Why, and how,
Beauty Dupe was built
What touches your skin is the ingredients, not the brand name. The same hyaluronic acid, niacinamide and ceramides can sell for more than a tenfold price gap once a brand name and the packaging are wrapped around them. Beauty Dupe uses ingredient data to show that gap clearly. We do not recommend anything we would not recommend to our own family and friends.
How does the analysis work?
Enter a product name or a full ingredient list and the result comes back through the three steps below. A human-defined verification check runs at every step.
Ingredient extraction
When you enter a product name, OpenAI (gpt-4o) pulls together the publicly listed ingredients. A photo of an ingredient label is read by OCR and converted into text.
Cross-check against the MFDS database
Each extracted ingredient name is matched against the MFDS cosmetic raw-ingredient database to filter out spelling errors and hallucinations. Any ingredient missing from the database is flagged separately.
Value-driven dupe matching
From a curated pool we pick products that share the same core active ingredients. We then verify their prices against the Naver Shopping and Coupang market and present them by value or luxury tier.
What data do we use?
We are open about the source data behind every result. Each source is checked and refreshed on a regular schedule.
Ingredient data · MFDS cosmetic raw-ingredient database
We pull the public cosmetic raw-ingredient dataset published by the Ministry of Food and Drug Safety (nedrug.mfds.go.kr) and use it to verify ingredient spelling and usage-limit information.
Price data · Naver Shopping + Coupang Partners
Dupe prices are pulled live from the Naver Shopping search API and the Coupang Partners API. Prices shift over time and may differ from the real checkout price, so we show the lookup time alongside each result.
Academic citations · peer-reviewed papers (PubMed and others)
The mechanisms and clinical-report figures in our magazine articles are cited directly from peer-reviewed papers indexed on PubMed, with the source named at each point of citation. We do not cite unverified marketing material.
How are the AI efficacy scores calculated?
Each result carries five scores from 0 to 100 (moisturizing, brightening, elasticity, soothing and safety) plus an ingredient-composition similarity percentage. They are worked out on the basis below.
Moisturizing
We look at water-binding and water-holding ingredients such as hyaluronic acid, glycerin, ceramides, panthenol and sodium PCA, and where they sit in the formula order.
Brightening
This is based on the ranking of ingredients linked to melanin synthesis such as niacinamide, arbutin, vitamin C derivatives and azelaic acid.
Elasticity
We analyze ingredients that support collagen synthesis or skin firmness, including retinol, the peptide family, collagen and EGF derivatives.
Soothing
This reflects whether soothing ingredients are present and how they rank, such as centella and madecassoside, allantoin, azulene, bisabolol and green tea extract.
Safety
Points are deducted depending on whether ingredients such as fragrance, alcohol, parabens and sulfate-based surfactants are present.
Ingredient similarity
We compare the core active ingredients of the original product and the dupe, then combine the share of overlapping ingredients with how closely their formula order lines up.
The limits of our analysis, and our responsibility
Our analysis is for reference only, and we want to be clear about the limits below.
The AI analysis is an estimate built from publicly listed ingredients. It cannot fully account for a maker's own formulation technique or differences in the origin and purity of a raw material. Always check the ingredient panel on the package itself before you buy.
Everything on this site is written for general cosmetic-ingredient education and does not replace a medical diagnosis, prescription or treatment. If you have skin trouble or a sensitive reaction, or you are weighing up products during pregnancy or while breastfeeding, please consult a dermatologist or pharmacist.
The efficacy notes in our magazine articles and on the results screen follow cosmetic advertising-language standards, using phrases such as "clinical report", "a main active ingredient" and "widely used". We avoid absolute claims such as "proven effect" or "clinically validated". We do not assess the safety of a particular brand or ingredient ourselves, and we do not promise drug-level effects.
Our editorial standards for ingredient articles
Every Beauty Dupe ingredient-analysis article is written to the four standards below.
Objectivity first
We write from INCI ingredient names, clinical reports and MFDS data rather than personal taste or brand marketing language.
Transparent source data
Every result is grounded in credible peer-reviewed papers and the MFDS ingredient-labeling rules.
No commercial bias
We never bend a product's analysis for paid ads or affiliate marketing. A recommendation is an option, never a push.
Kept up to date
We monitor products on a regular schedule so we can re-run the analysis as soon as a reformulation or ingredient change is confirmed.
Ingredient-analysis articles
Each article below is an in-depth report where AI analyzes a product's full ingredient list and cross-checks it against the MFDS raw-ingredient data.
Report an error or get in touch
If you find a data error, have a product you would like analyzed, or want to suggest an improvement, reach the atelier anytime.