Reverse-engineer any product on the shelf.

theformulator.ai reverse-engineers a marketed cosmetic product into a probable bench-ready formula. Start from an INCI list, a product-name search across 1M+ products, or a photograph of the back label; the engine regroups decomposed supplier blends, estimates each concentration with a confidence tier, and screens the result against the markets you submit. Same credits and report tiers as formulation.

Paste an INCI list, search our database of 1M+ marketed products, or photograph a back label. Get a probable bench-ready formula with concentrations, phase assignments, and regulatory screening, in minutes.

Designed for professional formulators and R&D teams. Not a consumer recipe tool.

Start free, no credit card required.

1M+ marketed products searchable44,000+ ingredients with technical profiles21 regulatory markets screened429,000+ peer-reviewed findingsSupplier-Agnostic

Search 1M+ Products

Type a product name. If it's in our database of 1M+ marketed products across 21 markets, the INCI list auto-populates instantly. Coverage spans global giants like L'Oréal, Shiseido, and Estée Lauder to indie brands like The Ordinary, Drunk Elephant, and mCaffeine, across 27,000+ brands worldwide.

3-Layer Blend Detection

Commercial INCI lists decompose supplier blends into constituent names. Our engine re-groups them using thousands of ingredient technical records, statistical co-occurrence from 1M+ products, and functional chemistry rules. See what the formulator actually weighed out.

Concentration Reconstruction

INCI order gives you descending concentration above 1%. We go further, cross-referencing regulatory limits, TDS usage ranges, clinical efficacy data, and market norms to estimate probable concentrations with confidence tiers.

🔍 Product Search📋 Paste INCI📷 Label Photo

Same credit cost as formulation. Same 17-layer retrieval. Same regulatory screening across 21 markets. Same INCI-only, supplier-agnostic output. Every ingredient scored across six independent hazard axes — read our safety scoring methodology.

BLEND DETECTION

From label to lab — how blend detection works.

A commercial INCI list decomposes supplier blends into individual constituent names. Our engine re-groups them into the raw materials a formulator actually weighed out, using three independent detection layers validated against 1M+ products.

What the INCI list shows

1.Aqua
2.Glycerin
3.Cetearyl Alcohol
4.Caprylic/Capric Triglyc.
5.Niacinamide
6.Ceteareth-20
7.Dimethicone
8.Phenoxyethanol
9.Butylene Glycol
10.Tocopheryl Acetate
11.Ethylhexylglycerin
12.Centella Asiatica Ext.
13.Carbomer
14.Sodium Hydroxide
15.Xanthan Gum
16.Disodium EDTA
17.Caprylyl Glycol
18.Propanediol

18 individual INCI names — blends invisible, components scattered

Detection layers

01

Technical data matching

44,000+ ingredient profiles

02

Statistical corpus analysis

1M+ products

03

Functional chemistry rules

Chemical logic

Powered by regulatory intelligence across 21 markets, ingredient safety methodology, market intelligence from real products, and the chemistry knowledge base.

What the formulator actually used

Emulsifying wax

Cetearyl Alcohol

Ceteareth-20

Preservative blend

Phenoxyethanol

Ethylhexylglycerin

Botanical active

Centella Asiatica Ext.

Butylene Glycol carrier

Gel network system

Carbomer

Sodium Hydroxide neutraliser

18 INCI names → 9 raw materials + 4 detected blends

Validated against 1M+ commercially launched products

HOW IT WORKS

Search. Detect. Reconstruct. Screen.

01

Search or input

Type a product name to search 1M+ marketed products, paste an INCI list from any source, or photograph a back label with your phone. Vision AI extracts ingredient panels from curved bottles, foil packaging, and multilingual labels.

product name · INCI paste · label photo

02

Blend detection

Commercial INCI lists decompose supplier blends into constituent INCI names. The engine re-groups them into the raw materials the formulator actually used, matching against technical data sheets, co-occurrence patterns from 1M+ products, and functional chemistry rules.

technical records · co-occurrence · chemistry rules

03

Concentration reconstruction

Probable concentrations estimated for every ingredient using regulatory ceilings, TDS usage ranges, clinical efficacy data, and market norms. Each estimate carries a confidence tier, not a guess.

regulatory ceilings · TDS ranges · confidence tiers

04

Regulatory screening

The reconstructed formula is screened across your target markets: same 21-market regulatory engine, same safety scoring, same prohibited/restricted/permitted status per ingredient.

21 markets · safety scoring · prohibited/restricted/permitted

Ready to reverse-engineer your first product?

Start free, no credit card required. Same credits, same reports as formulation.

See pricing and report levels →

Frequently asked questions

How accurate is a reconstructed concentration?

It is a probable value with a stated confidence tier, not a measurement, and the tier tells you how much to trust it. INCI order gives descending concentration above 1% and nothing else, and the engine enforces that constraint. From there each ingredient's estimate is cross-referenced against its regulatory ceiling in the submitted markets, the usage range on technical data sheets, and the concentration norms observed for that product type in 1M+ launched products. Treat high-tier values as a sound starting point and lower tiers as candidates to confirm at the bench.

What is blend detection and why does it matter?

A commercial INCI list decomposes every supplier blend into its constituent names, so an emulsifying wax appears as "Cetearyl Alcohol" and "Ceteareth-20" several lines apart, and a preservative system as "Phenoxyethanol" and "Ethylhexylglycerin" scattered across the list. The engine regroups them into the raw materials the formulator actually weighed out, using technical-data matching against 44,000+ ingredient profiles and statistical co-occurrence across 1M+ products, with a chemistry layer that labels the result. A list of, for example, 18 INCI names can resolve to nine raw materials and four detected blends, which is what makes the output a recipe rather than a list.

Can I deformulate from a photo of the label?

Yes. Vision AI extracts the ingredient panel from the photograph, and the extracted INCI list follows the same pipeline as a pasted or searched list. Extraction is free and the list is editable, so check it before you spend a credit on reconstruction; if a panel is unreadable, a second photo costs nothing either, and the reconstruction runs only when you confirm the list.