../

Branding

How to prompt a chat assistant (ChatGPT, Claude, Gemini) and image generators (Midjourney, Flux, Ideogram, Adobe Firefly) for brand strategy, naming, voice and visual identity when you are not a brand designer. Read fundamentals first for the general method; campaigns and copy are in marketing, print layouts in posters and print, screen design in UI/UX, and the science of palettes in color theory.

How experts think

Marty Neumeier's definition in The Brand Gap is the one most designers use: "A brand is a person's gut feeling about a product, service, or company." You don't own the brand; your customers hold it. What you control is the identity: the name, words, visuals and behavior that shape that gut feeling.

Expert habitWhat it meansWhat it changes in your prompt
Strategy before pixelspositioning, audience and personality are decided before anyone draws a logoask for the strategy first; feed it into every visual prompt
Different beats bettera brand wins by being distinct and memorable in its category, not by claiming "quality"ask "what do all competitors say?" and then for the opposite
A logo is not a brandthe logo is one asset in a system of color, type, imagery, voice and behaviorask for a system with rules, not one pretty mark
Consistency compoundsthe same assets used the same way for years build recognitionask for rules and examples of misuse, not just the ideal
Design for the worst casea mark must work at 16 px, in one color, embroidered, on a busy photodescribe the hardest uses up front
Ownable, not just likablea name or mark you can't register or that confuses buyers with another firm is a liabilityscreen names and marks legally before you fall in love
Test with the audiencestakeholders' taste is not evidence; target customers' reactions areask the model for test plans, not just opinions
research ─► strategy ─► verbal identity ─► visual identity ─► guidelines ─► rollout
 audit,      positioning,  name, voice,      logo system,       rules,          touchpoints,
 interviews  personality   messaging         color, type,       examples,       launch,
                                             imagery            misuse          measure

Vocabulary

TermWhat it meansSay it like
Brandthe reputation and feeling people associate with a company or product"Describe the brand as a customer would after one year of use."
Brand identitythe designed, controllable expression: name, logo, color, type, voice"Propose a brand identity system, not just a logo."
Brand strategythe plan for who the brand serves, how it differs and how it should be perceived"Write the brand strategy before any visuals."
Purposewhy the organization exists beyond making money"Draft three purpose statements under 15 words; avoid clichés like 'empower'."
Mission / visionwhat you do now / the future you're working toward"Separate the mission (now) from the vision (10 years out)."
Valuesprinciples that guide behavior; useful only if they rule something out"For each value, give a decision it would force us to make."
Positioningthe place you want to occupy in a customer's mind relative to alternatives"Position us against spreadsheets, not against other apps."
Positioning statementan internal sentence: for whom, which category, what benefit, unlike what"Write a positioning statement in Geoffrey Moore's format."
Target audiencethe specific people you are for (and implicitly not for)"Our audience is independent bookkeepers, not enterprises."
Category / frame of referencethe market you say you compete in; it sets buyers' expectations"Suggest three category framings and what each implies about price."
Competitive alternativeswhat customers would use if you didn't exist, including doing nothing"List the competitive alternatives, including manual workarounds."
Point of differencean attribute you have that alternatives lack and customers value"Which of these is a real point of difference, and which is table stakes?"
Point of paritysomething you must match just to be considered"List the points of parity we must not neglect."
Competitive mapa 2-axis chart (perceptual map) placing competitors on attributes buyers care about"Plot these 8 brands on a competitive map: price vs. playful–serious."
Brand promisethe one experience customers can count on every time"State the brand promise in one sentence a support agent could keep."
Reason to believe (RTB)proof that makes a claim credible: data, features, history"Give two reasons to believe for each claim."
Brand personalityhuman traits the brand expresses, usually 3–5 adjectives"Pick 4 personality traits and their 'but not' counterparts."
Archetypea stock character (Hero, Sage, Jester…) used as shorthand for personality"Which archetype fits, and which would be a cliché in our category?"
Voicethe consistent personality in words; it doesn't change"Define our voice with 'this, not that' pairs."
Tonehow the voice adapts to the situation (an error message vs. a launch)"Show the same message in our tone for a celebration and an outage."
Messaging hierarchyordered set of messages: tagline, value proposition, pillars, proof"Build a messaging hierarchy from this positioning."
Taglinea short public line that sums up the promise"Give 15 tagline options, grouped by approach."
Boilerplatethe standard "About us" paragraph used in press releases"Write a 60-word boilerplate in our voice."
Naming briefthe requirements a name must meet: meaning, tone, length, markets"Here is the naming brief; propose names that meet every criterion."
Trademarka legally protected sign (name, logo) identifying the source of goods or services"Flag names that look likely to conflict; I'll run the legal check."
Wordmark (logotype)the name set in distinctive type, no symbol"Explore wordmarks with custom letter details."
Lettermark (monogram)initials only"A lettermark for 'NW' that works as an app icon."
Symbol / brandmarka standalone icon: pictorial (a recognizable thing) or abstract (a shape)"Abstract symbols only; no literal coffee cups."
Combination marka symbol and a wordmark together"A combination mark with a stacked and a horizontal version."
Emblemtext enclosed inside a shape, like a badge or seal"An emblem for merchandise; a simplified mark for digital."
Lockupa fixed arrangement of logo elements (and sometimes a tagline or sub-brand)"Define horizontal, stacked and symbol-only lockups."
Clear space / minimum sizethe empty margin required around a logo / the smallest it may appear"Specify clear space as a multiple of the symbol's height."
Palette rolesprimary, secondary, accent, neutral and semantic colors, each with a job"Assign each color a role and a rough usage share."
Type pairingthe combination of typefaces (e.g. a display face for headlines, a text face for body)"Pair a characterful headline face with a neutral text face."
Imagery style / art directionrules for photos and illustration: subject, light, crop, color treatment"Write art direction for photography: subjects, light, framing, what to avoid."
Iconographythe family of small symbols in a product or on signage"Describe an icon style: stroke weight, corner radius, grid."
Distinctive assetselements people recognize without the name: a color, shape, sound, character"Which of our assets could become distinctive if we used them for years?"
Brand guidelinesthe rulebook: strategy, logo use, color, type, imagery, voice, examples"Outline brand guidelines for a 5-person team without a designer."
Brand architecturehow a parent brand and its products or sub-brands relate"Should the new product be a sub-brand or stand alone?"
Rebrand vs. refreshchanging strategy and identity / updating the look while keeping the equity"Is this a refresh or a rebrand? Argue both, then decide."
Brand equitythe value (recognition, trust, preference) the brand has built up"What equity would we lose if we changed the name?"
Touchpointany place a person meets the brand: site, invoice, packaging, support call"List every touchpoint and what must change on each."
Mood boarda collage of references that sets a visual direction before design"Describe three mood board directions in words I can give an image tool."
Vector vs. rastershapes defined by math, scalable forever (SVG, EPS, PDF) / a grid of pixels (PNG, JPG)"This is a raster sketch; list what a designer must fix when redrawing it as vector."

