I generated a brand for zero rubles. That's what went wrong
Main chat
A chat for vibe coders: news, guides, live cases, marketplace, and finding executors.
ный 16 min reading · AI branding · generative design · startup
In February 2025, I launched a small product. The deadlines were burning, there was no branding budget, and I decided: I will deal with it myself using AI. Midjourney for images, ChatGPT for naming and texts, Looka for logo, Canva for templates. Eight hours of work and I had a brand.
Three months later, I was doing everything from scratch.
Not because it was ugly. The logo was quite decent. The palette is harmonious. The font is appropriate. The problem was different: I had a set of visual elements, but no brand. Every time I had to create something new, I started from scratch because there was no system. Every time I showed them, people couldn’t remember “what we look like.” And most importantly, the brand was no different.
The story is not unique. AI tools have created the illusion that branding has become affordable and fast. It became more accessible. But “generate visual” and “build a brand” are different tasks.
What AI does well in branding
Honestly, a lot. To list “AI is bad” is not true. AI tools have radically changed what is possible for small businesses and startups without a budget.
Generation of options in the research phase. Previously, a mudboard was a watch in Pinterest and Google Images. Midjourney now generates 20 mood choices in 10 minutes. This is not a final design, this is a study of directions. Used so – AI speeds up the process at times.
** Quick drafts to test the concept. ** Need to check if the idea works? The AI logo in 15 minutes allows you to show the concept to investors or partners before spending money on professional design.
**Creating basic templates.**Social media templates, presentations, documents – Canva and analogs do this well. If you already have a system (colors, fonts, principles) – AI templates are an effective way to scale it without a designer on every task.
Copywriting and naming. AI helps generate variants of names, slogans, descriptions. Not all are good — but sorting 50 options is faster than coming up with 10 from scratch.
**Iterations. ** Showed the client not to like it. Try another option. With AI, it's minutes, not days.
Where AI Branding Systemically Fails
Problem 1: AI optimizes for “like” rather than “works”
Tools like Looka, Brandmark, Hatchful are trained on what users choose. Users choose what looks “professional” and “beautiful” right now. This creates a systematic craving for the average.
The AI logo would be good. It will be balanced, harmonious, technically competent. It won’t be unique – because uniqueness is a deviation from the norm, and AI is optimized against deviation.
In three years’ time, most of the 2024 AI logos will look the same – as all stock photos from 2010 look the same now. These aren't bad logos. These are the template logos of the era.
Problem 2: No strategy, no system
AI generates elements. Not the system. Not a strategy. I don’t understand why this color, this typeface, this form.
When there is no strategy, any new brand application becomes a separate solution. You need a banner, you know what to do. You need to pack, you know what to do. You need a presentation - think about what to do. And that's what I've experienced: every new challenge requires a start from scratch.
A brand is a decision-making system. AI does not create that system. It creates the result of one particular decision.
Issue 3: Rights and uniqueness
This is a real legal problem that many people ignore.
AI generators are trained on existing designs. There is no guarantee that the generated logo is not confusingly similar to an existing registered mark. Several documented cases: Brands that launched AI logos received trademark infringement claims.
Before using the AI logo as the main brand, you need to check the databases of trademarks. In Russia: Rospatent and FIPS base. WIPO Global Brand Database. This is not an optional move.
In most jurisdictions, AI-generated works are not automatically copyrighted. That means a competitor can use a similar AI logo and you can’t challenge it.
Problem 4: AI doesn’t know your audience
Midjourney doesn't know that your clients are 45-year-old engineers who trust a rigorous visual, not a startup aesthetic. GPT doesn't know that the word "innovative" in your niche is sceptical because it's been abused by competitors.
AI works with common patterns. A brand works in a specific context – a specific audience, a specific market, a specific moment. You know that, not a model.
How to use AI in branding correctly
Model: AI as a tool, man as author
The difference between “AI made me a brand” and “I used AI to build a brand” is fundamental. In the second case, AI speeds up work. In the first, it replaces thinking.
