In a recent LinkedIn post, Ruben Hassid discusses a common pitfall in AI-generated visuals: the tendency for them to look generic and indistinguishable. Hassid argues that the issue often lies not with the AI itself, but with how users prompt and guide the creative process. He offers a practical, step-by-step approach to help users move beyond bland outputs and achieve more distinctive designs.
Hassid begins by addressing the perceived limitations of AI, stating:
Your infographic looks like every other AI image. Same gradient & mush. But you can fix it in 10 mins:
He suggests that the first mistake is often describing a style to the AI, as Hassid believes AI lacks inherent taste. Instead, he advocates for a visual-first approach. According to Hassid, users should leverage platforms like Pinterest to find visual inspiration. The process he outlines involves searching for a single word, identifying an image that resonates, and then using that image as a reference point.
Leveraging Visual References for AI Design
Hassid’s methodology centers on providing the AI with concrete visual examples rather than abstract stylistic descriptions. He details a method using Claude’s Opus model, where users upload 2-3 inspiring images.
The core of his technique, as Hassid explains, is to instruct the AI to extract the ‘design system’ from these examples. This involves identifying key elements such as:
- Color palettes
- Typography choices
- Spacing and layout principles
- Unique signature elements
Hassid emphasizes the importance of reviewing the AI’s preview before final generation. This iterative step allows for adjustments based on the extracted design system, ensuring the output aligns with the desired aesthetic.
Addressing the “AI Look” in Design
Hassid laments the prevalence of uninspired AI-generated content on professional networks, noting:
My LinkedIn feed is filled with designs that scream “made by AI without looking at the output”
He contends that the blame is frequently misplaced, with users pointing fingers at the AI when the root cause is a lack of specific guidance. Hassid asserts that the solution is straightforward: users must actively show the AI what constitutes ‘good’ design through carefully selected examples and targeted prompts.
For those in web design, Hassid also suggests alternative sources of inspiration like Awwwards, Godly, and Refero, indicating that the principle of seeking and extracting design systems applies broadly across digital media.
The Power of “Show, Don’t Tell” in AI Prompting
Ruben Hassid’s advice boils down to a fundamental principle: effective AI prompting requires demonstrating desired outcomes rather than merely describing them. He encapsulates this in his concluding thought:
Forget “make it look good.” Show it what good looks like.
Hassid offers a free, comprehensive guide detailing his exact design-system prompt, screenshots, and comparisons between models for those seeking a deeper dive into his techniques.
📝 About This Content
This article is based on insights shared by Ruben Hassid on LinkedIn.
📅 Originally posted on August 5, 2026 | View original post on LinkedIn →