# Visual review rubric

Review the rendered page, slide, image or PDF. Passing the build establishes useful technical facts; the review below checks what a reader receives. Answer each question with evidence from the actual output, and fix a failed check before handoff.

| Check | Question |
|---|---|
| Purpose | Within a few seconds, can someone identify the subject and the main idea or next action? |
| Hierarchy | Is the first read obvious, followed by a clear supporting element? Do headings, labels and body copy have distinct roles? |
| Register | Do the human/digital choices match the meaning? Would labels and structure still explain the relationship without color? |
| Composition | Do alignment, spacing and changes of pace guide reading? Does the structure suit the content rather than repeat a generic card grid? |
| Density | At the intended viewing size, is every label and qualification readable? Was copy edited or content split before type was shrunk? |
| Evidence | Are methods, sources, uncertainty and limitations near the claims they qualify? Are every sample figure and illustrative scenario visibly identified? |
| Contrast | Do text and controls meet the specification in both themes, including on tinted surfaces, gradients and interaction states? |
| Access | Can someone follow the reading order and operate every control with a keyboard? Are focus, labels, touch targets and diagram equivalents present? |
| Adaptation | Does the page reflow at 360px and with enlarged text? Does reduced motion preserve the complete information and usable state? |
| Export | Are margins, type, images and page breaks intact in the actual exported file, with no clipping and with sample labels still visible? |

For a page, check both themes at a phone width and a large screen. For fixed artwork, review its intended size and the smaller size at which it will be encountered. For a deck, inspect every canvas and the exported pages, not just the cover. The 30-word default applies to calm spoken slides; an evidence canvas can be denser if its hierarchy, qualifications and readable type hold up.

If a check does not apply, record why. For example, a static PNG has no keyboard controls, but still needs an accompanying text description when shared on a platform that supports one. A print-only piece needs a print contrast/readability review rather than a dark-mode variant.

At handoff, name the composition choice, what was checked, and any sample content still needing replacement. Distinguish automated checks from visual judgment; neither substitutes for the other.