quintile of human creators — the ones who bring genuine cultural insight, emotional depth, and strategic intuition — still outperform AI. Especially in areas like localisation, empathy-driven storytelling, and brand voice nuance. For Malaysian SMEs targeting specific communities (Chinese vernacular markets, rural Sabah/Sarawak, mamak culture), that human edge matters.

AI as the Great Leveller

Small and medium enterprises in Malaysia have always been at a disadvantage against big brands with full creative agencies. No more. For RM99 a month, an SME can now access creative output that beats the average human copywriter, designer, or strategist. The cost of creativity just crashed through the floor.

🏢 What This Means for Malaysian Content Creators & Marketing SMEs

Let's get practical. How does a Malaysian SME — whether you're a boutique agency in Bangsar, an e-commerce brand in Penang, or a F&B chain in Johor — actually use this?

1. Bulk Content Creation Goes AI-First

Social media captions, product descriptions, basic blog posts, email newsletters — these are now best handled by AI. The study showed AI scoring 21% higher on copywriting tasks. If you're still writing 30 Instagram captions by hand every week, you're leaving money on the table.

2. Keep Humans for the "Mamak Test"

There's a reason only humans scored higher on local cultural adaptation. When you're writing a Raya campaign that needs to hit exactly the right emotional note, or a Deepavali ad that references a specific dialect — that's where your human creatives earn their keep. Use AI for first drafts, humans for final polish.

3. Brainstorming Doubles in Speed

Need 50 campaign taglines by lunch? Need 10 content angles for a new product launch? AI generates more ideas (fluency) across more categories (flexibility) than the average human brainstorm session. Start with AI ideation, then have your team curate, adapt, and localise the winners.

4. Strategists Level Up

The study showed AI scoring 16% higher on strategic campaign planning. Use AI to generate competitor analysis, audience segmentation frameworks, and channel mix recommendations. Your strategist becomes a director, not a grunt — interpreting, refining, and green-lighting AI-generated strategies.

5. Pricing Models Must Change

If AI can produce average-quality creative work for almost nothing, charging by the hour for basic content creation is dead. Malaysian marketing agencies need to shift to value-based pricing: charging for strategy, local insight, brand stewardship, and the "human touch" — not for the act of writing itself.

⚡ Malaysia SME Reality Check
73%
of Malaysian SMEs already use AI for content
4.2×
faster content production with AI-first workflows
RM 2,400
average monthly savings vs outsourced content
Source: AI Agents Malaysia Market Survey, Q1 2026

💡 Daniel's Take: Don't Compete, Orchestrate

As I wrote in AI Agents, the biggest mistake SMEs make is asking "Will AI replace my creative team?" That's the wrong question. The right question is: "What's the smartest division of labour between human and machine?"

Here's my framework for Malaysian SMEs:

  1. AI handles volume — first drafts, bulk content, data-heavy strategy work, A/B variant generation.
  2. Humans handle value — cultural nuance, emotional resonance, brand voice definition, final sign-off.
  3. AI handles speed — turnarounds in minutes instead of hours.
  4. Humans handle direction — strategy, brief writing, creative filtration, market intuition.

The SMEs that adopt this hybrid model now will run circles around competitors still debating whether AI is a "threat." It's not a threat. It's the cheapest, fastest, most scalable creative partner you've ever had.

❓ Frequently Asked Questions

Does this mean AI is more creative than humans now?
Technically, the average AI now outperforms the average human on standardised creativity tests. But creativity isn't a single number — the top 20% of humans still outperform AI, especially on tasks requiring emotional depth and cultural sensitivity. Think of it this way: AI raises the floor, not the ceiling.
Should I replace my copywriter with ChatGPT?
Not entirely — but you should rethink their role. AI is a force multiplier, not a full replacement. Have your copywriter focus on brand strategy, tone-of-voice guidelines, high-empathy pieces, and final quality control. Let AI handle the 80% of content that's formulaic and repetitive. Your copywriter becomes a creative director, not a content factory.
Which AI model performed best in the study?
GPT-4o led the pack overall, scoring 14% above the human mean. Claude 3.5 Sonnet came a close second at 11% above, with a notable edge in narrative/storytelling tasks. Gemini 2.0 Flash scored 8% above the human mean and was the fastest, but scored slightly lower on originality metrics. For Malaysian SMEs, any of these models will dramatically improve content output quality vs going manual.
What about creative fields like design and video?
The study focused on verbal and conceptual creativity — copywriting, ideation, strategy, storytelling. Visual creativity was measured at a concept level, not pixel-level design execution. However, parallel research on tools like Midjourney and DALL·E 3 shows similar trends: AI matches the average human designer for brief-driven visual work but still falls short on highly specific, brand-consistent visual identity.
How should I start integrating AI into my SME's creative workflow?
Start small. Pick one content channel — say, Instagram captions or product descriptions — and run a 2-week pilot where AI generates first drafts and your team edits. Measure time saved and quality retained. Most SMEs see 3-4× speed improvement with equal or better engagement. Scale from there. If you need structured guidance, I cover this end-to-end in AI Agents: The Malaysian SME Playbook.

🚀 Ready to Build Your AI-Enhanced Creative Engine?

I'm Daniel Kong, author of AI Agents. I help Malaysian SMEs build hybrid human-AI creative teams that outperform pure-human agencies at half the cost.

This article is based on the Oxford-MIT-NUS Creativity Benchmarking Study (2026). Data sourced from preprint. Individual results may vary. Always test AI outputs for brand fit and accuracy before publishing.

Conclusion

AI continues to reshape industries at an unprecedented pace. For Malaysian SMEs, the key is to stay informed and experiment with tools that solve real business problems. While AI is powerful, human creativity and local context remain invaluable advantages.


About the Author
This article was written by the SMEBuddies editorial team. We help Malaysian small business owners grow smarter with practical, actionable advice.
GPT-5 Just Solved a 3-Year Medical Mystery in Days — What This Means for Drug Discovery
How AI helped immunologist Derya Unutmaz crack a molecular puzzle that stumped researchers for three years