Correct sequence:
**1. Strategy - without AI (or with AI as an interlocutor, not a generator) **
- Who is the audience and what is important to them
- How we are different from our competitors
- What kind of brand we build
- What a person should feel at every contact with a brand
This is not a job for an AI generator. It’s a job you have to do yourself — you can talk to Claude, you can use ChatGPT as a conversationalist, but the answers have to come from real knowledge of your market.
2. Research - AI is useful
- Generate mudboards through Midjourney by specific characteristic n
- Analyze competitors through search and synthesis
- Collect naming options for manual selection
3. Concept - AI as a draft
- Quick logo in Looka to check direction, not as final
- Several color palettes to evaluate which is closer to the character
- Draft texts for tone assessment
4. Finalization - with an understanding of the system
- Or the designer who builds the system
- Or yourself, but with clearly defined principles that you can apply consistently
What exactly did I do
After three months with the AI brand, I went through the following:
I have written 200 words about who our customers are and what matters to them. It took two days - because I hadn't thought about it specifically enough before.
He defined three principles of visual language: not colors, not fonts, principles. No decoration for the sake of decoration. "Every element makes sense." “We look like a tool, not a window.”.
Only then did he choose a color (the other one didn’t match the principles), a font (the other one was too “startup”), and created three templates that matched the principles.
Now any new element is tested according to principles. It takes minutes, not hours.
Checklist: is the AI-brand ready to launch
- Check for similarity to registered trademarks
- Principles of visual language (not just elements)
- Each element can be explained through brand character and audience
- Logo checked in all formats (favicon, billboard, monochrome, embroidery)
- Color combinations tested by contrast (WCAG AA)
- Fonts licensed for commercial use
- Create templates for key content types (not just elements)
- The principle of making new visual decisions is described
If at least three points are not met, it is not a brand, it is a set of elements.
AI and Branding: How to Use Claude
Prompt: Formulate strategy prior to visual generation
Help me formulate a brand strategy before creating a visual language.
On business:
- What we sell: [product/service description]
Audience (as specific as possible): [age, profession, situation when they are looking for us]
The main problem we are solving is [specifically]
Why we are better or different than our competitors:
- Competitors: [3-5 straight]
- What we don't want people to think of us:
Help me determine:
1. Positioning in one sentence ("we are X for Y who want Z")
2. Three Brand Characteristics You Should Feel in Every Contact
3. Two or three visual directions that correspond to this character (not specific colors – directions)
4. What I need to test on a real audience before going into designPrompt: Check out the AI-generated brand
Help me evaluate the AI-generated brand before launching.
What's generated:
- Logo: [Description]
- Colors: [HEX or description]
- Fonts: [title]
- Tagline/positioning: [text]
Business context:
- What we sell: [Description]
- Who: [Audience]
Direct competitors and how they look: [description]
Estimate:
1. Whether the visual corresponds to the declared nature of the business
2. Does it stand out from competitors or is lost in their background?
3. What risks (legal, reputational, strategic) need to be checked
4. What's missing to make it a system, not a set of elements?
5. Three things to do before launchPrompt: Write a brief for Midjourney/Stable Diffusion
Help create a prompt for generating visual references in Midjourney.
Character of the brand: [three words]
Audience: [who]
Category: [market/industry]
What I want to explore: [mood/color gamut/art style/photo style/general aesthetic]
What I don't want: [anti-examples]
Composition:
1. 5 prompts for Midjourney with different angles of study
2. Keywords to add for each direction
3. Negative Prompts (which to exclude) for more accurate results
4. How to select results – what to look at when evaluatingPrompt: Create Principles from AI Results
I have some AI-generated results that I like. Help me turn them into principles.
I selected the following:
- References that you like: [description - what they have in common: colors, shapes, mood, style]
- What I like about the selected logo options: [description]
- What I like about the chosen colors: [description of sensations]
What I don't like about the rest:
- [Description of rejected options and why]
Help me:
1. Extract 3-5 Principles from What I Choose (Why This Is What Is Common)
2. Formulate rules for the application of each principle
3. How a new designer will be able to determine whether a new element meets these principles
4. Identify potential contradictions between principles and how to resolve them
