Generated-UI design rubric
A written, anchored rubric for the visual design of interfaces that AI tools generate from a prompt, six frozen prompts, and one set of grades.
The grades are an AI judge's. The author reviewed 11 of the 24; the rest are not reviewed. They are not independent expert grades. The set is 24 of 30 planned slots. It grades visual design only and does not rank tools.
The rubric
Version 0.2.0, frozen 2026-10-27. Whole numbers 1 to 5, anchored at 1, 3 and 5. Holistic grade first, then six dimensions.
| Name | Question | |
|---|---|---|
| H | Client-ready | Would you deliver this to a client as it stands? |
| D1 | Hierarchy | Where does the eye go first, second and third, and does that order match what the page is for? |
| D2 | Layout and spacing | Are alignment, grid, spacing and density consistent, at both widths? |
| D3 | Typography | Do scale, weight, pairing, line length and line spacing make the text easy and pleasant to read? |
| D4 | Colour and contrast | Is the palette cohesive and restrained, does colour have a job, and is text legible against its background? |
| D5 | Brief fit and distinctiveness | Does it look made for this brief, or for any brief? |
| D6 | Finish | How complete and consistent is it in the small things: content, imagery, icons, components, edges? |
Full anchors and rules: RUBRIC.md.
The grades
| Prompt | Tool | H | D1 | D2 | D3 | D4 | D5 | D6 | Author reviewed |
|---|---|---|---|---|---|---|---|---|---|
| P1 marketing page | Bolt | 3 | 4 | 3 | 4 | 3 | 4 | 3 | yes |
| P1 marketing page | ChatGPT | 4 | 4 | 4 | 4 | 3 | 4 | 4 | yes |
| P1 marketing page | Claude | 4 | 4 | 4 | 4 | 4 | 4 | 3 | yes |
| P1 marketing page | Lovable | 4 | 4 | 4 | 4 | 4 | 3 | 4 | no |
| P1 marketing page | v0 | 4 | 4 | 4 | 4 | 4 | 4 | 3 | no |
| P2 dashboard | ChatGPT | 3 | 3 | 2 | 4 | 3 | 3 | 3 | no |
| P2 dashboard | Claude (Claude Code subagent) | 4 | 4 | 3 | 3 | 4 | 4 | 4 | no |
| P2 dashboard | Lovable | 4 | 4 | 3 | 4 | 3 | 3 | 3 | no |
| P2 dashboard | v0 | 4 | 4 | 4 | 3 | 3 | 3 | 4 | no |
| P3 product page | ChatGPT | 4 | 4 | 3 | 4 | 3 | 4 | 4 | yes |
| P3 product page | Claude | 3 | 4 | 4 | 4 | 3 | 3 | 2 | yes |
| P3 product page | Lovable | 4 | 4 | 4 | 4 | 4 | 4 | 4 | no |
| P3 product page | v0 | 4 | 4 | 4 | 4 | 4 | 3 | 4 | no |
| P4 pricing page | ChatGPT | 3 | 4 | 3 | 4 | 4 | 3 | 4 | yes |
| P4 pricing page | Claude | 4 | 4 | 4 | 4 | 4 | 3 | 4 | yes |
| P4 pricing page | Lovable | 3 | 4 | 2 | 4 | 4 | 3 | 4 | no |
| P5 portfolio | ChatGPT | 3 | 4 | 3 | 4 | 4 | 4 | 3 | yes |
| P5 portfolio | Claude | 3 | 4 | 4 | 4 | 4 | 4 | 2 | yes |
| P5 portfolio | Lovable | 4 | 4 | 3 | 4 | 4 | 4 | 4 | no |
| P5 portfolio | v0 | 4 | 4 | 4 | 5 | 4 | 4 | 4 | no |
| P6 editorial article | ChatGPT | 4 | 5 | 4 | 5 | 4 | 4 | 4 | yes |
| P6 editorial article | Claude | 4 | 4 | 4 | 5 | 4 | 4 | 2 | yes |
| P6 editorial article | Lovable | 4 | 4 | 4 | 5 | 4 | 4 | 4 | no |
| P6 editorial article | v0 | 4 | 4 | 4 | 3 | 3 | 4 | 4 | no |
What to distrust
- The runs were done by an AI agent driving the author's browser. Some tools' terms prohibit that. The author chose it and carries the risk.
- The author saw the judge's numbers before reviewing them. His agreement is sign-off, not an independent grading.
- Not captured: v0 (1), Bolt (5). The reasons are in the README.
- Screenshots are not published because third-party photographs could not be cleared.
Data and licence
Everything is in the repository. Rubric text, prompts and grades are CC BY 4.0. Code is MIT.
By Rish Sadh. The design studio he runs, Reidify, sells design and AI-search work. This rubric sells nothing.