Brand strategy

Positioning

Positioning is the decision everything else expresses. Two well-known formats:

Geoffrey Moore, Crossing the Chasm (positioning statement)
For [target customers] who are dissatisfied with [current alternative],
our product is a [new product category] that provides [key benefit].
Unlike [the product alternative], we have [key differentiators].
 
April Dunford, Obviously Awesome (five components, worked in this order)
1. Competitive alternatives   what customers would do if we didn't exist
2. Unique attributes          what we have that the alternatives lack
3. Value (and proof)          what those attributes enable for customers
4. Target market              who cares a lot about that value
5. Market category            the context that makes the value obvious
(+ relevant trends, optional)

Dunford's point is the order: start from alternatives, not from the category you wish you were in. Ask the model to work through the five steps with you, one at a time, rather than to "write our positioning".

calm, personal hectic, impersonal slow service fast service open corner artisan café slow, pricey station chain fast, impersonal drive-through Slow Lane fast and calm 1. Pick two attributes that buyers care about 2. Plot every option incl. doing nothing 3. Find an open corner that your proof supports 4. Check it's wanted empty can mean unwanted: ask target customers
A competitive map makes white space visible, but only if buyers care about both axes.

Personality and archetypes

Personality is usually 3–5 traits, each with a "but not" to stop it sliding into a cliché: confident, but not arrogant; warm, but not cute. Archetypes come from Margaret Mark and Carol Pearson's The Hero and the Outlaw (2001), which adapted Jungian ideas into twelve brand characters. Treat them as a vocabulary for briefing, not as science: they are a framework from practice, not a validated model of how buyers behave.

ArchetypeCore desireTypical voiceRisk in use
Innocentsafety, simplicityoptimistic, plainbland
Explorerfreedom, discoveryadventurous, independentgeneric "adventure" imagery
Sageunderstandingexpert, measuredlecturing
Heromastery, winningbold, motivatingchest-beating
Outlawdisruptionrebellious, bluntedgy for its own sake
Magiciantransformationvisionary, mysteriousvague promises
Everyman (Regular Guy/Gal)belongingfriendly, down-to-earthforgettable
Loverintimacy, pleasuresensual, warmover-promising
Jesterfunplayful, wittyjokes in serious moments
Caregiverprotecting othersnurturing, reassuringpatronizing
Creatormaking something newimaginative, craftedself-indulgent
Rulercontrol, stabilityauthoritative, polishedcold

Ask the model which archetype every competitor already uses. If they are all Sages, a Sage brand disappears.

Voice and tone

Voice stays constant; tone flexes with the situation. Nielsen Norman Group describes four tone dimensions that make a voice concrete: funny vs. serious, formal vs. casual, respectful vs. irreverent, enthusiastic vs. matter-of-fact. Place the brand on each, then write "this, not that" pairs with examples.

TraitWe areWe are notSounds likeDoesn't sound like
Directclear, shortcurt"Your invoice is overdue. Pay in one tap.""We noticed that perhaps your invoice may be…"
Warmhumangushing"Nice work: that's your first month closed.""OMG you're amazing!!!"
Expertspecificjargon-heavy"Matched 212 of 215 transactions.""Leveraging AI-driven reconciliation synergies."

Messaging hierarchy

Tagline           one public line                 "Books closed by Friday."
Value proposition who + outcome + how, 1–2 lines   for firms of 1–10 bookkeepers
Message pillars   3 themes, each a benefit         speed · accuracy · calm
Proof points      2–3 per pillar: numbers,         "matches 95% of lines"
                  features, customers, awards      (only if true and sourced)
Boilerplate       60–100 word "About"              for press and partners
Books closed by Friday. for firms of 1–10 bookkeepers: month-end in days, not weeks, with automatic bank matching speed accuracy calm 2-day close bank feeds 95% matched 212 of 215 300 firms 4.8 rating About us: 60–100 words for press and partners tagline one public line value proposition who + outcome + how pillars 3 benefits proof points 2–3 per pillar, boilerplate true and sourced
Each level supports the one above it; every proof point needs a source you can show.

Naming

Name typeExample patternTrademark strength (US)Trade-off
Descriptivesays what it does ("Best Plumbing")weak; hard to register without acquired distinctivenessinstantly clear, impossible to own
Suggestivehints at a benefit ("Netflix")strongneeds a moment's thought; often taken
Arbitrarya real word unrelated to the product ("Apple" for computers)strongmeaning must be built by marketing
Coined (fanciful)invented word ("Kodak")strongestempty until you fill it; spelling issues
Compound / portmanteautwo words joined or blendedvariesmany obvious ones are registered
Acronym / initialslettersweak to moderateforgettable unless already famous
Founder / place namea surname, a townvaries; surnames can be weakpersonal; hard to sell the company later

The USPTO describes this spectrum from fanciful and arbitrary (strongest) through suggestive to descriptive (weak) and generic (not protectable). A model can generate hundreds of names; it cannot tell you which are free to use.

SCREENING CHECKLIST (the model helps with 1–5; 6–8 are your job)
1. Meets the naming brief (meaning, tone, length)
2. Easy to say and spell after hearing it once
3. No unfortunate meanings or slang in target languages
4. Doesn't describe a feature you might drop
5. Not confusingly close to a competitor's name
6. Domain and social handles available (check registrars yourself)
7. Trademark search in your classes and countries
   (USPTO search, WIPO Global Brand Database, EUIPO TMview)
8. Professional clearance by a trademark attorney before launch

Visual identity

Logo types

TypeBest whenWatch
Wordmarkshort, distinctive name; new brand that needs name recognitionlong names get tiny at small sizes
Lettermarklong name, or initials already usedinitials alone carry no meaning for a new brand
Symbol (pictorial or abstract)brand will appear without words: app icons, favicons, productsa new symbol means nothing until people learn it
Combination markmost new brands: the word teaches the symbolneeds rules for when each part appears alone
Emblemheritage, education, sports, craft goodsdetail fills in at small sizes; often needs a simplified version
slowlane SL slowlane SLOW LANE · STATION COFFEE · slowlane SL slowlane SLOW LANE · STATION COFFEE · wordmark short, distinct names lettermark long names, known initials symbol app icons, favicons combination most new brands emblem heritage, craft, merchandise the same marks at favicon size, about 16 px: text too small still reads still reads use symbol only detail fills in
Test every mark at its smallest use: at favicon size, only simple ones survive.

