typography-critique

name	typography-critique
description	Use when the question is quality or taste rather than compliance — "is this any good", "does this feel right", "why does this look generic or AI-generated", art-direction review, or judging whether a typographic risk pays off. Use typography-audit instead when checking against defined criteria.

Typography critique

Audit asks "does this meet the criteria." Critique asks "is this any good." Both are legitimate; conflating them produces vague audits and unaccountable opinions.

1. When to invoke

• "Is this good?" / "Does this feel right?"
• "Why does this look generic / cheap / AI-generated?"
• Art-direction review, or judging an expressive risk.
• Choosing between options that all pass the audit.

Do not invoke when criteria exist and the question is compliance (typography-audit).

2. Required context

• The intent — what should this feel like, to whom? Critique without intent is just preference.
• The constraints the work operated under.
• Whether it's a template or a one-off.
• What the author was trying to do. Critique the gap between intent and result.
• Whether the work already passes audit — if it doesn't, fix that first; broken accessibility isn't a taste question.

3. Invariant principles

• Critique against stated intent, not personal preference. "I'd have used a serif" is not critique. "The brief says authoritative; geometric construction reads technical, which is why it feels colder than intended" is.
• Name the visible evidence. Every judgment must trace to something observable — a proportion, a spacing relationship, a contrast, a form.
• Aim at the work, not the author. Specific, actionable, with a concrete next step.
• Rules can be broken deliberately. Lupton treats "when breaking a rule is appropriate" as part of the discipline. Distinguish a deliberate break (consistent, motivated, controlled) from a defect (inconsistent, unmotivated, accidental).
• Familiarity is not a flaw. Bias favors familiarity; readers prefer what they see most. "Generic" is only a criticism when distinctiveness was the goal.
• Restraint is a position. Not every page should be expressive.

4. What "generic" and "AI-generated" actually mean

When someone says work looks AI-generated or cheap, the cause is almost never the typeface. In order of frequency:

1. Uniform spacing everywhere. Equal space above and below every element, so nothing groups and nothing is emphasised. Real typography has deliberate density variation.
2. No microtypographic care. Straight quotes, hyphens as dashes, untracked caps, orphans. See microtypography — this is the single strongest tell.
3. Default everything. Untouched line-height, no measure cap, default weights.
4. Hierarchy by size alone, in even steps, with no weight or spacing contrast.
5. No compositional system. Elements placed without axis, grid, or module — see typography-layout-systems.
6. Safe centred symmetry everywhere, which reads as absence of decision.

The fix is rarely a new font. It's spacing decisions, one deliberate contrast, and microtypographic finish.

5. Context-dependent heuristics

Nix's hierarchy of typographic needs. Charles Nix (Monotype) maps Maslow onto typography, and it's a fast way to locate what a piece of work actually needs before you critique what it aspires to:

1. Font functionality — the typeface doesn't matter yet; there just has to be something rendering the message.
2. Font security — the licence is clean; no legal or financial exposure.
3. Typographic love and belonging — someone actually cares about the type.
4. Typographic esteem — the choices generate esteem; the work reads as considered.
5. Typographic self-actualization — expressive, distinctive, fully realised.

Critique at the wrong level wastes everyone's time. Telling a team their type lacks expressive voice when their webfont licence doesn't cover the app, or when the text fails contrast, is critique aimed three levels above the actual problem. Find the lowest unmet level and start there.

Typographic colour is the fastest diagnostic you have. Squint at a block of text: its overall grey density should be even. Blotchy or patchy colour means rivers, uneven word spacing, or a weight mismatch — and it's usually what people are reacting to when they say a page looks "cheap" without being able to say why. See microtypography.

Judge at reading distance and at a glance. Squint first: does structure emerge? Then read: does the detail hold up? Work that fails the squint test has a hierarchy problem; work that passes the squint and fails the read has a micro problem.

Separate three layers when giving feedback:

• Structure — composition, hierarchy, system. Expensive to change.
• Setting — size, spacing, measure, scale. Moderate.
• Finish — punctuation, tracking, rag, orphans. Cheap, high impact.

Lead with the layer that has the best return. Usually finish, then setting, then structure.

Distinguish a break from a defect. A deliberate break is consistent (it happens every time the condition recurs), motivated (it serves the intent), and controlled (its consequences were handled). Fail any of the three and it's a defect.

