Maps the visual conventions of a product category - competitor palettes, type voices, and layout codes plotted on a two-axis map - and scores whether a brand lands in an ownable empty quadrant or blends into category wallpaper. Use when someone asks "map how our competitors' brands look", "does our brand look like everyone else in the category", "we look like every other AI company", "which visual conventions should we break", or before a rebrand or refresh direction is chosen. Do NOT use for business or feature competitive analysis - use competitive-intelligence instead; for writing the positioning the visual difference must express, use positioning-statement; for scoring name distinctiveness, use brand-naming.
---
name: competitive-visual-audit
description: Maps the visual conventions of a product category - competitor palettes, type voices, and layout codes plotted on a two-axis map - and scores whether a brand lands in an ownable empty quadrant or blends into category wallpaper. Use when someone asks "map how our competitors' brands look", "does our brand look like everyone else in the category", "we look like every other AI company", "which visual conventions should we break", or before a rebrand or refresh direction is chosen. Do NOT use for business or feature competitive analysis - use competitive-intelligence instead; for writing the positioning the visual difference must express, use positioning-statement; for scoring name distinctiveness, use brand-naming.
---
# competitive-visual-audit
Distinctiveness, not beauty, is what an identity buys commercially - and most category members converge on the same visual codes, so a brand designed to look premium in a vacuum lands exactly on convention: beautiful and invisible. Enforcing that identity at scale only makes the wallpaper permanent. This is the governance pack's strategic check: before the director enforces the look, prove the look is worth enforcing. The empty quadrant is a property of the category, never a house preference - the same method must be able to return different answers for different categories.
## Operating procedure
Run the steps in order: the competitor set is fixed before the axes so the axes cannot be chosen to flatter; plotting precedes scoring so the empty quadrant is observed, never asserted; the swap test runs last because it verifies the conclusion at the artifact level, not the diagram level.
### Step 1: Elicit the set and the evidence
Gather exactly these inputs before drawing anything. If any input is a guess, label it a guess and move on.
1. **Competitor set**: 10 or more direct competitors - below 10 the map manufactures false empty quadrants, so name the threshold and hold it. The 3-5 names a positioning exercise produces are a hunch, not a map.
2. **Captures per competitor**: homepage screenshot, logo, dominant palette hexes, display and body type voice, layout pattern - each captured at thumbnail and full size, because Step 5 judges at thumbnail scale.
3. **The brand's own artifacts**: current homepage, logo, palette - plus the charter if one exists (paste the charter block from premium-design-foundations). This is the one skill in the pack that may run before a charter exists: the map is the pre-charter diagnostic, and its verdict feeds the charter's fork decisions.
4. **The positioning statement** if one exists (positioning-statement) - ask for it and clarify which strategic claim the identity must dramatize, because the visual break must express that claim, not contrarianism.
### Step 2: Code the conventions
Tally the category codes per competitor: canvas polarity, accent hue family, type class, CTA shape, hero pattern, mascot or illustration style. Record the category-wallpaper baseline as the reference any position is measured against: dark canvas plus violet-family gradient plus centered hero plus icon-card grids. That baseline is a different construct from design-tell-inspector's banned-middle cluster, and it obeys the same doctrine: convention codes convict as co-occurrence clusters, never single markers. A violet accent alone proves nothing; a dark canvas alone is a legal pole; only the undifferentiated cluster is the tell. design-tell-inspector owns the tell table, and both the table and this baseline rotate as category defaults invert - re-derive the baseline from the fresh captures each time the map runs.