The logo system

A professional delivers a system, not a file:

PartWhat to specify
Primary lockupthe default arrangement, usually horizontal
Alternate lockupsstacked, symbol-only, wordmark-only; when each is used
Color versionsfull color, one color, black, white (reversed) on dark
Clear spacemargin defined by a part of the logo (e.g. the height of one letter)
Minimum sizein px for screens and mm for print; a simplified mark below it
Misusedon't stretch, recolor, outline, add effects, rotate or place on busy images
Filesvector masters (SVG, PDF/EPS) plus PNG exports; favicon and app-icon sizes

Color, type, imagery and icons

ElementRoles and rulesAsk the model for
Color paletteprimary (recognition), secondary (support), accent (calls to action, sparingly), neutrals (most surfaces), semantic (success, warning, error); give HEX/RGB for screen, CMYK and a spot color for printroles, usage proportions, accessible text pairs (WCAG 2.2 AA: 4.5:1 body text, 3:1 large text and UI)
Typographya display face for headlines, a text face for body, sometimes a mono or a superfamily that does both; check the license covers web, app and print3 pairings with reasons; fallback system fonts; a type scale
Imagerysubject, lighting, framing, color treatment, diversity, what never appearswritten art direction you can paste into an image tool
Iconographygrid size, stroke weight, corner radius, filled vs. outline, metaphor rulesan icon spec and a list of the first 20 icons needed

See color theory for contrast checks and tonal scales.

Image tools for logo exploration

Image generators are good for fast, cheap exploration: dozens of directions in an afternoon, to react to and to brief a designer with. They are poor at the final mark.

They do wellThey do badly
mood boards and visual directionsspelling: lettering is often garbled or inconsistent
rough symbol ideas and silhouettesexact repeats: the "same" logo drifts between images
testing a mark in context (sign, cup, app icon mockups)clean geometry, consistent stroke weights, optical balance
style exploration (flat, line, geometric, hand-drawn)vector output: most produce pixels, not editable shapes

Prompting them. Describe the mark, not the brand's feelings: type of logo, shape language, style, color count, background, and what to exclude. Generate in batches and vary one thing at a time.

flat vector-style logo symbol, a single abstract shape suggesting a
folded page and a checkmark, geometric, thick even strokes, one color
(dark navy) on a plain white background, centered, no text, no
gradients, no shadows, no 3D, no mockup, simple enough to read at
favicon size

Why the final logo must be vector and redrawn by a designer:

  • A logo is used from a 16 px favicon to a building sign. Raster images blur or pixelate when scaled; vectors don't.
  • Auto-tracing a raster image produces lumpy curves with hundreds of points. A designer rebuilds the mark with clean geometry, consistent stroke weights and optical corrections (circles slightly larger than squares to look equal).
  • The wordmark should be set in real, licensed type (or custom-drawn letters), never generated lettering.
  • Print needs exact color specs and single-color versions, which a generator won't give you.
  • Some tools now offer vector output (e.g. Adobe Illustrator's text-to-vector feature), but the paths still need cleanup, and features change quickly.

Guidelines, architecture and rebrands

What brand guidelines contain

brand guidelines
guidelines/01-strategy/      # positioning, audience, promise02-personality/   # traits, archetype, voice, tone03-messaging/     # tagline, pillars, proof04-logo/          # lockups, clear space, misuse05-color/         # roles, values, contrast pairs06-typography/    # faces, scale, licenses07-imagery/       # photo and illustration rules08-iconography/   # grid, stroke, examples09-applications/  # site, slides, email, packaging10-assets/        # files and where to get them

Small teams need a short version: one page each for logo, color, type, voice, plus ten real examples. Long guidelines nobody opens are worse than a one-pager everyone uses.

Brand architecture

David Aaker and Erich Joachimsthaler's brand relationship spectrum (2000) runs from one master brand to many independent ones.

ModelHow it worksExample patternChoose it when
Branded houseone master brand; products are descriptors"FedEx Express", "FedEx Ground"products share audience and promise; you want every launch to build one name
Sub-brandsmaster brand plus a named product with its own character"Master + ProductName"a product needs distinct personality but benefits from the parent
Endorsed brandsan independent brand, vouched for by the parent"Courtyard by Marriott"the parent adds credibility but would constrain the product
House of brandsseparate brands, parent mostly invisible to buyersProcter & Gamble's Tide, Pampersproducts serve conflicting audiences or price points; risk must be isolated
one master brand many independent brands how much buyers see the parent Branded house one audience, one promise, one name FedEx Express, FedEx Ground Sub-brands own personality, parent still helps master name + product name Endorsed brands parent adds trust but would constrain Courtyard by Marriott House of brands clashing audiences or risk to isolate P&G: Tide, Pampers Acme Pay Acme Books Acme Payroll Acme Nimbus Acme Torch Nimbus by Acme Torch by Acme Nimbus Torch Fern Acme: owner
Further right, each brand must be built on its own, and trouble with one stays contained.

Rebrand or refresh

RefreshRebrand
What changeslogo polish, palette, type, photography, templatesstrategy, often name and identity
Keepsrecognition, core assetslittle; equity has to be rebuilt
Good reasonsdated look, inconsistent use, new channelsmerger, new market or audience, name conflict, reputation problem, strategy has moved
Bad reasonsa new executive's taste, boredomthe team is tired of it (customers usually aren't)
Risklowhigh: lost recognition, confused customers, cost across every touchpoint

Brief the model

Each input has a slot in the prompt skeleton from fundamentals; the Slot column says where it goes.

