Version 3 is live

AI skin analysisthat refusesto guess.

White-label skin analysis for beauty brands. Your customer answers five questions and takes one selfie. Claré checks seven skin concerns, drops any claim it can't back with evidence, and builds a routine from your own catalog.

  • 7 skin concerns
  • Your catalog only
  • Photo never stored

01 · Scan

One selfie, checked before it counts.

Your customer answers five quick questions, then takes one selfie in your store. The camera checks framing, light and blur in the browser. Nothing uploads until the photo passes, and a rejected photo never costs you a credit.

  • 5 questions
  • Consent first
  • Works on a shop tablet

02 · Analyze

Every claim needs evidence. Or it's dropped.

The AI has to describe what it sees before it may score anything. Our validator then removes any finding it can't justify, like shine under a camera flash or a "dark spot" that is really a shadow.

  • 7 concerns
  • Mild, moderate, severe
  • Location on the face

03 · Recommend

A routine from your shelves. Never ours.

Claré picks four products from your catalog, one for each step of the routine, ranked by your customer's own concerns. A declared allergen removes a product outright. Every pick says which concern it targets and why.

  • Allergen-safe
  • Budget-aware
  • One product per step

04 · Report

Not buying today? The report goes home with them.

Your customer downloads a PDF with their skin results and the products Claré recommended from your catalog. Every product links straight back to your store, so the sale can happen tonight or next week. The report is built in their browser, with no photo inside.

  • Your brand on it
  • Buy link per product
  • No photo inside

Why Claré

A quiz picks a bundle. Claré builds a routine.

A shopper opens your catalog, scrolls past twenty bottles and leaves. Quizzes map four answers to a pre-written set, and shoppers feel how shallow that is. Claré looks at the face in front of it, checks what it sees, and builds the routine from your shelves.

A scan, not a lookup

Five questions plus one selfie. The routine reflects what is visible today and what your customer says they care about.

Your shelves only

Every product comes from the catalog you enter. We never recommend ours, and a declared allergen removes a product outright.

A report to take home

Not ready to buy? The PDF carries their results and your recommended products, each with a buy link back to your store.

Honest by design

The AI describes before it scores, and a validator drops any claim it cannot justify. Cosmetic language only.

7skin concerns checked
4products in every routine
96automated checks on the engine
0photos stored. Ever.

The cost of guessing

When a shopper guesses wrong, your brand pays.

A customer who does not know their skin buys on a guess. If the product does not suit them, they blame your brand, and a customer who might have stayed for years is gone. Claré gives every shopper a personalised recommendation from your own range, so the first purchase fits.

Evidence first

Every claim has to earn its place.

Getting a vision model to answer is easy. Making the answer true is the work. We found five ways it goes wrong on real faces, and fixed each one in the system, not just in the prompt.

01

Leniency bias

Faces that human raters scored 52 to 55 came back from the model as 62.

Fixed byThe headline score is computed by a fixed function, fitted on hand-scored scans. Examples: 62 becomes 52, 72 becomes 62.
02

Spatial hallucination

“T-zone oiliness” showed up on almost every scan: a prior pretending to be an observation.

Fixed byEvidence comes before numbers. A region with no evidence string is dropped, and oiliness is dropped under flash or glare.
03

Confounders scored as skin

Beard stubble read as blemishes, moles as dark spots, brow shadow as pigment.

Fixed byExcluded features are named first. A blemish claim in a facial-hair region is discarded.
04

Double counting

One inflamed spot was counted as a blemish and again as a dark spot.

Fixed byDark spots are suppressed unless flat marks were actually counted, and each condition has a strict definition.
05

Age read as damage

A person of about 65 was told fine lines were severe, 82, and overall skin health 35.

Fixed byAge changes the framing, never the score. It is a skippable dropdown and is never estimated from the face.
+

Read the full story

The field-order mechanism, the computed score, the validator and the fairness finding we are proudest of.

Technology →

Seven concerns

What Claré looks for.

Each concern is reported as mild, moderate or severe, with a score and the areas of the face where it was seen, in the image frame. Tap a spot to see how it is defined.

Dark spots

Recommendation engine

A routine, not a leaderboard.

Three layers, all pure functions: the same input always gives the same routine. Safety is a wall. Preference is a weight.

