The 'Two Traps' Beginners Fall Into When Making Flyers with Generative AI: The Barrier Between DTP and Smartphone Screens

The trap of the 'sense of omnipotence' that everyone has become a creator

In the last few years, the evolution of generative AI has been staggering.
By just tweaking the prompt (instruction words) a little, anyone can now 'pop out' high-quality images that look like they were taken by professional illustrators or photographers.

While that in itself is wonderful, this ease of use has brought a very dangerous illusion to some AI beginners.
It is a sense of omnipotence without substance, the belief that 'if I can create images, I should be able to create perfect flyers (printed materials) with just this one AI, including text and layout.'

Recently, an SOS message came to me saying 'It didn't go well when I made a flyer with AI' (I will omit the details as the individual could be identified), (I will omit the details as the individual could be identified), and the reason for the failure was so 'typical of the AI adoption phase' that I had to hold my head in my hands.

There are mainly two traps they are falling into.

Trap 1: The insurmountable wall between smartphone screens and 'physical formats'

First, many of them try to complete everything on a 'smartphone' screen.
Smartphone screens are basically vertically long (16:9 or 20:9) and are premised on the ability to scroll infinitely.
They pack elements tightly into that and export the image, thinking, 'Alright, I've filled the screen! This is a flyer!'.

However, in the real world, there are absolute aspect ratios (silver ratio) such as 'A4' or 'B5'.
What happens if you try to print a vertically long image, like a hanging scroll for a smartphone, onto A4 paper at a convenience store copier?
A major accident occurs where the top and bottom are chopped off, or huge white margins are left on the left and right sides.

In the first place, printed materials (DTP) have absolute physical rules such as resolution (350dpi), color mode (CMYK), and bleed.
Printing an RGB image that looks beautiful on screen as-is and then panicking because 'the colors turned muddy' or 'the edges were cut off' is the result of drowning in the sense of omnipotence of the tool and ignoring the physical format.

Trap 2: The text drawn by AI is not 'language' but 'patterns'

What's even more troublesome is that recent AI has become able to draw 'things that look like text'.

Deceived by this, some people type in long prompts saying, 'I'll have the AI write everything from the catchphrase to the detailed explanatory text'.
What happens as a result?
When you actually enlarge it and look, Japanese characters with many strokes completely undergo gestalt collapse, and unknown hieroglyphs are generated.

For generative AI, text is not 'meaningful language', but merely 'pixels (patterns) in that shape'.

Moreover, because it is output as a single combined image (JPEG or PNG), 'Oh, I want to fix just this one character' or 'the phone number changed' becomes physically impossible to fix just the text.
The maintainability is absolutely zero.

True 'automation' and what we should be creating

In the end, for that SOS, I used the image created by the AI only as 'pure background material', and handed over a PDF that I re-laid out to A4 size using Illustrator.
It was DTP done by human hands, calculating eye guidance and margins.

'It's super convenient to automate with AI!' The people who say this with a smug look are often doing nothing more than 'spinning the AI gacha on their smartphones, holding their heads over mysterious hieroglyphs, and trying to fix them bit by bit by hand', which is extremely laborious and inefficient work.

Paradoxically speaking, if they really want to claim 'automation', I wish they would at least build a system where 'Excel text data is sent via API, automatically poured into a DTP PDF template using Python, and print data is mass-produced without any human intervention' before saying so.

Well, even if you built such a system, all you would end up with is a pile of 'soulless, low-quality spam flyers', so I would never build one.

AI is not a magic wand.
It is 'just a tool' to be used in the right place for the right task.
Instead of getting drunk on a sense of omnipotence and dumping everything on AI, let the AI create the materials, and have humans set the rules (physical formats) and provide direction.
As long as you can't handle that basic division of labor, you will never be able to escape from 'mysterious hieroglyph flyers'.