In partnership with
Quick question before you read on. Van Gogh's Sunflowers — the famous wall of yellow, the one on all the mugs and tote bags. How many sunflowers are in it?
Hold your guess. Most people say "a lot," because that's how the painting works — it's an impression of abundance, not a countable bouquet. You feel fifteen-ish without ever counting to fifteen.
Which is exactly what makes it the perfect trap for an AI.
We gave two of the best image models — Google's Gemini 3 Pro and OpenAI's GPT-5 Image — a careful written description of the painting and asked each for a photorealistic version. No reference image. Just words: fifteen sunflowers, a pale-yellow background, a two-tone vase, a tall portrait canvas.
Both came back with something genuinely beautiful — warm, textured, unmistakably Sunflowers. If you glanced at either on a phone, you'd nod and keep scrolling.
Then we started counting.
Why this painting
Van Gogh's National Gallery Sunflowers (Arles, 1888) is the one historians call "yellow on yellow" — yellow flowers, yellow wall, yellow vase, barely three shades of yellow "and nothing else." No contrast to lean on. That's a flex when a human does it and a landmine for an AI, which can't use strong color separation to organize the scene — it has to actually understand the arrangement.
What they nailed
Both images are convincing photographs. Real petal texture, natural light, the earthenware two-tone vase. On mood, palette, and craft, both did beautifully — better than most people expect.
What they missed
Everything you'd want to count. The original has fifteen flowers; ask for fifteen and you get "a bunch" — a plausible mass that reads as roughly right and lands on thirteen, or seventeen, or a number that's hard to pin down. The models paint the idea of fifteen, not fifteen.
And then the one thing a machine can check with zero judgment: the canvas shape. The real painting is a tall portrait, noticeably taller than wide. We asked for exactly that. Gemini got it right — a proper 3:4 portrait. GPT-5 Image returned a perfect square, which isn't a near-miss, it's ignoring the instruction.
Why it matters — and it's not about paintings
A few years ago, AI images were easy to dismiss: six fingers, melted faces. The failure was visible, so it was safe. That era is over. These images are excellent. The failures didn't disappear — they moved, from the surface where you'd notice to the details where you won't unless you look.
Swap "sunflowers" for line items in a contract, figures in a chart, steps in a recipe, or citations in a report, and you have the real reason this matters. The polish is real. The facts underneath still need checking — more now, not less, precisely because nothing on the surface tells you to.
Grade them yourself
On the site there's a live scoring rubric — not a screenshot, an actual tool. Tier A is the machine-checkable canvas shape, already scored. Tier B is countable: you count the flowers. Tier C is the subjective stuff, capped at 20% so taste can't decide the winner. There's a slot to load the real painting (public domain — Van Gogh died in 1890) and score all three side by side.
None of this means the models are useless. Both images would pass for real at a glance — that's exactly the point. The danger was never the obvious garbage; you catch that. It's the output that looks completely right and is wrong in one place you didn't think to check. The next time an AI hands you something polished — a summary, a table, a draft that looks done — treat "it looks right" as the reason to check the facts, not the permission to skip it. The polish is free now. The facts still cost attention.
The live rubric, both AI versions, and the full write-up are on the site.
Try the AI that knows your customers. No commitment.
Most platform evaluations start with a demo request and end three weeks later in a conference room. This one takes 15 minutes and puts you directly inside Gladly's interface — navigating it on your own terms.
See how AI surfaces real-time customer context before a conversation starts. Watch how a single conversation thread pulls in purchase history, channel history, and account details without a handoff.
No installation. No commitment. Start the interactive demo and see the platform for yourself.