Skincare is being reranked by AI shopping agents. Is your catalog readable?

A skincare shopper today rarely starts with your homepage. They open ChatGPT, Gemini, or Perplexity, describe a skin concern and a budget, and get a shortlist back in one turn. Adobe Analytics found AI-driven traffic to U.S. retail sites jumped 805% year-over-year on Black Friday 2025, and those AI-referred shoppers converted 38% more often than shoppers from other channels (Adobe). Beauty and skincare are near the front of that shift, because ingredient and routine questions map almost perfectly onto how these agents work. The catalogs that show up in the answer are the ones an agent can actually read.

What "ask an AI to recommend" looks like now

Picture a shopper with reactive, acne-prone skin typing this into an AI shopping agent: "recommend a fragrance-free moisturizer with ceramides and niacinamide, non-comedogenic, under $40, that won't sting after retinol." That is a real, specific query pattern, and it is exactly the kind of question industry researchers say beauty shoppers ask more than shoppers in almost any other category (BeautyMatter).

To answer it, the agent needs to filter on attributes most product feeds don't carry in a usable form: fragrance status, comedogenicity, named active ingredients and their concentration, and compatibility with a retinoid routine. If that data lives only in a marketing paragraph or a scanned ingredient photo, the agent can't confidently match it, so it skips the product and recommends a competitor whose page states the same facts as clean, queryable fields.

This is also why a small set of brands keeps showing up in these answers. Research from Business of Fashion found ChatGPT recommended La Roche-Posay in a striking share of facial skincare queries (Business of Fashion) — not because it's the only good product on the market, but because its ingredient and use-case information is unusually explicit and consistent across retailers. Consistency is a data problem before it's a marketing problem.

Why thin data makes a catalog invisible

Most skincare feeds were built for a search bar, not a reasoning model. A typical PIM record has a title, a hero image, a price, and a few paragraphs of brand copy. That's enough for a shopper who already knows the product name. It's not enough for an agent trying to decide, among forty moisturizers, which ones are fragrance-free, which are safe with retinol, and which suit oily-acne-prone skin at a given price point.

Three gaps show up constantly in skincare catalogs:

• Skin type and concern fields are missing or use inconsistent free text ("for sensitive skin!" instead of a structured sensitive tag).
• Active ingredients and concentrations sit inside a paragraph instead of a parseable list, so an agent can't confirm "contains niacinamide" versus "mentions niacinamide as a category."
• Fragrance-free, non-comedogenic, and cruelty-free claims aren't tied to any verifiable field, so agents that weight trust signals have nothing structured to check.

None of this is exotic information. It's the same information a knowledgeable store associate would rattle off. The catalog just never wrote it down in a form a machine can act on.

What machine-readable skincare data actually looks like

Here's the same moisturizer before and after enrichment:

The enriched row doesn't add hype. It adds facts an agent can filter on and a shopper can trust. That's the difference between a product that surfaces in a comparison and one that never enters the candidate set.

Underneath the table, this data should also exist as structured markup an agent's crawler can parse without guessing. Industry analysis of AI-cited pages found 65% of pages cited by Google AI Mode and 71% of pages cited by ChatGPT include structured data, with JSON-LD accounting for roughly 90% of that markup because it's cleanly separated from page HTML and doesn't require parsing the visual layout (Alhena). For skincare specifically, that means Product schema carrying brand, GTIN, and price alongside explicit fields for active ingredients, skin-type suitability, and fragrance status, not buried inside a description string.

The mechanism, not the magic

None of this requires chasing every AI platform's algorithm. The mechanism is simpler and more durable: agents recommend what they can verify quickly, and they verify fastest against structured, consistent fields. A catalog with complete skin-type tags, named actives with concentrations, and accurate fragrance and comedogenic flags gives every AI shopping agent, current and future, the same clean surface to reason over. That's a data maintenance problem, not a one-time SEO project, because ingredient reformulations, new claims, and pricing changes happen constantly and the fields drift out of sync with reality if nobody's watching them.

Where Anglera fits

Your PIM stores the data. Anglera does the work of keeping it complete: it scores every skincare listing for missing skin-type tags, unparsed ingredient lists, and unverified fragrance or comedogenic claims, then gap-fills and standardizes those fields so they read the same way to a shopper and to an AI agent. It plugs into whatever PIM or commerce platform you already run, no rip-and-replace, and it keeps re-checking the catalog as ingredients, claims, and prices change so the data doesn't go thin again six months later.