InputWhy it mattersExampleSlot
What they all saylets the model find white space"Everyone says 'simple', 'powerful', 'AI'." (paste their headlines)<materials>
What you do, in one plain sentencethe model fills gaps with generic startup language"We sell bookkeeping software to firms with 1–10 bookkeepers."<context>
Audience, specificpersonality and voice depend on who listens"Owners of small accounting firms, 40–60, skeptical of hype."<context>
Category and competitorsdifference is relative; without names you get clichés"Competitors: Xero, QuickBooks, spreadsheets, a part-time assistant."<context>
Desired perceptionthe gut feeling you're aiming at"After a month, customers should feel 'I'm in control'."<context>
Proof you actually havestops invented claims"Founded by two former bookkeepers; 300 paying firms; 4.8 app-store rating."<context>
Hard usage contextsdrives logo type and detail level"App icon, invoice footer, embroidered polo shirts, one-color fax."<context>
Existing assets and equityso it doesn't throw away what works"Customers know our green; keep it."<context>
Markets and languagesnames and colors carry meanings"UK, US and Germany; must work in German."<context>
References and whytaste is easier to show than describe"We like Mailchimp's voice for its warmth, not its jokes."<examples>
Personality with "but not"a trait without a limit drifts"Confident but not arrogant; plain-spoken but not cold."<requirements>
Constraintsbudget, timeline, tools, who executes"No designer yet; I'll use Figma and an image generator."<requirements>
Decision criteria and deciderthe model can rank options against them"The two founders decide; criteria: memorable, ownable, calm."<quality_bar>
Output formatsaves reformatting"A table, then a one-page summary."<output_format>

What to ask for

StageDeliverables to ask for by name
Discoverybrand audit (how the brand shows up today, touchpoint by touchpoint); competitor audit (claims, colors, archetypes); stakeholder interview guide; customer interview guide; summary of themes
Strategypositioning statement; Dunford positioning worksheet; brand platform (purpose, promise, values, personality); competitive map; archetype analysis; messaging hierarchy
Verbal identitynaming brief; name longlist grouped by type; shortlist with rationale and risks; voice chart with "this, not that"; tone by situation; tagline options; boilerplate
Visual identitydesign brief for a designer; three mood board directions in words; logo exploration prompts; palette with roles and contrast pairs; type pairings; art direction; icon spec
Rolloutguidelines outline; touchpoint inventory with priorities; launch announcement; internal FAQ; customer FAQ for a rename
Reviewcritique against the strategy; misuse examples; a test plan with target customers

The prompt chain

The main deliverable is a brand platform for a product: positioning, personality and voice, a messaging hierarchy, then a direction for the visual identity that a designer can work from. Run stages 1–3 in one chat. Run stage 4 in a fresh chat, because a model reviewing its own draft shares its blind spots.

1. Interview

Use first, when you have notes about the product and customers but no agreed strategy.

<context>
I need a brand platform for [business, e.g. bookkeeping software for small accounting firms]: positioning, personality and voice, a messaging hierarchy, then a direction for the visual identity.
Stage: [stage: new brand | refresh | rebrand]. I am not a brand strategist.
</context>
 
<task>
Interview me first: ask the questions a brand strategist asks in discovery, before any strategy is written.
</task>
 
<requirements>
- At most 10 questions, grouped under: the business and audience, competitive alternatives (including doing nothing), proof, desired perception and personality, hard usage contexts, constraints and who decides.
- For each question: why it matters to the platform, and 2-3 typical answers I can pick from or adapt.
- Ask explicitly what proof I have (numbers, customers, history); I may not know what counts.
- Do not draft positioning, names or visuals yet.
</requirements>
 
<process>
Ask the questions, then stop and wait for my answers.
</process>

2. Options

Use once the interview is answered, to choose a positioning territory before anything is built on it.

<materials>
[paste the interview questions and my answers, and competitors' homepage headlines if you have them]
</materials>
 
<task>
Propose 3 materially different positioning territories for this brand.
</task>
 
<requirements>
- Work in April Dunford's order: competitive alternatives first, then unique attributes, value and proof, best-fit audience, market category.
- For each territory: the alternative it positions against, the category frame and what it implies about price, the point of difference and the proof behind it, a fitting archetype (as a briefing shorthand, not research), which archetypes competitors already use, and the main risk.
- Mark any point of difference with no proof in my answers [NEEDS PROOF].
- Recommend one and say what would have to be true for each of the others to win.
</requirements>
 
<process>
Give the territories as a table, then the recommendation, then stop and wait for my choice.
</process>

3. Produce

Use after you have chosen a territory: the full skeleton, filled in for the platform.

<materials>
Interview answers and the territory I chose: [paste]
Competitors' claims: [paste their homepage headlines, taglines and About paragraphs]
</materials>
 
<context>
Business: [one plain sentence, e.g. "We sell bookkeeping software to firms with 1-10 bookkeepers."]
Audience: [audience, e.g. owners of small accounting firms, 40-60, skeptical of hype]
Desired feeling after a month: [feeling, e.g. "I'm in control"]
Proof we have: [real facts only, e.g. founded by two former bookkeepers; 300 paying firms]
Hard uses: [contexts, e.g. app icon, invoice footer, embroidered shirt, one-color stamp]
Stage: [stage: new brand | refresh | rebrand]
Logo type to explore first: [logo type: wordmark | lettermark | symbol | combination mark | emblem | let the strategy decide]
The platform is for [readers, e.g. the two founders, who decide, and a freelance designer who will be briefed from it].
</context>
 
<examples>
[optional: 1-3 brands outside our category whose voice or identity you admire, and exactly what you admire, e.g. "Mailchimp's voice: the warmth, not the jokes"; 1 you don't want to resemble, and why]
</examples>
 
<task>
A brand platform: positioning, promise, personality and voice, messaging hierarchy, then a direction for the visual identity that a designer can work from.
</task>
 
<requirements>
- Positioning: a statement in Geoffrey Moore's format, built from Dunford's five components, against the alternative I chose, plus a one-line homepage version.
- Brand promise: one sentence a support agent could keep.
- Personality: 3-5 traits, each with a "but not"; one archetype from Mark and Pearson's twelve as a shorthand, presented as a framework, not as research.
- Voice: a position on Nielsen Norman Group's four tone dimensions (funny-serious, formal-casual, respectful-irreverent, enthusiastic-matter-of-fact), a "we are / we are not" table with a sample line for each, and the tone for an error, a price increase and a celebration.
- Messaging hierarchy: 5 tagline options, a value proposition, 3 pillars, 2-3 proof points per pillar, and a 60-100 word boilerplate.
- Identity direction: logo type and why, shape language, palette roles (primary, secondary, accent, neutrals), type direction and imagery rules, each traced to a line of the strategy. No finished logo.
- Never invent numbers, names, quotes or sources: write [NEEDS PROOF] where a claim has no proof I supplied.
- Do not say any name or mark is available or trademark-free.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The competitor test: put a competitor's name in place of ours; rewrite every line that still works until it couldn't.
- Every pillar and proof point traces to my materials, or is marked [NEEDS PROOF].
- Each "but not" rules out something a competitor does.
- The identity direction survives the hard uses: 16 px, one color, embroidery.
</quality_bar>
 
