Using AI to Design Your Engagement Ring

23 June 2026 · By Johan Poggenpoel · 1 min read

Times have changed.

Again.

Couples have always arrived at our studios with ideas in whatever form they could manage — hand-sketched drawings, a screenshot from Pinterest, a photo of a ring spotted on a stranger's hand.

Since 80% of our work is custom… we've always loved this part. It tells us something about how a person sees beauty, and it gives us something to work with.

Now, increasingly, they arrive with AI-generated images as part of the brief. And honestly? We're not complaining.

AI is a genuinely powerful brainstorming tool. It's fast, it's visual, and it lets you explore ideas at a pace that no design consultation ever could. If you've spent an evening with ChatGPT or Midjourney pulling ring concepts together, you've probably landed somewhere interesting — something that feels close to what you've been trying to articulate. That's real value, and we want to help you bring it to life.

Looking beautiful in a render and wearing beautifully on a hand are two very different standards.

And since more and more couples are walking in with AI designs in hand, we thought it was worth being honest about where that gap tends to show up.

The core problem: AI has seen the pictures. It doesn't know why they look the way they do.

Engagement ring design isn't an open frontier. I’d wager a guess there are roughly two thousand genuinely distinct ring designs in the world — and that's not a limitation of imagination, it's the result of decades of refinement.

Every curve, every proportion, every claw placement has been stress-tested against the reality of daily wear. What looks beautiful in a well-made ring is usually inseparable from what makes it durable and buildable.

AI samples from this refined catalogue. It blends, morphs, and recombines. And in doing so, it frequently steps over the very boundaries that made those original designs work. It pulls the flavour without understanding the recipe.

The result is images that look stunning — and sometimes describe rings that can't be made. Or at least, can't be made well.

Here's where it tends to go wrong:

The stones

AI hallucinates diamond shapes that don't exist. It renders stones at sizes below what's commercially available. More critically, it regularly places diamonds in positions where there simply isn't enough metal to hold them securely for everyday wear.

When we assess an AI-generated design, the first questions are always: do these stone shapes actually exist? Are the sizes practical? And is there enough metal structure around each stone to keep it safe for a lifetime of wear?

These aren't stylistic considerations — they're non-negotiable.

The silhouette

AI will often generate what it presents as a single ring from multiple angles — and when you look carefully, the top view, the side view, and the profile view describe three completely different rings. The AI has rendered each angle independently rather than as views of the same coherent object.

A beautiful image isn't always a coherent design.

Before anything else, a ring needs to make structural sense as a three-dimensional object, with minimum thresholds of metal thickness and width that allow it to survive in the real world.

The detail

This is where vintage-inspired designs tend to suffer most. AI draws on scrollwork, filigree, and ornamentation from architecture, embroidery, and decorative arts — references that don't always translate into the compressed space of a ring shank.

The detail in AI-generated vintage rings is frequently so fine that it creates two problems. First, it's impossible to manufacture to the required standard because the tools fine enough to work at that scale don't exist. Second, ring-making is a reductive process — material is removed as you work — which means overly fine detail gets finer still. What looked delicate in the image becomes structurally compromised in metal.

There's a real ceiling on how much ornamental detail can exist in a given area of a ring and still be finished, polished, and worn daily without deteriorating.

The metal colours

Rose gold is a good example. Certain warm, dusty, peachy hues that AI renders simply don't exist in real alloys. Falling in love with a specific gold colour in an AI image before you've seen the actual metal options is a setup for disappointment.

Come and see the colours in person — the real range is beautiful, and you'll choose better for having held it.

Why this keeps happening

It's worth understanding that this isn't a problem that's getting better on its own. AI systems learn from the images they're trained on, and as AI-generated ring designs proliferate across the web, those impractical images are being fed back into the training data. The gap between what AI generates and what can actually be made isn't narrowing — in some respects it's widening.

There's also no feedback loop from the workshop. When a couple downloads an AI image they love, the system registers that as a success. What happens when that design meets a goldsmith — the compromises made, the elements that had to be reworked or abandoned — never makes it back into the model. The AI keeps confidently generating the same categories of problems.

So where does this leave you?

In a better position than you might think.

You have a tool that can help you explore and articulate your vision faster than ever before. That's genuinely useful. Bring us your AI images — we want to see them. They tell us what you're drawn to, what aesthetic language you're working in, what matters to you.

What we bring is the accumulated knowledge of what makes a ring work. Not just look good in an image, but wear well, hold its stones, take a polish, survive decades of daily life, and still look beautiful on your hand many moons later.

A skilled designer isn't less important in the age of AI. If anything, the translation layer — from generated image to buildable ring — matters more than ever.

Bring us your ideas, in whatever form they've taken. We're here to help make something real.

You can reach my studios here: