Why Generic Image AI Fails at POP Displays

A designer at a display manufacturer showed me a folder of AI renders this summer. Twelve images of what was meant to be a cardboard floor display for a beverage brand. All twelve looked great at thumbnail size and fell apart the second you looked at them as something a factory has to build. Shelves floating with no side panel carrying them. A header card in a proportion no die line would ever cut. The brand's logo, uploaded as a reference, rendered as a plausible-looking word that was not the brand name.

She hadn't done anything wrong. She'd written a careful prompt, attached references, and iterated. The tool did what it's built to do, and what it's built to do is not this.

I've spent close to four years in point-of-purchase and I build AI POP Displays. General image models are not bad at images. They fail at this particular job, in a way that more prompting doesn't fix, and what I want to explain is why, and what has to sit on top of them before a display comes out the other end.

What is a general image model actually doing?

This is about the general image models, the ones in ChatGPT, Midjourney and Gemini. The criticism below is not the tired one about AI images looking obviously fake. They don't any more.

These models learned from an enormous amount of imagery, retail photography included. They know what a point-of-purchase display looks like in a photo and almost nothing about how one is constructed, because construction isn't visible in the training data. Board thickness, fold lines, joints, load paths, print area, pallet footprint. None of that appears in a JPEG of a supermarket aisle, so none of it gets learned.

What the model optimizes for is an image a person accepts as real. That's a different target than an object that resolves as a structure, and the gap between the two is exactly where display work lives.

The consequence that matters most is that these models have no failure state. Nothing ever comes back saying the geometry doesn't close. A picture comes back, at the same confidence, whether it's correct or nonsense, and the errors are the ones that read fine on screen and cost money in a quote.

Why does the geometry come back unbuildable?

This is the failure people notice last, because it hides well at small sizes. Zoom in and you start finding shelves cantilevered off nothing, side panels three millimeters thick at the top and twenty at the bottom, a base narrower than the mass sitting on it, and corrugated structures with no fold, no tab, no visible way of going from flat sheet to standing unit.

None of that bothers the model, because photographically it's fine. Light behaves, shadows land, materials read. It only breaks when someone at the manufacturer has to work out how to make it, redesigns the structure, and the render turns out to have been a mood image rather than a concept.

That distinction is the whole game. A concept is something a client approves and a factory quotes. A mood image gets a nod in a meeting and then gets rebuilt from scratch.

Why are the format proportions wrong?

Point-of-purchase runs on named formats, and each name carries real dimensional constraints. An FSDU has to sit on a pallet footprint and survive shipping. An endcap has to fit the gondola end of a specific retailer. A counter unit has to leave the cashier room to work. A glorifier has to hold one hero product at eye height without dominating the shelf around it.

A general image model treats those words as style, not geometry. Ask for an FSDU and you often get the visual language of a floor display at proportions no floor display has ever had. Ask for a counter unit and it may come back at chest height. There's rarely a scale cue in the frame, so nobody can tell until someone puts a dimension on it.

This is the one that quietly wastes the most time, because the image passes internal review, goes to the customer, gets loved, and then dies when the drawing comes back and it doesn't fit the space it was sold into.

Why does the brand logo come back warped?

Text rendering in image models has improved enormously, and people conclude from that that brand assets are safe. They aren't, because a brand mark isn't a text problem. It's an exact-reproduction problem, a specific mark, in a specific color, at an angle across a folded or curved surface, at whatever resolution that corner of the image happens to get.

The vendors say as much themselves. OpenAI's image generation guide lists among its limitations that the model can still struggle with precise text placement and clarity, and that it may occasionally struggle to maintain visual consistency for brand elements across multiple generations. That is a precise description of a logo that has to come back the same on a curved pack front or an angled header card.

So the model redraws the mark approximately. Letterforms shift, the kerning goes soft, a symbol grows an extra stroke, the color lands near the brand color rather than on it. A logo that's almost right is worse than no logo at all. It's the one thing a brand manager is trained to catch on sight, and it turns a concept review into a conversation about your process.

Why is the shelf count fantasy?

A display exists to hold product, so the product load is the spec, not decoration. This is where general models drift furthest, because they place as many facings as look good in the composition rather than as many as physically fit.

You get nine facings across a shelf that takes five. You get packs floating a centimeter clear of the surface. You get large and small SKUs rendered at the same apparent size, because the model is balancing a picture rather than merchandising a fixture. Sometimes the product photo you attached is ignored and a generic stand-in gets drawn.