<output_format>
Sections in this order: positioning, promise, personality, voice (tables), messaging hierarchy (a table: level | content | proof | source), identity direction (one short paragraph per element). Under 1,500 words, then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
Write the positioning first and check it against the quality bar before building the rest on it.
</process>

4. Review

Use when the platform is back: open a fresh chat, paste the brief and the platform, and ask for a ranked critique, not a rewrite.

<materials>
The brief: [paste the stage 3 prompt]
The brand platform: [paste]
Competitors' homepage headlines: [paste]
</materials>
 
<context>
This is a fresh chat. The platform below was written in another conversation; review it as someone seeing it for the first time, before it is shared with a designer.
</context>
 
<task>
A review of this brand platform. Do not rewrite it.
</task>
 
<requirements>
- Judge against: strategic fit, distinctiveness, simplicity, scalability of the identity direction, ownability, consistency of the voice across situations, honesty of claims, and whether a non-designer could use it.
- Run the competitor test on every line of the positioning and messaging, and list the lines that fail.
- List every claim that needs proof, and say whether the brief supplied it.
- Quote the line behind every finding.
</requirements>
 
<quality_bar>
Before answering, check that every finding names a criterion, quotes the line, and proposes a specific fix.
</quality_bar>
 
<output_format>
A table: rank | finding | criterion | quoted line | fix, most important first. Then one line: ready, fix first, or rethink the positioning.
</output_format>
 
<process>
Rank the issues by how much they weaken the brand; do not produce a revised platform.
</process>

Goals and loops

In an agent that runs commands and checks its own work, such as Claude Code, the Produce stage can run until it measurably passes: /goal keeps starting turns until a condition holds, and /loop repeats a prompt or skill on an interval. The checker reads only the conversation, so every condition below names a check whose output Claude prints, the constraints that must survive, and a bound. How to write your own: Goals and loops.

TaskCompletion condition
Naming longlist/goal names.md lists 20 candidates, each with its naming type and the whois result for its .com printed as available or taken; or stop after 8 turns. Trademark clearance is still a job for you and counsel
Brand guidelines/goal brand-guidelines.md has every section of the template, every color lists hex, RGB and CMYK with its contrast ratio against white and black printed by python contrast.py, and every lockup links an SVG; or stop after 10 turns
Voice across pages/goal vale site/ with a Vale style built from the voice guide reports 0 errors (show its summary), and no page was rewritten beyond the flagged sentences; or stop after 10 turns

Loops that fit: a monthly /schedule that collects competitors' new messaging, campaigns and rebrands from their sites and summarizes what changed against your positioning map.

What a checker can't judge: whether the brand means anything to the people you sell to. Test names and positioning with real customers.

Prompt library

One entry per deliverable a brand designer or strategist produces, besides the platform in the chain. Quick is for when you're in a hurry; Full is for when the result matters.

Competitor audit

Stage: produce · Use when: you need a point of difference and suspect you sound like everyone else.

Quick

Here are the homepage headlines, taglines and About paragraphs of [number, e.g. 6] competitors: [paste]. Build a table: brand, main claim (quoted), archetype, personality traits, dominant color, typical imagery.
Then list the claims everyone makes, the claims nobody makes, and 3 open positioning territories with the risk of each. Quote; don't paraphrase.

Full

<materials>
[paste each competitor's homepage headline, tagline and About paragraph, labeled by brand, plus a short description of each logo and its colors]
</materials>
 
<context>
We are: [one plain sentence, e.g. "bookkeeping software for firms with 1-10 bookkeepers"]
Our audience: [audience, e.g. owners of small accounting firms]
Alternatives our customers consider: [alternatives, e.g. QuickBooks, spreadsheets, a part-time assistant, doing nothing]
</context>
 
<task>
A competitor audit that shows where every brand in the category sits and which positioning territories are open to us.
</task>
 
<requirements>
- For each brand: main claim (quoted), archetype as a shorthand, 3 personality traits, dominant color, typical imagery, logo type.
- The claims everyone makes (points of parity) and the claims nobody makes.
- A competitive map: propose 2 axes buyers actually choose on, and place every alternative on it, including doing nothing.
- 3 open territories, each with the proof we would need to own it and the risk (an empty corner can mean nobody wants it).
- Use only what is in the materials and mark anything you infer as inferred. Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where a fact is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every claim attributed to a brand is quoted from the materials.
- The map's axes are things buyers choose on, not things we happen to be good at.
- Each open territory says how we would test that customers want it.
</quality_bar>
 
<output_format>
A table with one row per brand, the map as a list of coordinates on the two axes, the territories in under 100 words each, then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Naming longlist

Stage: produce · Use when: you have a naming brief and want breadth, then a defensible shortlist to take into legal checks.

Quick

Generate 40 names from this naming brief: [paste], 8 each that are suggestive, arbitrary, coined, compound and metaphor-based. For each: one-line rationale, pronunciation and main risk.
Don't claim any name is available; I'll check domains and trademarks myself.

