AI can today generate an ad in a few seconds.
An eye-catching image. Perfect lighting. A luxurious interior. Gold lettering. Icons. A slogan. A price. A phone number. A CTA. A few benefits.
You can also add the website address, WhatsApp, an extra message, a logo, another advantage, and information that “absolutely has to be included.”
Technically, everything checks out.
There’s just one problem.
The user did not come to the internet to read the entire ad.
And that is exactly where the interesting problem of modern design begins.
AI made the graphic. But who is reading it now?
Just browse the ads published online to see a certain characteristic pattern.
Large headline. Second headline. Subheading. A few benefits. Price or amount. Phone number. Website address. Logo. Icons. Extra message. CTA.
And because the graphic is supposed to look “rich,” there are also lighting effects, gradients, gold elements, decorative illustrations, an AI-generated image, and several different typography styles.
Everything looks impressive. But at the same time everything is fighting for attention. And if every element is trying to be the most important, visual hierarchy stops working.
This is very important, because visual hierarchy is not decoration. It is a way of guiding the viewer through the message.
It says:
- look at this first,
- then read this,
- the most important information is here,
- you can take action here.
If all elements receive similar weight, the user has to decide for themselves where to look.
And most of the time, they won’t.
They’ll just scroll on.
An ad is not a newspaper
One of the biggest problems with modern ads is the attempt to fit all product knowledge onto a single graphic.
The company has seven selling points, so it shows seven. It has three competitive advantages, so it shows three. It has a phone number, WhatsApp, a website, a logo, a certificate, and extra information, so everything ends up on the graphic.
Then there’s also the slogan. And the photo. And the promotion. And the button.
As a result, you get something that is formally an ad, but visually starts to resemble a notice board.
On a large monitor, it may still look “full.”
On a phone, it often turns into a block of small text that no one will bother enlarging.
The user will not analyze every element.
They will not look for the most important message.
They will not wonder what the author meant.
They’ll scroll past.
The user does not read graphics. The user scans them
This is not just a matter of the designer’s taste.
Designing digital communication is based, among other things, on the fact that the user has limited attention and very often scans content instead of analyzing it element by element.
That is why a good ad graphic should have a clear structure.
What do we see first?
What do we notice a moment later?
What can we skip?
What is supporting information?
What is supposed to prompt action?
Eye-tracking studies on the visual complexity of ads show that more complex banners may require greater cognitive effort. In one study, ads with low visual complexity were observed more often and for longer, and their reception was more favorable than in the case of more complex ads.
This leads to an important distinction: noticing an ad does not yet mean the user has processed its message.
We can make the graphic very visible.
We can use huge text.
We can use contrast, animation, and intense colors.
And yet the user may still not remember what we wanted to tell them.
And that is where banner blindness comes back
The internet has taught us to ignore certain visual patterns.
We refer to this as banner blindness.
Interestingly, it does not literally mean that the user never looks at an ad. Eye-tracking studies show a more complex picture: an ad can be noticed, but its elements may not be effectively processed or remembered. Among other things, this is influenced by the task the user is currently performing.
This is important, because in practice it means that simply “making something that stands out” does not solve the problem.
The user can see the ad and ignore it at the same time.
And modern AI can speed up this problem even more. Why?
Because never before has it been so easy to produce so many similar messages.
When all ads start to look the same
Let’s imagine a few thousand companies.
Each has access to similar generators.
Each can type: “luxury real estate ad, gold elements, sunset, elegant apartment, professional marketing”.
Or: “modern financial ad, gold lettering, documents, phone, premium, professional company”.
Or: “dynamic business ad, successful woman, laptop, charts, lights, modern office”.
The generator will do its job...
An attractive (well, let’s say...) image will be created.
The problem is that after a while attractiveness stops being a differentiator.
If everyone uses a similar aesthetic, that aesthetic stops standing out.
Visual noise emerges.
And the user begins to recognize not a specific brand, but an entire pattern: “another ad.”
And they scroll on.
The biggest problem is not that AI makes ugly graphics
That would be too simple a diagnosis.
AI can generate truly good images.
What is more, a recent Columbia Business School study involving more than 2 million ad observations, over 7,000 advertisers, nearly 50 product categories, more than 16 billion impressions, and 116 million clicks showed that ads with AI-generated images can achieve a higher CTR than ads with images created by humans - provided the generated images do not look too much like AI.
That is a very interesting result.
Because it shows that the problem is not the use of AI itself.
The problem is the lack of control over what AI produces and how that material functions in brand communication.
AI can be a great tool.
But a tool is not yet a design process.
Generation is not design
This is probably the most important sentence in the entire topic.
Generation is not design.
A generator can create an image.
The designer must decide:
- whether this image fits the brand,
- whether it suits the audience,
- whether it communicates the right emotion,
- whether it has the right hierarchy,
- whether the text is legible,
- whether the contrast works,
- whether the CTA is visible,
- whether the amount of information is too large,
- whether the ad doesn’t look like yet another random piece of material generated by AI.
And above all: what can be removed.
That last part is especially important.
Because AI can very easily answer the question: “What else can we add?”
A good designer can answer: “Nothing. It’s enough already.”
A graphic designer is not just a person who makes pictures
That is why it is hard to reduce the future of graphic design to the question: “Will AI take their job?”
