Industries · Beauty & skincare
AI search visibility for beauty brands.
Beauty is the category where somebody else writes the answer. When a shopper asks an assistant what to use for a damaged barrier or which sunscreen leaves no cast, the recommendation is assembled almost entirely from editorial roundups, retailer listings, health publishers and community threads. One 2026 analysis of 750 beauty answers put brand-owned sites at 2.4% of what the engines actually read.
The second problem is quieter and more fixable. The data that answers a beauty question — shade depth, undertone, finish, the full ingredient list — is usually shipped as a swatch image and a photograph of the carton. It is invisible to a parser, and it is the part you fully control.
Last updated 2026-08-22 · Deadpan Labs · Miami, FL
What buyers actually ask
Almost none of it is a product request.
Two patterns run through this list. Most of these are constraint sets — a skin type, a concern, an ingredient exclusion and a finish, all at once. And several of them span more than one product, which means no product page is even a candidate to answer them.
- “What moisturizer repairs a damaged skin barrier? Everything stings right now.”
- “Can I use niacinamide and retinol together, or do I alternate nights?”
- “Fragrance-free, non-comedogenic moisturizer for rosacea that works under makeup.”
- “Sunscreen for oily acne-prone skin with no white cast on deeper skin tones.”
- “What shade am I in this line if I'm NC25 in another brand?”
A page that says “hydrating moisturizer, 50ml, $32” answers exactly none of them.
Your list is not this list. A panel is built from your concerns, your actives, your exclusions and your actual competitors, which is the first thing the free check produces.
The diagnosis
Six reasons a beauty brand goes unrecommended.
The general reasons are on the pillar page. These six are specific to this category, and the first one changes what a sensible budget looks like more than the other five combined.
Owned media is a rounding error
Your own site is roughly two percent of what gets read.
AIVO's July 2026 analysis of 750 beauty answers across five engines reported that brand-owned sites accounted for 2.4 percent of what the engines actually read — 3,814 citations spread across 610 domains, overwhelmingly editorial, retail and community. Goodie, isolating 135,419 beauty citations in September 2025, found the top ten domains taking 36.7 percent of all beauty citations and noted that not a single traditional beauty brand made that top ten. On-site work still matters, but in this category it has a hard ceiling, and pretending otherwise is how a year of budget disappears.
The category is ingredient-led
Shoppers ask about a molecule; your page talks about a feeling.
Beauty demand is expressed in actives, concentrations, exclusions and skin concerns. A page written in mood, provenance and imagery has almost no lexical overlap with the question. The brands showing up in citation indexes for skincare are consistently the ones whose product pages read like a formulation sheet — the active, the percentage, the vehicle, and what is deliberately not in it.
Roundups own the head term
You do not out-write the roundup. You get into it.
Goodie found that Allure's “Best of Beauty” franchise alone appeared in 73 percent of “best [product category]” queries it measured. When a single editorial franchise is that load-bearing for a query class, publishing your own better article about the same category does not compete with it — the engine is not comparing your article to theirs, it is citing the source it already trusts. This is the failure mode most often misdiagnosed as a content problem.
Routine questions have no owner
Nothing on any product page answers half the list above.
Sequencing, layering, compatibility, purging, pregnancy — these questions span multiple products and often multiple brands, so no product detail page is a candidate to answer them. The engine reaches instead for health publishers, editorial and community threads. This is the largest genuinely unowned surface in the category, and it is one of the few places in beauty where first-party content can win on merit rather than on relationships.
Shade data is structurally unreadable
Google's variant schema was written for apparel, and cosmetics do not fit it.
Google documents exactly six axes a product group may vary by: color, size, suggested age, suggested gender, material and pattern. A shade range technically maps onto color, but most beauty catalogs express shade as a swatch image and undertone as marketing prose, with one URL for the whole family. Depth, undertone and finish — the three things a shade question is actually about — have no dedicated property anywhere. An engine cannot answer “what shade am I here” from a hex swatch, and it will not guess.
Ranking first proves nothing here
Beauty had the lowest overlap between AI citations and organic results of any niche measured.
SE Ranking, across 100,013 keywords in twenty niches, found Fashion and Beauty at the bottom for overlap between the links inside an AI Overview and the top ten organic results — 9.02 percent. That study is from 2024 and we would not quote its trigger rates today, but the structural point has not aged: in this category, being on page one and being in the answer are close to unrelated events. If your reporting only watches rankings, it is watching the wrong thing.
The part you control
Make the catalog answer the question.
You cannot instruct an editor to include you. You can make your catalog readable, and in beauty that is a bigger job than in almost any other category, because the product data that decides a purchase is the data most often shipped as a picture.
NielsenIQ reported in July 2026 that 84% of the consumers it surveyed are more likely to purchase when key product attributes can be easily compared. That is the shopper-side statement of the same problem an answer engine has.