Full

<materials>
[optional: paste the naming brief, if you have one]
</materials>
 
<context>
What it names: [thing: company | product | feature | sub-brand]
Positioning and personality: [paste, e.g. "fast service, calm feeling"; confident but not loud]
Must convey: [ideas, e.g. calm, speed, a pause in the day]
Must avoid: [words, e.g. the ones competitors use: "brew", "bean", "daily"]
Length: [limit, e.g. 2 syllables or under 8 letters]
Markets and languages: [markets, e.g. US, UK and Germany]
Competitors: [names, e.g. the 3 brands customers compare us with]
</context>
 
<examples>
[optional: 2-3 names you like from any category and why, e.g. "a short real word used in an unexpected way"; 1 you dislike, and why]
</examples>
 
<task>
A longlist of names that meet the brief, then a scored shortlist of the 10 strongest, to take into domain and trademark checks.
</task>
 
<requirements>
- 40 names: 8 each that are suggestive, arbitrary, coined, compound and metaphor-based. Label each with its type on the USPTO strength spectrum (fanciful and arbitrary strongest, suggestive next, descriptive weak).
- For each: a one-line rationale, pronunciation, and any risk (a meaning or slang in a target language, likely to be common, too close to a competitor).
- Score the 10 strongest 1-5 on: fits the brief, memorable, easy to say and spell after hearing it once, distinct from competitors, works in every target language, room to grow.
- Do not claim any name is available or trademark-free: write [CHECK REGISTER] beside every shortlisted name.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- No name describes a feature we might drop.
- No two shortlisted names are variations of one idea.
- Doubtful meanings in a target language are flagged, not kept silently.
</quality_bar>
 
<output_format>
The longlist as a table (name | type | rationale | pronunciation | risk), then the shortlist scoring table, then the checks I must run myself (domains, handles, trademark search in my classes and countries, attorney clearance), then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Voice and tone guide

Stage: produce · Use when: positioning is set, and copy from different people sounds like different companies.

Quick

Create a voice guide for [brand, e.g. a bookkeeping app] whose personality is [traits, e.g. direct but not curt, warm but not gushing]: a position on NN/g's four tone dimensions, a "we are / we are not" table with sample lines, the tone for an error, a price increase and a celebration, and 10 words we use and 10 we avoid.
Check that every sample line sounds like the same person.

Full

<materials>
Positioning and personality: [paste]
Copy to rewrite: [paste a real page or email]
</materials>
 
<context>
Audience: [audience, e.g. owners of small accounting firms, skeptical of hype]
Who will use the guide: [writers, e.g. two founders, a support team of four, a freelance copywriter]
Channels: [channels, e.g. app interface, support emails, website, social posts]
</context>
 
<examples>
[paste 3 pieces of copy that sound like us, or like who we want to be, and why; 3 that don't, and why]
</examples>
 
<task>
A voice and tone guide that people with no writing training can follow.
</task>
 
<requirements>
- A position on Nielsen Norman Group's four tone dimensions (funny-serious, formal-casual, respectful-irreverent, enthusiastic-matter-of-fact), each 1-5 with a reason.
- A "we are / we are not" table, 3-5 rows, each with a line that sounds like us and one that doesn't.
- Tone by situation: onboarding, error, price increase, outage, celebration, legal notice; one example line each.
- 8 concrete writing rules inferred from the examples (sentence length, person, contractions, jargon, humor, punctuation), plus 10 words we use and 10 we avoid.
- Rewrite the pasted copy with the rules and say which rule drove each change.
- Never invent numbers, names, quotes or sources: write [NEEDS PROOF] beside any claim in the rewrite that I haven't supported.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every rule can be tested on a sentence: a writer could say whether it was followed.
- The outage and legal lines still sound like the same brand, only more serious.
- The rewrite changes no facts and adds no claims.
</quality_bar>
 
<output_format>
The dimensions table, the voice table, the tone table, the rules, the word lists, then the rewrite with change notes. Under 1,200 words.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Logo exploration prompts

Stage: produce · Use when: the strategy is set and you want visual directions to react to, or to brief a designer with.

Quick

From this strategy: [paste], give 4 logo directions of different kinds (e.g. abstract symbol, pictorial, lettermark, emblem). For each: why it fits, and an image-generator prompt that describes only the mark (shape, stroke, color count, plain background) and ends "no text, no gradients, no shadows, no mockups, no 3D".
Say how each would fail at 16 px.

Full

<materials>
Strategy: [paste the positioning, personality and audience]
</materials>
 
<context>
Name: [name, e.g. Slow Lane]
Hard uses: [contexts, e.g. favicon, a hanging sign seen from 20 m, an embroidered polo shirt, a one-color stamp]
Category clichés to avoid: [symbols, e.g. cups, beans, steam]
These are explorations to react to and to brief a designer with, not the final logo.
</context>
 
<examples>
[optional: 1-3 marks from outside our category that you admire, described in words (shape, weight, color count) with why; 1 style you don't want. Describe them; don't upload competitors' logos]
</examples>
 
<task>
Four logo directions, each turned into prompts for an image generator, so I can explore before briefing a designer.
</task>
 
<requirements>
- The directions differ in kind, e.g. geometric abstract, pictorial, lettermark, emblem.
- For each: the idea in one sentence and the line of the strategy it expresses.
- An image prompt that describes only the mark: logo type, shape language, stroke, color count, plain background, and an exclusion list (no text, no gradients, no shadows, no mockups, no 3D). No lettering in any prompt: image tools garble it, and the wordmark will be set in real type. Describe a lettermark in words for a designer instead.
- 3 variations per direction, each changing one thing.
- How each would fail at 16 px, in one color and embroidered.
- No reference to existing brands or named living designers.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Each mark can be described in one sentence and drawn from memory.
- At least one direction avoids the category's literal symbols.
- No prompt asks for text or a mockup.
</quality_bar>
 
<output_format>
Per direction: idea, strategy link, prompt (plain text I can paste), 3 variations, failure cases.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Designer brief

Stage: produce · Use when: you're hiring a designer and want to brief them like a professional would.

Quick

Write a one-page identity design brief for a freelance designer from this strategy: [paste]. Include the audience, positioning, personality, competitors' visual codes to avoid, hard uses, deliverables (logo system, palette with roles, type pairing, icon style, short guidelines, vector masters), timeline, budget and how options will be judged.
Check that a designer could quote from it without a call.