Downstream, that's the expensive one. If the concept implies a load the unit can't carry, either the sale rested on a number that doesn't exist, or the structure grows to reach it and the price grows with it.

If there's a real pack and a real shelf count on your desk right now, I'd rather you run that brief than take my examples on trust.

Can't you just prompt harder?

Not where it counts. With enough references and patience, a general model will sometimes hand back one display image that passes review.

The trouble arrives with the second image. There's no persistent object underneath the render, only a distribution of plausible pictures that your prompt is steering. So when the client asks for the header twenty percent taller and the material changed to acrylic, you regenerate, and the shelf spacing moves, the base changes, the product load reshuffles and the lighting shifts. Fix that, and something else drifts. Every revision round becomes a fresh negotiation with the model instead of an edit to a design.

That's why the cost of general models here isn't the first render. It's rounds two through six, where the hours in this business have always gone.

Where a general model still belongs

Before the brief exists. Mood boards, exploring a visual territory, color and material atmosphere, environment plates, and the "show me twenty directions" session at the start of a pitch. Their looseness suits that stage, because nobody has committed to anything yet. Until the brief lands, the image communicates a feeling. From the brief onward it communicates an object that someone will quote, engineer and build, and that takes a tool that knows what the object is.

What changes with a purpose-built tool

AI POP Displays runs on Google's Nano Banana Pro. Nothing in the paragraphs above gets fixed by better pixels, because none of it was a pixel problem. What changes is what sits on top of the engine.

The failures stop being prompt guesswork and start being inputs. Display format comes from a real taxonomy with proportions attached, so the word FSDU carries geometry rather than vibes. Materials are a choice from a list that behaves, not an adjective. Product load is built from the packs you uploaded. Retail context is a background you select, a supermarket aisle or a pharmacy counter rather than a studio sweep. References travel labeled by role, so the product photo, the brand mark and the structure to follow each get used for what they are. And the briefing step reads the customer's email, so the format, the retailer and the facings come out of what they wrote rather than what you remembered to type.

Every one of those is the work that decides a display, done once, inside the tool, instead of supplied again on every brief and every revision round. That encoding is what four years in this industry bought.

A concept render is not a production file. Structural CAD, dielines, board grades and print specs still happen at the manufacturer, and the render's job is to be the brief that work starts from.

The quickest way to check any of this is to put your last brief through it and set the result next to the render you already have.

The specific head-to-heads are separate, against ChatGPT and against Midjourney, and there's a wider map of the tools people use for this if you're deciding what belongs in your stack.

What we're building next: Bellto

Everything above ends at the concept render, and so does every tool in this comparison. The structure still gets drawn at the manufacturer. Bellto is what we're building to close that gap, the only entrant here that takes a campaign to a manufacturable package, an AI agent for brand and shopper marketing teams and their agencies that goes from brief to shop drawings.

You tell it what you're launching (brand, product, channel, stores, budget, date) and it asks for what's missing. It proposes the campaign mix and materials, floor stand, glorifier, shelf strip, stopper, with the budget split per store. It designs every piece with you at real scale in parametric 3D, proportions derived from the product dimensions and the facings, which is exactly where the general models above come apart. Change a dimension and the whole piece recomputes and re-runs its manufacturing checks in seconds. Every approved piece comes out as a 3D model, part-by-part cutlist, dimensioned drawings, STEP of the assembly and a cut DXF per part, a package a workshop can quote without redrawing, and it suggests manufacturers that fit by material, format and volume. The quote still comes from the manufacturer.

We're building it now. The Bellto waitlist is open to any brand or agency, and we're contacting the first ones soon to run the first real campaigns end to end. It opens by invitation, in small groups. Pricing goes first to the people on the list. No card. POP manufacturers can sign up too.

The test that matters

One question sorts good display renders from pretty ones. Could your manufacturer quote from this without redesigning the structure?

Run it on your last AI render. Every shelf lands on something that carries it, or it doesn't. The proportions match the format, or they don't. The facings match what fits, or they don't. None of that is subjective.

If the brief on your desk this week is a real display, signup is free with 15 one-time credits and no card required. Pro is $49 a month for 150 concept generations at the founding price, with the $69 list price stated openly. Nothing you upload or render trains AI models, on any plan, because most of this work sits under NDA and that shouldn't be a paid feature. And if what's on your desk is a whole campaign rather than one display, Bellto is where it goes next.