code skøn

I didn’t get this idea from a trend report or a conference slide.

It started, as many things do in beauty, over coffee with my dear friend Julie in Copenhagen – someone I’ve shared desks, spreadsheets and far too many late‑night debrief calls with.

Julie and I can stay on the phone for hours, picking apart what’s changing in our jobs: formulas, budgets, creators, the strange new romance between beauty and data.

Somewhere between “how’s your latest project going?” and “have you seen what TikTok is doing now?”, we landed on a sentence that stuck with me: “If the AI can’t read your serum, it doesn’t exist.”
Not in the literal sense, of course. The bottle is there. The texture is there. The people who already love it are very real.

But in a world where more and more shoppers start with a question in a search bar or a chat window a product only counts if the machine can understand it well enough to recommend it.

That’s what this piece is about.

Not just the technical layer, but the silent metrics behind: who gets seen, who disappears, and what it means for beauty brands when “good data” starts to matter almost as much as a good formula.

Can machines “read” a serum?

Before this turns into a tech seminar, let’s stay concrete.
When Julie says “if the AI can’t read your serum”, she isn’t thinking about robots sniffing your INCI list. She’s talking about something much less glamorous: how your product information is stored and labelled behind the scenes.

On most beauty sites, there are two layers:

• The layer you design for humans: photos, storytelling, texture shots, that perfect line about “cloud‑like gel that calms angry skin”.

• And the layer you (sometimes) design for machines: the fields where you tell search engines and AI “this is a serum, it contains niacinamide, it’s for redness and sensitivity, it costs 29€, it’s in stock, people gave it 4.7 out of 5”.

That second layer is what people mean when they talk about structured data or schema. It’s a way of turning your product page into a list of clear facts a machine can understand and trust: category, ingredients, concern, price, reviews, FAQs.

Now imagine two almost‑identical serums:

• Brand A has a beautiful page but no proper schema. The ingredients live in an image, “for who” is hidden in a paragraph, reviews are there but not marked up in a way AI recognises.

• Brand B has filled its Product, Review and FAQ schema properly. For the machine, it’s instantly legible as “fragrance‑free niacinamide serum, under 30€, suitable for redness‑prone sensitive skin, 4.7 stars, in stock”.

When someone asks an assistant for “a gentle niacinamide serum for redness under 35€”, the AI isn’t “discovering” anything magical.

It’s pulling from the products it can actually parse – which, in this case, means Brand B is far more likely to appear in that invisible, digital shelf than Brand A.

Indie vs. giants: who the machine sees first.

Once you see how this works, the next question is obvious: who actually benefits from being “readable” – and who quietly disappears?

The intuitive answer is “big brands win, small brands lose”, but reality is a bit more twisted.

Large groups still treat product data almost like a second formula. They have teams and tools to keep every shade name, ingredient list, skin concern, price and review aligned across their own site, Sephora, Douglas, marketplaces and Google.

Most indie brands don’t live in that world. Their “database” is usually a mix of Shopify fields, spreadsheets, maybe a Notion table and whatever copy made it into the last launch.

And yet, some of the most agile players inside the algorithm are indie.

Because they move faster, some founder‑led brands treat their back‑end like a test lab: clear ingredient lists, simple “for this skin / for this concern” labels, names that work on TikTok, and a very direct link between what people say online and what appears on their product pages.

Still, when AI assistants and new search features scan the landscape at scale, gravity pulls towards what’s easiest to parse and trust:
• Pages with complete Product/Review/FAQ schema.
• Retailers and brands that appear consistently across many sources.
• Listings hosted by big marketplaces that already dominate traditional SEO and link signals.

In the metric world of AI‑driven results for beauty and personal care, that bias shows up clearly: large marketplaces and major retailers capture a disproportionate share of mentions, while a small cluster of highly optimised indie brands manage to punch above their weight – and the rest risk to stay invisible.

The violence isn’t that no indie can win; it’s that only the ones who speak “machine” fluently are invited onto the shelf.

The ethical knot of AI shopping.

Once machines start deciding which products are worth surfacing, we’re no longer just talking about tech – we’re talking about taste, power and responsibility.

The first tension is how commercial all this is about to become.

OpenAI has already rolled out shopping features inside ChatGPT, where the assistant can suggest products and even take users to instant checkout for certain items.

Users have also spotted “shop at Target” and “connect Peloton” prompts inside answers, which sparked a public debate: were these the first ads, or just very ad‑like shopping features?

OpenAI publicly framed them as shopping integrations rather than paid advertising, then later disabled some of these promotional messages after criticism that they looked too much like ads.

Parallel to this, marketing and SEO circles are already preparing for the next step: clearly labelled “Sponsored” recommendations embedded directly into conversational answers.

Think: you ask “best anti‑ageing routine for sensitive skin” and, somewhere inside an otherwise helpful answer, a specific brand or retailer appears with a small “Sponsored” tag.

For beauty, that means the same interface that feels intimate and advisory could also become a new kind of media buy.

That leads straight to the second tension: can you trust a recommendation that shares a screen with money?

In classic media, we at least know where the lines are supposed to be: editorial vs. advertorial, organic vs. sponsored. In a chat window, the boundaries are softer by design.

If an AI assistant is trained on data where big retailers are already over‑represented, and then starts layering sponsorship on top, how do you tell the difference between “this is genuinely the best fit for your skin” and “this brand paid, and the model already liked it to begin with”?

The third tension is about what quietly gets edited out.

Slow fragrances that need time on skin, treatments for rare conditions, textures that are hard to capture in quick filters or five‑star ratings – everything that doesn’t fit neatly into “sponsored slot + clean data + mainstream demand” risks slipping off the radar.

If AI‑driven visibility and conversational ads become the default way people discover products, the beauty landscape will shrink to what’s easiest to sell and safest to sponsor.

Humans are built for adaptation.

If you’ve made it this far, you probably understand why my phone calls with Julie last hours and why our coffee dates quietly turn into brunch. Once you start pulling on the thread of “who gets to exist on the invisible shelf?”, it’s very hard to stop.

CodeSkøn is not outside this system either.

While I write about it, I’m also working to put this magazine on that same “media shelf” alongside more established titles – cleaning up categories, coding things properly, reviewing the invisible structure one small tweak at a time.

And like it or not, you’re part of this inventory simply by scrolling, searching and tapping on your phone.

The question – for brands, editors and readers – is how much of that definition we’re willing to hand over, and how much we still want to decide for ourselves.