Full

<materials>
Strategy: [paste the brand platform]
Competitors' visual codes: [paste the competitor audit, or describe their colors, marks and imagery]
</materials>
 
<context>
Stage: [stage: new brand | refresh | rebrand]
Assets to keep: [assets, e.g. "customers know our green", or "none"]
Designer: [designer, e.g. a freelancer found through a portfolio site, briefed by email]
Timeline: [dates, e.g. first concepts in 3 weeks, final files in 8]
Budget range: [range, e.g. $5,000-8,000]
Who decides: [decider, e.g. the two founders]
</context>
 
<examples>
[1-3 identities from outside our category that you admire, and exactly what, e.g. "the restraint: one color, one typeface"; 1 you don't want to resemble, and why]
</examples>
 
<task>
A one-page identity design brief, so a designer can quote, plan and start without a discovery call.
</task>
 
<requirements>
- Sections: background, audience, positioning, personality, competitors and their visual codes to avoid, hard usage contexts, deliverables, what I will provide, timeline, budget, how options will be judged, and who decides.
- Deliverables by name: a logo system (primary and alternate lockups, color versions, clear space, minimum size), a palette with roles, a type pairing with licenses covering web, app and print, an icon style, short guidelines, and vector masters (SVG, PDF) plus PNG exports.
- Judging criteria drawn from the strategy, not taste, e.g. distinct from competitors, legible at 16 px, works in one color.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- It fits on one page, about 500 words.
- Someone other than me could check every judging criterion on a concept.
- It asks for a system with rules, not one mark.
</quality_bar>
 
<output_format>
One page with the sections above as short headings, then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Brand guidelines

Stage: produce · Use when: an identity exists and a small team needs rules they will actually read.

Quick

Write short brand guidelines for a [team, e.g. 5-person team with no designer] from these inputs: [paste]. Sections: strategy (one page), logo (lockups, clear space, minimum size, 6 misuse examples), color, type, imagery, voice, 10 application examples.
Keep each section under 200 words, each with a "when in doubt" rule.

Full

<materials>
[paste the positioning, personality, voice guide, a description of the logo files, palette HEX values and typefaces]
</materials>
 
<context>
Team: [team, e.g. 5 people, no in-house designer, making slides, emails and social posts]
Brand architecture: [architecture: single brand | branded house | sub-brands | endorsed brands | house of brands]
What the team makes most: [list, e.g. sales slides, invoices, social posts]
</context>
 
<task>
Short brand guidelines a non-designer can make a correct slide from.
</task>
 
<requirements>
- Sections: strategy on one page; logo (lockups, clear space, minimum size in px and mm, color versions, 6 misuse examples); color (roles, values, usage share, accessible text pairs); typography (faces, weights, scale, fallbacks, license notes); imagery; voice; 10 application examples (email signature, slide, social post, invoice and so on).
- Text color pairs must meet WCAG 2.2 AA (4.5:1 for body text, 3:1 for large text and UI); write [VERIFY] beside every ratio so I check it with a contrast tool.
- Each section under 200 words, with one "when in doubt" rule.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every rule comes with an example of correct use, and the logo section with examples of misuse.
- Nothing contradicts the strategy or the voice guide.
- The whole document can be read in 10 minutes.
</quality_bar>
 
<output_format>
Markdown with one heading per section and tables for color and type, then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Refresh or rebrand

Stage: options · Use when: someone proposes a new name or look and you need a reasoned decision before any design work.

Quick

We're considering [change: refresh | rebrand | new name] because [reason, e.g. a merger, a name conflict, a dated look]. Argue for a refresh, a full rebrand and no change: equity kept or lost, cost across our touchpoints, risks.
Recommend one, say what would have to be true for the others, and stop there.

Full

<materials>
[paste what customers say about us, any recognition data, and the list of touchpoints: site, app, invoices, signage, packaging, vehicles]
</materials>
 
<context>
Why change is proposed: [reason: dated look | inconsistent use | merger | new market or audience | name conflict | reputation problem | new leadership's taste]
What customers know us for: [assets, e.g. the green, the owl symbol, the name]
Who decides, and when: [decider, e.g. the board, in 6 weeks]
</context>
 
<task>
A reasoned recommendation on whether to refresh, rebrand or change nothing, so the decision rests on strategy rather than taste.
</task>
 
<requirements>
- Argue each option in turn, at its strongest: equity kept or lost, cost and effort across the listed touchpoints, risks, and timeline.
- Separate good reasons for change (the strategy has moved, a name conflict, a merger) from bad ones (boredom, a new executive's taste), and say which this case is.
- If the name may change, list the legal checks needed and who does them.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Each option gets its strongest case, not a straw man.
- The recommendation names the evidence that would change it.
- Costs are listed per touchpoint, not as "significant".
</quality_bar>
 
<output_format>
A comparison table (option | equity kept | equity lost | cost by touchpoint | risk), the recommendation in under 150 words, then the assumptions.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on. Give the options and the recommendation, then stop; don't plan the rollout yet.
</process>

Logo concept critique

Stage: review · Use when: you have a sketch or generated image and want to know what a professional would change.

Quick

[attach image] An early logo concept for [brand, e.g. a coffee shop by a commuter station], used on [contexts, e.g. cup sleeve, hanging sign, app icon]. Critique it: concept clarity, what it resembles, legibility at 16 px and at 20 m, one-color and reversed versions, stroke consistency, and what must change when it is redrawn as vector.
Rank the fixes; don't redraw it.

Full

<materials>
[attach the image or images]
Strategy: [paste the positioning and personality]
</materials>
 
<context>
Brand: [brand, e.g. Slow Lane, a coffee shop by a commuter station]
Hard uses: [contexts, e.g. cup sleeve, a hanging sign seen from 20 m, app icon, one-color rubber stamp]
Where it came from: [source: image generator | sketch | a designer's first round]
</context>
 
<task>
A critique of this logo concept before it goes further, so I know what a designer would change. Do not redraw it.
</task>
 
<requirements>
- Judge against the rubric: strategic fit, distinctiveness, simplicity, scalability, ownability.
- Check: the concept in one sentence; common category symbols or existing logos it resembles; legibility at 16 px and at 20 m; one-color, reversed and embroidered versions; balance and stroke consistency; any generated lettering.
- Say what must change when it is redrawn as vector with licensed type.
- Never invent numbers, names, quotes or sources: write [NEEDS SOURCE] where support is missing. Name a resembling logo only if you are confident, and recommend a reverse image search either way.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- Every finding points at a specific part of the image.
- The fixes are ranked by how much they change the result, not by how easy they are.
- No finding is taste alone, without a criterion behind it.
</quality_bar>
 
<output_format>
A table: rank | finding | criterion | fix. Then 3 revision directions, one sentence each.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

Worked example

Weak prompt

Make me a logo for my coffee shop. Something modern and cool.

What's wrong

  • No strategy: who it's for and how it differs, so every direction is a guess.
  • "Modern and cool" means nothing to a designer or a model; it produces the category average.
  • No hard uses (sign, cup, app icon, stamp), so nothing forces a simple mark.
  • It asks an image tool for a finished logo, which it can't deliver: the lettering comes out garbled and the file is pixels.

