Fix generic or inaccurate AI output

Overview

AI generation is strongest when the source data and refinement prompt are both clear. If a site feels generic, first improve the Data Tree, then use a targeted refinement request.

Common causes

Key points

• The Data Tree has broad labels like "developer" or "designer" without projects, outcomes, or audience.
• Several sections repeat the same information, so the AI gives the duplicate details too much weight.
• The selected vibe is too broad, such as "modern" without direction.
• The refinement prompt asks for everything at once, which makes the changes harder to control.
• Photos are low resolution, too dark, or not relevant to the story the site is trying to tell.

Better refinement prompts

Key points

• Make the hero sound more senior and less salesy. Emphasize product engineering, developer tools, and founder collaboration.
• Keep the visual style, but rewrite the project section so each card explains the problem, stack, and result.
• Use a calmer editorial layout with more whitespace. Remove hype language and make the copy more direct.
• Replace the contact call to action with a consulting-focused CTA for startups that need launch help.

Refine in small passes

One focused request usually works better than a long list of unrelated changes. Save versions between big changes so you can restore a better draft.

When to use manual editing

Overview

Use manual editing for short text fixes, image crop positions, and tiny polish after the generated structure is right. Use AI refinement for larger changes to layout, tone, sections, and visual direction.