The consequence is that a beauty catalog has to state, in text a parser can retrieve, the things it currently ships as pictures: shade, undertone, finish, and the full ingredient list. Doing that properly is more work than it sounds, because the schema was not designed for this category and the workaround is not obvious — as the failure above says, three of the things a shade question is actually about have no property to live in anywhere.
It is also the fastest win in most beauty catalogs, and the only half of this problem you fully control. The other half is written by editors and strangers on their own schedule.
Measurement
What a panel measures in this category.
A prompt panel is a fixed set of real buyer questions asked across the answer engines on a fixed schedule. In beauty a panel built on your product line measures the wrong thing, because your product line is not what these questions are about — somebody else is writing the answer, and they write it in the language of the question.
So the panel has to be built in that language instead, and deciding what belongs in it is most of the difficulty. A generic beauty panel will hand you a number every month and never touch the places you are actually losing.
We run these panels monthly, including on our own brands. That is not a beauty case study and we will not dress it up as one — we have no signed client work in this category to point at. It is simply why we know how noisy the underlying measurement is.
Who this is not for
Five reasons to not hire us.
- You will not publish a full ingredient list. In this category that is the raw material for half the questions buyers ask; without it there is very little we can do that matters.
- You want us to buy your way into the roundups. We do not pay for placements or sponsored inclusion, and we will tell you when someone is offering to sell you one.
- Your catalog is a single hero product with no variant depth and no concern positioning. The honest read is that a monthly engagement will not pay for itself yet.
- You need this to move before a launch six weeks out. Editorial and community citations are other people's publishing schedules, and nobody controls them.
- You want a guaranteed placement in an AI answer. Nobody can sell that honestly, and in a category this concentrated the promise is especially cheap.
Questions buyers ask us
Frequently asked.
Why is beauty harder than other ecommerce categories?
Because the answer is written by other people. AIVO's July 2026 analysis of 750 beauty answers across five engines put brand-owned sites at 2.4% of the citations the engines actually read, and Goodie's September 2025 analysis of 135,419 beauty citations found the top ten domains taking 36.7% of them with no traditional beauty brand among them. In most categories on-site work moves the needle a long way. Here it has a ceiling, and the work that clears the ceiling is earned rather than published.
We rank on page one for our category terms. Doesn't that carry over?
Much less than you would expect. SE Ranking looked at 100,013 keywords across twenty niches and found Fashion and Beauty had the lowest overlap of any of them — 9.02% — between the links cited inside an AI Overview and the top ten organic results. The study is from 2024 and its trigger rates are stale, but the structural finding has held: in this category, ranking and being cited are close to independent. It is the single most common surprise we see.
How do shade ranges break AI visibility?
Google documents six axes that a product group may vary by — color, size, suggested age, suggested gender, material and pattern. That model was built for apparel. Shade nominally maps to color, but the three things a shade question is really about are depth, undertone and finish, and none of them has a dedicated property anywhere. Most beauty catalogs then compound it by shipping one URL for a whole family with a swatch picker on top. The practical consequence is that an assistant asked to match a shade across two brands has nothing to read, so it answers from a community thread instead.
Should we be trying to get mentioned on Reddit?
You should be present and accurate wherever your customers already discuss the category, and Reddit is unambiguously one of those places — Goodie found it accounted for 76.5% of the social and community citations in its beauty set. What you should not do is build a plan on it. Community citation share is volatile enough between measurement windows that treating it as a channel you control is a mistake, and manufactured community presence tends to get identified as exactly that. Be there honestly, and count it as one layer among several.
What actually moves first?
It depends on which of the six you are carrying, and most brands are carrying several. What is reliably true is that the part you fully control moves on a schedule you control, and the earned part does not — editorial and community citations run on other people's publishing calendars. NielsenIQ reported in July 2026 that 84% of the consumers it surveyed are more likely to purchase when key product attributes can be easily compared, which is the same problem stated from the shopper's side. Sequencing the controllable half against the earned half correctly is most of what an engagement decides in month one.
What can Deadpan Labs honestly claim in this category?
That we measure it and run the work that changes it. We have no signed clients in beauty, we have no vertical case study, and we are not going to manufacture either. What we do have is our own record: we run our own products on the same platform we sell, we took one of our own from zero to more than 900 organic downloads with no ad spend, and we run monthly AI-visibility prompt panels on ourselves, which is how we know what the run-to-run variance in these answers looks like.
Keep reading
The rest of it.
- /ai-search-visibility
AI search visibility for ecommerce
The pillar: what it is, how it differs from ranking, and the six general reasons a store goes unnamed.
- /industries/supplements
Supplement & wellness brands
A different problem entirely — there, the constraint is on what the brand is able to say at all.
- /industries/home-and-garage
Home, garage & equipment retailers
Where the answer is a number, and the specs are trapped in a PDF.
- /pricing
What it costs
Published: the audit price, the monthly bands, and the factors that move them.
The free check
See who the engines name instead of you.
Give us your concerns and your hero actives. We’ll run them against ChatGPT, Perplexity and Google AI and show you which brands come back, and which sources decided it.