01

Hard filters

A declared allergen removes a product. Very sensitive skin only sees sensitive-safe products. Budget is a soft ceiling, relaxed by one tier only if fewer than three products survive, and labelled “above your budget”.

02

Scoring

The concerns your customer names, in priority order, get a scoring floor and a rank multiplier. Diminishing returns stop a product tagged for everything from out-scoring a focused one.

03

Composition

At most one product per routine step. Step one always addresses a stated concern. Match percentages descend with rank, so the list reads honestly.

White-label

Make it yours in a minute.

Set the colour, the welcome message and the disclaimer. Customers stay inside your experience, and the report carries your brand name. Try it.

You set the colour, the welcome message and the disclaimer text. The report carries your brand name.

9:41
Your BrandSkin analysis

Find the routine that fits your skin.

Five quick questions, then one selfie. Your photo is analysed and never stored.

Start my analysis
Cosmetic analysis only. Not medical advice.
QUESTION 2 OF 5

What would you like help with?

Dark spotsBlemishesDrynessFine lines
GUIDED CAPTURE
FramingLightFocus
YOUR ROUTINE
1Azelaic acid 10% serum95%
2Oil-free ceramide moisturiser94%
3Mineral SPF 5088%
4Brightening eye cream79%
Download my report

The report

The sale does not have to happen today.

A shopper who is not ready to buy downloads a two-page PDF: their skin results, and the four products Claré picked from your catalog. Every product links straight back to your store, so the purchase can happen tonight or next week.

  • Your brand name in the header
  • A buy link on every recommended product
  • No photo inside. It is built in their browser from the analysis only

Privacy by construction

The photo is never stored.

Not a policy page: a guarantee in the code. The selfie lives in memory for one analysis, then it is gone.

No image on disk, ever

The scan record has no image field. The photo is not written to a database or a log.

A report with no photo

The PDF is generated in the customer’s own browser from the analysis, so a photo cannot be inside it.

Dictation stays on the device

Voice input uses the browser’s speech recognition. Audio never leaves the device, only an editable transcript is sent.

For brands

Built for how you sell.

On your website, or on a tablet on the shop floor. You see what is happening, and you only pay for scans that work.

Shop tablets

Run Claré in your outlets with a scan key per outlet, so you see where each scan happened. Automatic reset between customers is on our roadmap.

A scan key per outlet

Label each key, for example “Outlet Dolmen Mall”, and see where scans happen.

Failed scans are free

A credit is spent only after a successful analysis. A blurry photo costs nothing.

Brand dashboard

Manage products, scan keys and credits in one place.

FAQ

Questions brands ask first.

What is Claré AI?

Claré AI is a white-label skin analysis and product recommendation tool for beauty brands. Your customer gives consent, answers five questions and takes one selfie. Claré checks seven skin concerns, drops any claim it cannot back with evidence, and builds a routine of four products from your own catalog.

Who is Claré for?

Beauty and skincare brands that want to help shoppers choose from the brand’s own range. It runs on your website or on a shop tablet in your outlets.

What does the customer do, step by step?

They agree to consent, answer five quick questions (skin sensitivity, up to three concerns in priority order, allergies, current routine and budget, plus an optional age band they can skip), then take one guided selfie. After the analysis they see their results and a routine from your catalog, and can download a PDF report.

Which skin concerns does Claré check?

Seven: blemishes, dark spots, dark circles, oiliness, dryness, redness and fine lines. Each one is reported as mild, moderate or severe, with a score and the areas of the face where it was seen.

Is this medical advice or a diagnosis?

No. Claré is a cosmetic analysis tool only. It never diagnoses, never detects disease and never uses medical terminology. This boundary is enforced in the prompts and checked again by a validator, and every report says so.

Is the customer’s photo stored?

No. The selfie exists in memory for one analysis and is then discarded. It is never written to disk, a database or a log. Only the anonymised analysis is saved. The PDF report is built in the customer’s own browser from that analysis, so it contains no photo.

Founder mission

Why we built this.

Claré AI was built to close the personalization gap in beauty e-commerce. Guesswork is the most expensive part of skincare shopping, and most AI beauty tools are not built with the accuracy or affordability that independent brands actually need.

100% focused on D2C and beauty retailCosmetic match onlyNon-diagnostic platform

Get in touch

Want to see it on your store?

We take conversations with beauty brands every week. Write to us, follow our progress, or book a demo.

contact.clareai@gmail.com