How to make AI event posters look less generic

AI-generated event posters do not have to resemble the same pastel craft-fair template. The key is to give image generators a clear visual direction instead of accepting their first result. John Hartnup writes for ERE I AM – JH! that repeated default designs, rather than outright poor quality, are what make many AI posters feel tiresome.

Hartnup tested this idea with an invented spring fayre. His initial prompt asked for a clean, bright layout and explicitly excluded pastel, airbrushed, and oil-painting-style imagery. Yet the result still resembled the familiar local-event aesthetic he wanted to avoid.

A follow-up request for a completely different design direction produced a more distinctive geometric poster. ChatGPT described the result as Bauhaus-inspired modernism, combining simple shapes, strong typography, limited colors, and an information-led layout.

Ask for an aesthetic, not just a poster

The experiment suggests that content teams should name a visual reference or design system in the prompt. Hartnup’s examples include:

• Risograph printing, with limited colors and imperfect ink layers
• Japanese minimalism, with empty space and a single strong graphic
• Wayfinding and signage, using icons, arrows, and structured information
• Memphis design, with bright colors and playful postmodern shapes
• Photocopied 1980s punk fanzines or 1990s drum and bass flyers

Specific instructions also help define what to exclude. A prompt can request a bold graphic, real typographic hierarchy, and a restricted color palette while ruling out bunting, hand-drawn flowers, generic crowds, or “craft fair” imagery.

Hartnup notes that AI-made work may still be recognizable as AI. But distinctiveness can matter more than hiding the tool. He also advises starting with the desired style rather than repeatedly modifying an earlier image, since unwanted text and visual details can carry over through the chat context. For work that needs later changes, he points to tools that can produce editable layouts with real text and layers rather than a single flattened image.