AI room design quality benchmark
A transparent 20-point test for AI room designs.
Published September 1, 2026. This rubric explains exactly how Interizo reviews a room concept—and which mistakes cause an automatic rejection regardless of how polished the image looks.
The score: 20 points across five checks
Each concept is reviewed beside its source photograph at the original viewpoint. Fixed architecture receives up to five points, circulation up to five, furniture scale and function up to five, design coherence up to three, and practical actionability up to two.
- Architecture and viewpoint: 0–5
- Circulation and door clearance: 0–5
- Scale and room function: 0–5
- Coherence: 0–3
- Actionability: 0–2
Automatic failures come before the total
A high visual score cannot rescue a blocked entrance, missing window, invented room boundary or materially changed camera position. Those failures make the concept unreliable as a representation of the source room and trigger rejection before the total is considered.
- Furniture blocks a doorway or required route
- A fixed opening disappears or moves
- The viewpoint becomes a different room
- The room loses its intended function
How a review is recorded
The reviewer keeps the source and concept visible together, records one score per category and writes a short reason for every deduction. A concept must score at least 16 out of 20 with no automatic failure to pass the visual benchmark.
This release publishes the scoring method, not invented aggregate results. Sample-level performance will only be published after a sufficiently large, repeatable review has been completed.
- Paired source-and-concept review
- Written reason for every deduction
- Pass threshold: 16/20
- No claim of measured or professional certification
What the benchmark cannot certify
This is a visual review. It cannot verify centimetre-level dimensions, structural feasibility, accessibility, fire safety, electrical work, waterproofing, live prices or whether a generated product exists. Those decisions still require measurements, retailer verification and qualified advice where appropriate.
- Visual plausibility, not construction approval
- Generated dimensions are estimates
- Retail facts require source checks
- Consequential work needs a qualified person
From photograph to room design
- Upload a clear photo of the room.
- Choose the room type and style.
- Download the completed design from your account.
AI room design quality benchmark FAQ
Why use a 20-point benchmark?
A fixed score makes the review criteria visible and repeatable instead of relying on whether an image simply feels impressive.
Does 16 out of 20 guarantee that furniture fits?
No. Passing means the concept is visually plausible under this rubric. Every real dimension and clearance still needs measurement.
Will Interizo publish actual test results?
Yes, but only after a sufficiently large sample has been reviewed with the same method. This page deliberately does not invent aggregate performance.
Can another AI room tool use this rubric?
Yes. The criteria are intentionally tool-agnostic so anyone can compare a source photograph and generated concept more critically.