7 mistakes in AI design, because of which you will not be hired

## Mistake 1. Portfolio from AI screens without explanation of solutions

The problem is not that the designer used AI. The problem is that the case does not tell whether he was managing the process or simply accepting what was generated.

The two cases look equally beautiful. In one, the designer explains, “Stitch suggested the cards, but I made the list — because the user scans the content line by line, rather than selecting from the set.” The other is just the final screens.

The first is thinking. The second shows the ability to press the generation button.

AI has made the entry threshold for a “beautiful result” zero. Beauty is no longer a signal of level. The signal was justification.

How to fix it. ** For each key decision in the case, add one sentence: why so and not otherwise. Even if the decision came from generation, explain why you made it.

## Mistake 2. Generic design: took the first result and did not edit

There is a recognizable visual pattern: rounded cards with shadows, gradient buttons of the blue-violet spectrum, illustrations in the style of 3D-blob, a large centered Hero. It’s not a bad design – it’s the model’s default response to make it modern.

When such a screen appears in a portfolio unchanged, an experienced designer or lead reads it unequivocally: the person did not make visual decisions. He accepted the result.

AI generates by probability – takes the most frequent decisions from training data. That's why without a referral, he always gives a hospital average. The task of the designer is to set and then edit the result for a specific context.

How to fix it. Before generation, write a design brief: tone, audience, what you want to leave visually. After that, be sure to edit with your hands. If the result screen is indistinguishable from the first result of the generation, it is a signal that you have not yet worked.

## Mistake 3. Prompt as TK - and no iteration

Beginner AI designers often work like this: they wrote a prompt, got the result, slightly tweaked in Figma – ready. One cycle.

This is evident from the result: the screen is technically correct, but does not solve the problem precisely. Almost the wrong hierarchy, the accents are wrong, the structure is formally correct, but uncomfortable.

A good workflow with AI is iterations. The first result is structure. The second prompt clarifies the problem areas. Three, details. In parallel, there is an editing at Figma. It's not a asked-to-receive dialogue, it's a dialogue where the designer leads.

A hiring manager who works with AI himself understands this. And if he sees the result of one cycle in the case, he understands that the person has not mastered the workflow.

How to fix. Show in the case not only the final, but also the evolution: the first result → what did not suit → as specified → the result. This demonstrates that you control the tool, not the other way around.

## Mistake 4. Blind trust in AI in UX solutions

AI knows visual patterns well. They don’t know your specific user well.

A typical situation: the designer asks to generate onboarding, gets a five-step float with a greeting, a choice of goals, a tutorial - and leaves it to work. Looks logical. But no one checked whether onboarding is necessary at all, whether it is easier to let the user immediately into the product.

AI doesn't ask why. He solves the problem as it is formulated. If you ask “do onboarding” – he will do onboarding, not ask if he needs it.

A designer who accepts a UX structure from AI without critical evaluation shifts responsibility for decisions to the model. This is evident in the case: there is a structure, there is no justification.

How to fix it. AI is a tool to implement a decision, not to make it. First, answer yourself: what task the user solves this screen, which path is optimal for him. Then generate.

## Mistake 5. Not Noticing Generation Artifacts

Each AI tool has characteristic artifacts: slightly uneven indentations, icons from different visual languages in one interface, text stub that does not fit into real data, buttons with the same visual weight at different importance.

An experienced designer sees these things and fixes them. The beginner does not notice, because in general "looks normal.".

When such a screen enters the portfolio, the lead on the review sees the artifacts immediately - he works with these tools himself. And this is not a question of “bad design”, it is a question of discretion and attention: a person does not notice what catches the eye of a professional.

How to fix. After generation, go through the checklist: a single visual language of icons, indentations are multiples of 4 or 8px, the text is checked on real data (long name, empty field, error). Everything that gets knocked out is corrected by hands.

## Mistake 6. Use AI only to generate screens

It's a different kind of mistake -- not "doing wrong," but "using incompletely.".

Many designers have mastered interface generation, but do not use AI where it gives even more: task formulation, analysis of user scripts, writing UX texts, checking the logic of flow, generating variants for A/B.

In the interview, you can hear this: a person tells you how he used Figma Make for screens, but did all the other parts of the work the old way. This is a signal that the workflow is not built systematically.

Companies in 2026 are looking for designers who have integrated AI into the entire process, not just one operation.

How to fix it. Try to run each stage of your workshop through AI at least once: task statement, scripts, UX texts, hierarchy check. Find where it really speeds up — and leave it in the process.

## Mistake 7. Not being able to explain where AI is and where you are

In interviews in 2025-2026, this question is asked more and more often: “Show the case – what did AI do here, and what did you do?”

Designers who aren’t ready for this are lost. Either they begin to downplay the role of AI (“well, I used a little”), or vice versa – they can’t explain their contribution.

The correct answer goes something like this: “AI generated the screen structure and the first variant of the components.” I redesigned the hierarchy, replaced the typography with our design system, removed the two blocks that duplicated each other, and added error states – their AI did not cover. This shows that the person consciously worked with the instrument.

Trying to hide the use of AI, or failing to distinguish between contributions, both look equally bad.

Keep short notes while working: what came from generation, what was remade, what was added yourself. This will take 5 minutes per case and give you a clear answer to the interview.

---

## Main pattern

All seven mistakes are about one thing: AI has changed tools, but it hasn’t changed what a designer is hired for. They are hired for thinking, for decisions, for understanding the user.

A designer who knows how to work with AI and at the same time retains authorship on solutions is exactly the profile they are looking for now. I don’t know how to generate, but I can think and use AI to do it faster.

This is the difference between the candidate who is given a test and the one who immediately makes an offer.