Most brand concepts come out generic because the input was generic. This prompt makes Claude interrogate you before it decides anything, which is what a studio does on the kickoff call.
You are acting as the strategy lead on a small brand studio. I am going to describe my product in one paragraph, and your first job is to turn it into a proper brief, not to start designing.
Here is the product: [one paragraph, plain language, what it is, who it is for, what it costs, and what makes it different].
Ask me the six questions you most need answered before any naming or visual work would be sound. Ask them one at a time and wait for my answer, and go after the things I have probably not decided yet rather than the easy ones. If an answer is vague, say so and ask again in a narrower way.
When we are done, write the brief back to me in under 400 words with these headings: what we are making, who it is for, the one job it does for them, what we are deliberately not doing, and the three words the brand should feel like. Flag any assumption you made that I never actually confirmed.Positioning first, because the palette is downstream of it. A calm, premium palette on a brand positioned as loud and cheap is going to feel wrong no matter how pretty the hex codes are.
Using the brief we agreed, write the strategy layer for this brand.
Start with the audience: two short profiles of the people most likely to buy, each one covering what they currently do instead of buying this, what they are afraid of getting wrong, and the sentence that would make them stop scrolling. Keep them grounded in the brief rather than inventing demographics.
Then the positioning: one sentence in the form "for [audience] who [situation], [brand] is the [category] that [benefit], unlike [the obvious alternative]". Give me three versions of that sentence taking genuinely different angles, and tell me which one you would defend and why.
Then three name options. For each, give me the name, what it means or where it came from, how it sounds said out loud, the tagline that would sit under it, and one honest risk (hard to spell, already crowded, ages badly). Avoid invented words made by dropping vowels, and avoid anything that needs a paragraph of explanation to make sense.The trick in this one is forcing specifics. Ask for "a nice palette" and you get adjectives. Ask for hex values with defined roles and contrast ratios and you get something a developer or designer can build with tomorrow.
Now the visual identity for the direction we picked. Give me exact values I could hand to a developer, not descriptions.
Palette: six colours with hex codes and a defined role for each (page background, surface, primary text, secondary text, primary accent for buttons and emphasis, and one supporting accent). For every text and background pair I would realistically use, give me the contrast ratio and say whether it passes WCAG AA at normal body size. Explain in one line each why these colours suit the positioning, and include a hover or pressed variant for the accent.
Type: three font pairings, all of them fonts that exist on Google Fonts so I can actually use them for free. For each pairing give me the heading font, the body font, the weights to use, and a type scale with pixel sizes and line heights from h1 down to small print. Say what each pairing signals, and which one you would pick.
Logo: three directions I can build without an illustrator, described precisely enough to draw. Wordmarks, monograms and simple geometric marks only. For each, tell me the construction (what shape, what proportions, what the negative space does), how it behaves at 16 pixels for a favicon, and what it looks like in one colour on a dark background.This is the step that makes it look expensive. Instead of a chat full of hex codes, Claude writes a single page that displays the identity using the identity, so you are looking at the real fonts and the real colours rather than reading about them.
Build the whole identity as a single self-contained HTML artifact called the brand board. One file, no external dependencies except the Google Fonts link, and it must render the brand using the brand.
Lay it out top to bottom as: the logo lockup large on the brand background, the name and tagline, the positioning sentence, the palette as labelled swatches showing each hex value and its role, the type scale rendered live in the actual fonts with a line of real sample copy at each size, the three logo directions drawn as inline SVG (primary, one colour, and a 16 pixel favicon version), a short do and do-not list for using the mark, and one example UI card built from the palette so I can see the system working.
Make it print and screenshot friendly at 1200 pixels wide, keep generous spacing, and add nothing decorative that is not part of the identity. Use the exact hex values and font names we agreed, with no substitutions. If any font fails to load, fall back to a system stack rather than silently changing the design.What should be on the finished board, so you can check the output rather than hoping:
One-shot prompts give you the average brand. The version that looks like someone was paid for it comes from generating options and combining them on purpose.
Before we finalise anything, give me three genuinely different creative directions from the same brief, and make them differ in strategy rather than only in colour.
Name each one, and for each give me the core idea in a sentence, the palette, the type pairing, the logo approach, and the kind of customer it would attract most. Push them apart deliberately: one warm and human, one precise and technical, one bold and loud. If two of them start converging, redo the closest one further from the others.
I will tell you which parts of each I want. Then merge my picks into one coherent direction rather than stapling them together, resolve anything that conflicts, tell me explicitly what you had to change to make the combination work, and rebuild the brand board artifact with the merged result.