Strong prompt

<context>
I'm opening a coffee shop called "Slow Lane" next to a commuter train station.
Audience: commuters 25-45 who want a calm 3 minutes, not a lounge.
Competitors: a chain cafe in the station (fast, impersonal) and an artisan cafe 1 km away (slow, expensive).
We want to own "fast service, calm feeling". Proof we have: none yet; we open in March.
Hard uses: a cup sleeve, a hanging sign seen from 20 m, a loyalty app icon, a rubber stamp in one color.
</context>
 
<examples>
I like the restraint of old railway signage: one color, bold shapes, no decoration. I don't want the cozy hand-drawn chalkboard look every cafe nearby has.
</examples>
 
<task>
A short brand strategy and three logo directions in words, each with an image-generator prompt, so I can explore before hiring a designer.
</task>
 
<requirements>
- A positioning statement in Geoffrey Moore's format and 4 personality traits, each with a "but not".
- Three logo directions of different kinds (e.g. symbol, lettermark, emblem), each traced to the positioning.
- Each image prompt describes only the mark: shape, stroke, color count, plain background; it excludes text, gradients, shadows and mockups.
- Never invent numbers, names, quotes or sources: write [NEEDS PROOF] for any claim we can't back yet.
- End with a list of the assumptions you made.
</requirements>
 
<quality_bar>
Before answering, check your draft against:
- The competitor test: neither cafe could use these lines or marks.
- Every mark survives favicon size and one color.
- No image prompt asks for lettering.
</quality_bar>
 
<output_format>
The positioning, the traits as a table, then per direction: idea, why it fits, image prompt. Under 500 words.
</output_format>
 
<process>
Ask me up to 3 questions first if anything missing would change your answer; otherwise go on.
</process>

What came back

Illustrative, abridged:

Positioning: For commuters who want a calm moment on the way to the train, Slow Lane is a station coffee bar that serves you in under 3 minutes [NEEDS PROOF] in a quiet room. Unlike the chain cafe, it feels personal; unlike the artisan cafe, it won't make you miss your train.
 
Traits: calm, but not sleepy · quick, but not rushed · warm, but not cutesy · plain-spoken, but not blunt
 
Direction 1, symbol: a cup whose steam is a winding road.
Prompt: flat vector-style logo symbol, a single continuous line forming a steaming cup that doubles as a winding road, thick even stroke, rounded ends, one color (deep green) on plain white, centered, no text, no gradients, no shadows, no 3D, no mockup
 
Direction 2, lettermark: "SL" as two lanes merging.
Prompt: minimal monogram logo, the letters "SL" in a bold geometric sans serif, the S flowing into the L like two road lanes merging, one color (deep green), plain white background, no gradients, no mockup
 
Direction 3, emblem: the name around a ring like a station clock face …
 
Assumptions: deep green reads as calm; the sign is lit at night.

How an expert judged it

  • Strategic fit: pass. The positioning names both alternatives and the tension; "won't make you miss your train" is a line neither competitor could say.
  • Personality: pass. Each "but not" rules out something a competitor does.
  • Honesty: pass. The model marked the under-3-minutes claim [NEEDS PROOF] instead of asserting it; time real orders before using it.
  • Image prompts: fail on direction 2. The prompt asks the image tool to draw the letters "SL". Generators garble or vary letterforms, and a lettermark's whole value is in drawn letters, so this direction should be described in words for a designer, not generated. It also breaks the third quality-bar check: read the prompts against your bar, not just the pictures they make.
  • Scalability: weak on direction 3. An emblem with the name around a ring won't survive the app icon or the stamp; it needs a simplified version.

The follow-up

Direction 2's prompt asks the image tool to draw the letters "SL", which breaks the no-lettering check. Keep directions 1 and 3 as they are. Replace direction 2's prompt with a written brief a designer could use to draw the lettermark: the idea, how the letters are built, stroke weight relative to the symbol in direction 1, and how it must read at 16 px. For direction 3, add a simplified version for the app icon and the stamp. Change nothing else.

Judge the output

CriterionGoodRed flag
Strategic fiteach choice traces back to audience, difference or personality"modern, clean, innovative" with no link to the strategy
Distinctivenesswouldn't fit a named competitor if you swapped the namethe competitor test passes: it could be anyone's
Simplicitythe mark can be described in one sentence and drawn from memorymultiple ideas crammed into one symbol
Scalabilitylegible at 16 px and in one color, works reversedthin lines, tiny details, gradients doing the work
Ownabilityname and mark are unusual in the category and pass your searchesa dictionary word in your own category; a common symbol (leaf, swoosh, globe)
Consistencyvoice examples sound like the same person in every situationtone swings or turns into a different brand under stress
Honestyevery claim has a reason to believe you suppliedinvented stats, awards or customer counts
Usability of guidelinesa non-designer can make a correct slide from themrules without examples, or 80 pages

The same checks are built into every <quality_bar> on this sheet, so the model applies them before you see a draft; use the table to check what comes back. The quickest test of distinctiveness:

Put [competitor, e.g. the station chain cafe]'s name in place of ours in this platform. Which lines still work? List them with the reason; don't rewrite yet.

Pitfalls

PitfallWhy it hurtsFix
Starting with the logovisuals with no strategy are taste contestsstrategy first; brief visuals from it
Adjective soup ("innovative, trusted, modern")every brand says it; the model echoes ittraits with "but not", plus examples of sound-like and not
Believing "this name appears available"the model has no live trademark or domain datasearch registers yourself; hire an attorney for clearance
Using a generated image as the final logopixels, garbled lettering, look-alike risk, weak copyright positionuse it to explore; have a designer redraw in vector with licensed type
Literal symbols (a cup for coffee, a house for real estate)everyone in the category uses themask for abstract or metaphorical directions too
Designing only for the ideal casethe logo breaks on a favicon, invoice or embroiderylist hard uses in every prompt
Too many colors and typefacesinconsistent, expensive, hard to useone or two typefaces; palette with roles and proportions
Copying a famous brand's stylelooks derivative; legal riskreference qualities ("warm, plain-spoken"), not brands to copy
Rebranding out of boredomthrows away recognition customers rely onrefresh first; rebrand only for a strategic reason
Guidelines nobody readsthe system decays within monthsshort, example-led, with templates
Letting the model invent prooffalse claims damage trust and can break advertising lawsupply proof; ask it to flag claims needing evidence

References