Human-craft audit

Use this audit for Guideline 4.3(b), low-effort concerns, template similarity, or a product that feels generic. The goal is to make the app genuinely more specific, useful, and complete. It is not an AI detector or camouflage guide.

Core position

OFFICIAL: Apple does not publish a rule rejecting apps because AI helped write them. Apple reviews the submitted product under rules covering completeness, minimum functionality, templates, spam, dynamic code, privacy, metadata, and other observable behavior.

OFFICIAL: The June 8, 2026 wording of 4.3(b) rejects new submissions in mature categories unless they provide a meaningfully different or improved experience, and describes some low-effort categories as adding no value.

Audit the symptoms. Do not label an app AI-generated from visual style, code style, or copy.

When only a product description is supplied, complete the audit instead of stopping for files. Grade claims supported by that description conservatively, grade every screen, provenance, accessibility, and product-page dimension without evidence as UNVERIFIED, and make each intervention conditional on the missing evidence it addresses.

Evidence grades

Grade every dimension:

• DISTINCT: specific, coherent evidence makes the product hard to confuse with a generic category entry.
• CREDIBLE: complete and intentional, with some generic surfaces that do not erase the product's value.
• GENERIC: several important surfaces could belong to many apps and do not make the product's specific value legible.
• HIGH RISK: confirmed thinness, duplication, template provenance, or incomplete execution directly supports a 4.2 or 4.3 concern.
• UNVERIFIED: the required build, screens, metadata, or provenance was not supplied.

Grades summarize inspected evidence. They are not approval probabilities.

Apple-design lens

When supplied evidence includes a running build, screenshots, interaction recordings, motion, gestures, materials, typography, feedback, haptics, accessibility settings, or iPad adaptation, read apple-design-review.md completely. Map that evidence into the same five grades. Apple-design craft is not a sixth grade and does not replace product distinction.

Keep three things separate: a confirmed runtime observation, current OFFICIAL Apple guidance, and a Craft recommendation:. A Human Interface Guidelines preference is not automatically an App Review rejection rule. A craft issue becomes review risk only when the observed result makes a feature broken, incomplete, misleading, inaccessible, or difficult to operate.

1. Product distinction

Inspect:

• one-sentence purpose
• specific user and job
• first useful outcome
• core loop and repeat value
• capability that requires domain knowledge
• differentiation from sibling apps and obvious category comparators

DISTINCT evidence:

• The app solves a narrow problem with a workflow or insight that generic competitors do not provide.
• Distinction appears in the first session and does not depend on marketing copy.
• Product depth exists in data handling, interaction, offline behavior, integrations, or domain logic.

GENERIC or HIGH RISK evidence:

• The pitch is only "AI for X", "a better timer", "all-in-one", or another category label.
• The app wraps one API call or website without meaningful mobile value.
• Most navigation leads to static, placeholder, or paywalled screens.
• Differentiation consists only of colors, layout, or a renamed template.

Ask: what can this app demonstrate in three minutes that a thin clone cannot?

2. Binary and asset provenance

Inspect:

• starter kits, commercial templates, generators, and upstream repositories
• inherited module names, namespaces, copyright headers, and package URLs
• icon, illustration, sound, Lottie, and launch assets
• reuse across the developer's portfolio or prior rejected builds
• assistant files or internal notes included in the archive
• framework-common code versus app-specific behavior

Evidence rules:

• Template provenance is a policy issue under 4.2.6 only when the facts match the guideline.
• Exact asset reuse is reportable evidence. It does not prove Apple's internal matching method.
• A framework such as Flutter, React Native, Capacitor, Godot, or Unity is not a defect.
• DOCUMENTED CASE: an icon change was decisive in one published 4.3 recovery. Use it to justify comparison, not an icon-change ritual.

Under-a-day improvements:

• remove shipped internal artifacts
• replace genuinely reused primary assets
• clean inherited names that misstate ownership
• document licenses and transformation
• trim unused dependencies and sample resources

Do not recommend code obfuscation.

3. Visual identity and accessibility

Inspect screenshots or run the app. Do not grade from source filenames alone.

Icon

• recognizable at small size
• distinct silhouette and purpose
• no unreadable generated lettering or irrelevant stock symbol
• works in the supported appearance modes

Typography and hierarchy

• consistent text roles and labels
• Dynamic Type without clipping
• custom typography, if used, serves a deliberate role
• one primary action per state

Color and materials

• semantic palette and sufficient contrast
• translucency and materials preserve hierarchy
• Reduce Transparency is respected where relevant
• state is not communicated by color alone

Motion and haptics

• feedback marks meaningful state changes
• motion explains hierarchy or progress
• direct manipulation responds continuously and preserves the original grab position
• in-flight motion can be interrupted from its current visible state
• momentum-driven releases consider direction or velocity instead of position alone
• reversible surfaces enter and exit along a consistent spatial path
• gesture boundaries remain responsive instead of freezing at a hard stop
• Reduce Motion is respected
• haptics are not attached indiscriminately to every tap

iPad execution

• layout uses the available space intentionally
• popovers, sheets, keyboard, rotation, split views, and pointer interactions behave coherently
• the app does not look like a stretched phone screen unless the product justifies it

Evidence of craft is coherence, not decoration. Native controls can be distinctive when the product's content, hierarchy, and interaction are specific.

4. Microcopy and interaction states

Inspect every primary flow for:

• onboarding
• permission explanation
• empty state
• loading state
• error and recovery
• destructive confirmation
• purchase and trial terms
• account deletion
• success feedback

Flag:

• "Welcome to MyApp"
• "No data"
• "Something went wrong. Please try again."
• "Unlock the power of"
• "Manage your X"
• mixed tone from screen to screen
• permissions that repeat the system prompt without explaining the feature
• error messages with no next action

Rewrite formula:

1. name the object or problem
2. state the consequence or benefit
3. give the next action

Examples:

Specific copy must describe a feature that exists.

5. App Store product page

Inspect:

• name and subtitle
• first description sentence
• audience and core outcome
• screenshot sequence and captions
• keywords and categories
• privacy, support, and terms pages
• What's New history

DISTINCT evidence:

• The first sentence names a specific outcome and audience.
• Screenshots show real flows in a clear order.
• Each caption states a concrete benefit visible in the frame.
• Claims are supported by the build.
• The support page has a working contact and product-specific help.

GENERIC or HIGH RISK evidence:

• The copy could describe hundreds of apps.
• The name is a keyword chain.
• Screenshots are mostly marketing art or show UI absent from the build.
• Description and screenshots lead with a saturated category rather than distinct value.
• Repeated What's New entries say only "bug fixes and improvements."

Output

Mode C: Human-craft audit
Product distinction: <GRADE>
Provenance: <GRADE>
Visual identity and accessibility: <GRADE>
Microcopy and states: <GRADE>
Product page: <GRADE>

# Human-craft audit: <app>

## Why these grades
<path, screen, metadata, or supplied evidence for each>

## Top five interventions
1. <intervention>
   Evidence: <location>
   Why it matters: <specific 4.2, 4.3, completeness, or reviewer-legibility reason>
   Effort: <S | M | L>
   Verify: <observable outcome>

## Unverified surfaces
<screens, archive, metadata, provenance, or comparisons still needed>

If product distinction is HIGH RISK, lead with the product change required. Do not bury it beneath icon, color, copy, or animation suggestions.