You know that feeling when you generate an infographic and it looks exactly like every other AI image out there? Same tired gradients. Same mushy shapes. Same soulless vibe. I kept running into this myself, and I figured better prompts were the answer. Turns out I had it backwards.
I just came across a brilliant breakdown from an AI professional who nailed why our AI designs keep looking generic. The core insight from the original poster hit me hard: AI doesn’t have taste, so telling it to “make it look good” is useless. It has no idea what good means to you. The fix takes about 10 minutes, and it flips the whole approach on its head.
The problem nobody wants to admit
The creator pointed out something we’ve all seen. LinkedIn feeds are packed with designs that basically scream “made by AI without looking at the output.” And here’s the part that stuck with me. People blame the AI when the real problem is they never showed it what good actually looks like.
Think about it like hiring a designer who has never seen a single design you admire. You’d never expect magic from that. Yet that’s exactly what we do when we type “make a clean modern infographic” and hope for the best.
The 10-minute fix, step by step
Here’s the exact process this savvy professional laid out. Each step has a reason behind it, so follow the logic and not just the actions.
- Stop describing the style in words, because AI has no taste and can’t translate vague adjectives into real visual choices.
- Head to Pinterest and search ONE single word, keeping it simple so the results stay broad and inspiring.
- Click the image you love, then keep clicking through related results to build a strong visual trail.
- Upload 2 or 3 of those favorites to Claude, using the Home tab with the Opus model for the sharpest analysis.
- Ask it to extract the design system FIRST, before you generate anything at all.
- Have it pull out the palette, the typography, the spacing, and the one signature element that makes the look distinct.
- Look at that preview carefully before you make a single thing, so you catch the direction early.
- Only then generate the actual infographic, now that the AI knows what good looks like to you.
- For web design, swap Pinterest for Awwwards, Godly, or Refero to pull reference from the best sites around.
Why this order matters so much
The genius here is the sequence. Most people jump straight to generation and cross their fingers. The expert flips it: reference first, extraction second, creation last. By making the AI describe the design system before it builds anything, you turn a guessing game into a repeatable process.
The real taste is one screenshot away. When you show the AI concrete examples of great work, you stop relying on its imagination and start feeding it your standards.
Where to find great references
The person who shared this gave us a solid starter kit of sources. Here’s a quick rundown of what to use and when:
- Pinterest for infographics, posters, and general visual inspiration.
- Awwwards for award winning website design.
- Godly for curated, high end web layouts.
- Refero for real product and interface references.
Save a small folder of screenshots you keep coming back to. Over time you build your own reference library, and every project starts faster because you already know what good looks like to you.
My take on why this is huge
I was genuinely impressed when I saw how simple this shift is. It costs nothing. It takes minutes. And it works with tools most of us already have open. The mind behind this reframed the whole problem: it was never the model letting us down, it was us never showing it a target.
This connects to a bigger trend I keep noticing. As AI tools get more powerful, the bottleneck stops being the tech and becomes the taste and direction we bring to it. The people who win aren’t the ones with secret prompts. They’re the ones who feed the AI great input and know how to spot great output.
So next time your design comes out flat, don’t rewrite the prompt for the tenth time. Go grab a screenshot of something you love and hand it over. Forget “make it look good.” Show it what good looks like.
The original poster shared a full walkthrough with even more detail on this method. Check out the complete LinkedIn post to see the exact examples and go deeper.