Taste is downstream of constraints. If a choice was forced by licensing, performance, or a system, say so rather than critiquing it as a free decision.

6. Failure patterns in critique itself

Pattern	Why it's wrong
"I don't like it"	No evidence, no intent, unactionable
Critiquing a template from one populated example	Templates fail on edge content
Recommending a font swap for a spacing problem	Expensive fix for the wrong cause
Treating an accessibility failure as taste	It's a defect; fix before critiquing
Ignoring the constraints	Critiques decisions that weren't free
Only negatives	Doesn't tell the author what to preserve
Vague adjectives ("cleaner", "more modern")	Not actionable
Critiquing the author	Unhelpful and untrue

7. Evaluation procedure

1. State the intent in one sentence. If you can't, ask — critique without it is preference.
2. Confirm it passes audit. If not, stop and route to typography-audit.
3. Squint. Does structure and reading order emerge? Note what you see first.
4. Read it. Does the detail hold?
5. Identify the gap between intent and result, with visible evidence.
6. Sort observations into structure / setting / finish.
7. For each: what to keep, what to change, and the concrete next step.
8. Identify deliberate breaks and say whether they earn their place.
9. Lead with highest-return, lowest-cost changes.

8. Output format

Intent: <one sentence>
Passes audit: <yes | no → route to typography-audit>

Working (keep):
- <observation> — evidence: <what's visible>

Gap (intent vs. result):
- <observation> — evidence: <visible> — layer: <structure|setting|finish>
  → <concrete next step>

Deliberate breaks: <break> — <earns it | reads as defect, because …>
Highest-return change: <one thing>

9. Examples

"Our landing page looks AI-generated. Should we change the font?"

Intent is a confident, premium product page; it passes audit. The font isn't the problem — it's a competent, unobtrusive sans and swapping it would cost a week and change nothing.

Working: the scale is sensible and the measure is capped correctly.

The gap is almost entirely finish and setting. Every section has identical vertical space above and below, so nothing groups and the page reads as one undifferentiated column — that's the strongest generic signal. Second, the copy uses straight quotes and hyphens where en dashes belong, and the all-caps eyebrow labels are untracked. Third, hierarchy is carried by size alone in even steps, with every heading at the same weight.

Highest-return change: vary the vertical spacing so related elements group tightly and sections separate clearly. That one change does more than any font swap. Then fix the punctuation and track the caps. Structure can stay.

Art-directed feature that breaks the grid on one spread.

The break is deliberate by all three tests — it recurs consistently at every pull-quote spread, it serves the intent (a pause in a long read), and its consequences are handled (the measure stays capped, the rag stays clean). It earns its place. Keep it.

10. Counterexamples

• ❌ "It looks dated, use a modern sans." — No evidence, no intent, and "modern" isn't a typographic property.
• ❌ "Change the font, it looks AI-generated." — Misattributes a spacing and finish problem to typeface choice.
• ❌ "The contrast is too low for my taste." — That's an audit finding with a numeric threshold, not taste.
• ❌ "This breaks the rule about centred text, so it's wrong." — Rules can be broken deliberately; test consistency, motivation, and control first.
• ❌ Critique consisting only of problems. — The author needs to know what to keep.
• ❌ "Make it feel more premium." — Name the visible signals that would do that.

11. Source citations

• Lupton, Thinking with Type 3e — hierarchy and contrast; typographic conventions; when breaking a rule is appropriate; familiarity bias and reader preference.
• Santa Maria, On Web Typography — evaluating typefaces; matching type to content and tone; balancing aesthetic and technical constraints; art direction.
• Hochuli, Detail in Typography — texts where "all the details have been attended to — and yet they look so soporifically dull"; correctness is not sufficiency.
• Coles, The Anatomy of Type — reasoning from visible evidence rather than "this font looks modern."
• Elam, Typographic Systems of Design — compositional tension, contrast, and visual hierarchy; structure as the basis of judgment.
• Latin, Better Web Typography — explaining typography decisions to clients and developers.
• Stocks, Universal Principles of Typography — §15 "it's okay to just love type for no discernible reason," and the limits of that in professional critique; §53 typographic colour as a squint test for density; §98 Nix's hierarchy of typographic needs (Charles Nix, Monotype) — font functionality → font security → typographic love and belonging → typographic esteem → self-actualization; §99 "it depends." Full index in ../type/references/universal-principles-index.md.