If a designer’s task were only to create an aesthetic image, automation could indeed take over a large part of the process.
But a professional designer does much more:
- Builds hierarchy.
- Chooses typography.
- Controls space.
- Understands context.
- Designs the visual system.
- Adapts communication to the brand.
- Knows when a photo helps and when it becomes decoration.
- Knows when strong contrast draws attention and when it starts to look cheap.
- Knows that not every piece of information has to be on the graphic.
- And can give up an element that is attractive on its own, but ruins the whole composition.
This may be one of the most important design competencies in the era of generative AI. The ability to select.
We can have 100 versions. We need one good one
This is another change that AI brings to design.
In the past, the problem could have been creating one good version. Today we can create dozens in a few minutes. But a larger number of variants does not automatically mean a better design. Quite the opposite. If we do not have criteria for choosing, we simply get a larger number of possibilities. And possibilities are not yet decisions.
We can generate:
- 20 headlines,
- 30 photos,
- 15 layouts,
- 10 CTA versions,
- 50 color variants.
But what next?
Who will decide that one variant really communicates the brand best? Who will notice that a beautiful image doesn’t fit the product? Who will conclude that there is too much text? Who will see that the ad is eye-catching, but it’s not clear what it actually sells?
AI can increase the speed of production. It does not relieve humans of responsibility for the result.
“But people do click on such ads”
Of course.
And that is exactly why it is not worth basing this article on the assumption that every ad generated by AI is ineffective.
That is not true.
An ad’s effectiveness depends on many elements: the product, the offer, the audience, the context, the creative, the placement, the message, and the testing method.
Even current research shows that ads created with the help of AI can perform very well.
The problem is elsewhere.
You can get a click and at the same time create terrible brand communication.
You can attract attention and fail to convey meaningful information.
You can make an ad with a high CTR, but one that does not build trust.
You can create something that looks impressive in the feed, but completely does not fit the later user experience.
That is why ads should not be judged solely through the prism of whether “people click”.
Because the user also evaluates the brand
If an ad looks chaotic, exaggerated, generic, or careless, the user may transfer that impression to the company.
Especially in the case of brands that sell services requiring trust;
- Software house.
- Law firm.
- Financial company.
- Developer.
- Real estate agency.
- Medical company.
- Business advisor.
- Technology manufacturer.
Here, the ad is not just a carrier of product information.
It is also a fragment of the brand experience.
If the user’s first contact with the company is a graphic in which the text is microscopic, the information competes with itself, and half of the elements look like generated decoration, it is hard to expect the recipient to automatically think: “This must be a professional company”.
The technology may be advanced. The offer may be excellent. But the communication also has to show that.
AI may therefore create a new kind of blindness
This is probably the most interesting paradox.
Generative AI was supposed to help us create more appealing content. And it can. But at the same time it can lead to a situation where the internet is flooded with visually appealing but similar-looking materials.
And when everything looks impressive, impressiveness stops being enough.
The user starts filtering.
They do not analyze every graphic.
They do not check every slogan.
They do not stop at every “SUPER OFFER”.
They recognize the pattern and move on.
This is very similar to what happens with classic online advertising.
The more ads we see, the more we learn to ignore them.
This is not about going back to the world of “handmade graphics”
There is no point pretending that AI can be rolled back. It doesn’t need to be. AI is a great production tool.
It can speed up work.
It can help with concept development.
It can generate variants.
It can support art direction.
It can help designers with tasks that used to take many hours.
But that is exactly why the importance of the human who can evaluate the result is growing.
The cheaper and faster it becomes to create material, the more valuable the ability to choose the right material becomes.
This may be one of the biggest changes in design.
We will not have a problem with a lack of graphics. We will have a problem with their excess.
A good ad does not have to say everything
A good ad can say one thing. And do it well.
- It can show the product.
- It can evoke emotion.
- It can communicate one advantage.
- It can ask a question.
- It can direct the user to the website.
The rest of the information can be there.
- On the website.
- In the landing page.
- In the shop.
- In the form.
- In a conversation with a sales rep.
There is no obligation to close the entire sales process in one image.
An ad does not have to be an encyclopedia of the product.
Its task is, among other things, to create the right context for the next step.
The best designer of the future?
Perhaps the one who knows best how to say "no".
- No - to another icon.
- No - to another slogan.
- No - to a third CTA.
- No - to the text that "might come in handy."
- No - to the effect that is flashy but communicates nothing.
- No - to the photo that looks great but has nothing to do with the brand.
- No - to another generated variant just because it can be generated.
Because in a world where AI can produce an almost infinite number of versions, the value is no longer in production itself.
The value becomes the decision.
AI will not take jobs from designers. But it will change what they will be responsible for
We probably won't need fewer images. We will need more of them.
We won't need fewer variants. We will have far more of them.
We won't have fewer possibilities. We will have absurdly many of them.
That is why the designer's role may increasingly shift away from production itself toward art direction, UX, visual strategy, communication, selection, and quality control.
AI can generate - A human must decide.
AI can propose - A human must evaluate.
AI can create 100 ads - A human must know why 99 of them should go in the trash.
And perhaps that is exactly why the graphic designer will not disappear. The definition of their work will simply change.
Because the biggest problem of the internet of the future will not be a lack of content. It will be the excess of content that no one wants to look at.
And when every ad starts shouting, the user does something very simple.
They stop listening.
