Because the internet is flooded with professional product shots of watches set to 10:10, the AI has “learned” that “Clock = 10:10.” To the AI, a clock showing 4:30 looks “wrong” or statistically improbable compared to the millions of 10:10 examples it has digested in its training phase.
Franco Folini
Apparently, AI tools don’t know how to draw an analog clock! I’ve asked several AI image generators, such as ChatGPT, Gemini, Canva, Midjourney, and Photoshop (Firefly), to draw a clock showing a specific time, but all seem to default to 10:10.
The “Aesthetic Bias” in Big Data
Why does this happen? It’s a classic case of data contamination. AI image generators don’t “understand” the mechanics of gears or the concept of passing time; they are statistical engines that predict what an image should look like based on their training sets.
In the world of commercial photography and watchmaking, 10:10 is the “golden standard” for several reasons:
- Framing the Brand: The hands at 10 and 2 perfectly frame the manufacturer’s logo (usually located at 12 o’clock).
- The Smile Effect: Psychologically, 10:10 resembles a “smiley face,” making the product feel more approachable and positive.
- Visual Balance: It creates a symmetrical “V” shape that is more aesthetically pleasing than, say, 6:30.
Because the internet is flooded with professional product shots of watches set to 10:10, the AI has “learned” that “Clock = 10:10.” To the AI, a clock showing 4:30 looks “wrong” or statistically improbable compared to the millions of 10:10 examples it has digested.
Tokenization and the “Hands” Problem
Beyond the aesthetic bias, there is a technical limitation at play. AI models often struggle with spatial reasoning and fine motor details (this is the same reason AI famously struggles to draw five human fingers).
An analog clock requires precise geometry: the relationship between the hour hand and the minute hand is mathematical. Current Diffusion Models (the tech behind Midjourney and DALL-E) are great at textures and vibes, but they are notoriously poor at “counting” or understanding specific angular degrees. When you ask for 4:30, the AI tries to blend the concept of “Clock” with your request, but the gravitational pull of the 10:10 training data is simply too strong to overcome.
What This Tells Us About AI in Marketing
As a B2B marketing consultant, I see this experiment as a vital reminder: AI is a mirror, not a mind.
If we rely solely on AI for creative assets, we risk falling into a “sea of monotony.” If every AI-generated clock shows 10:10, and every AI-generated “professional” looks the same, our marketing becomes a reflection of averages rather than a beacon of original thought.
AI is excellent for generating drafts, but it still lacks the “common sense” to know that a clock is a functional tool, not just a decorative circle. For now, if you need a specific time on a clock for your campaign, you’ll still need a human designer or a very steady hand in Photoshop.
If we rely on AI for creative assets, we risk falling into a “sea of monotony.” If every AI-generated clock shows 10:10, and every AI-generated “professional” looks the same, our marketing becomes a boring reflection of averages rather than a beacon of original thought.
Franco Folini
