Last updated: August 29, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Identifying fake cosmetics using AI means photographing the box, tube, label, and batch code, then handing those photos to a multimodal chatbot — ChatGPT, Gemini, or Claude — and asking it to flag anything that doesn't match an authentic version. It's a genuinely useful second opinion for the kind of tiny print and layout details a tired eye skips at 11 p.m. after a marketplace purchase arrives. It is not a verdict. No AI model was built or trained specifically to authenticate makeup, and in my testing the same tool that caught a real inconsistency on one product confidently called a receipt-verified, store-bought product counterfeit on another.
Short answer: Photograph the box, tube, label, and batch code under even light, then ask ChatGPT, Gemini, or Claude what looks inconsistent with the real product. AI vision is genuinely good at catching ingredient-list typos and mismatched print details, but it also misreads glare and shadow as defects. Treat its answer as one input, not a final ruling, and cross-check with the retailer.
I ran this test with a $14 blush bought through a resale app against a receipt-verified bottle of the same shade purchased directly from the brand's site, plus a fragrance sample a reader sent me that she suspected was fake. I photographed each item — box, applicator, ingredient panel, and batch code — under the same kitchen light, then asked ChatGPT, Gemini, and Claude the same question about each set of photos: "does anything here look inconsistent with the authentic version?" All three caught something worth knowing. All three also got something wrong, in a way that would have misled me if I'd trusted a single answer.
What you'll need
You don't need a paid subscription to start — the free tiers of ChatGPT, Gemini, and Claude all accept photo uploads, just with a daily cap on how many you can send. What matters more is the photos themselves: even, diffused light (a window on an overcast day works better than an overhead bulb), and close-up shots of the box, the product itself, the full ingredient list, and the batch or lot code, which is usually a short string stamped or printed near the barcode. Pull up the brand's official product page or app in a second tab so you have something concrete to compare against — an AI model comparing your photo to "what cosmetics packaging generally looks like" is far less useful than one comparing it to the actual current listing. A few minutes and a phone camera is the whole kit.
Step-by-step: identifying fake cosmetics using AI
1. Photograph everything, not just the front of the box
Get the front and back of the packaging, the product itself out of its case, the full ingredient list in focus, and the batch code up close. In my testing, the ingredient panel and batch code were where the AI tools found the most, and where I found the least on my own — small type is exactly what a language model with vision reads more patiently than I do.
2. Pull up the real listing side by side
Open the brand's own site or a retailer you trust, screenshot the product photos there too, and hand the AI both sets of images in the same conversation. Asking "is this real?" with no reference gets you a guess. Asking "compare these photos of my item against these official photos — what's different?" gets you something closer to an actual answer.
3. Ask for specifics, not a yes-or-no
I got noticeably better answers phrasing the request as "list every difference you can see" rather than "is this authentic?" The yes/no framing pushed all three chatbots toward more confident, more overconfident answers. The list framing surfaced hedged, checkable details — this font looks heavier, this cap seam looks different — that I could then verify myself.
4. Check the batch code against the brand's own tool
Most major beauty brands run a batch-code lookup either on their own site or through a third-party checker they link to. An AI chatbot can't verify a batch code against a manufacturer's database — it can only tell you the code looks formatted oddly, which is a clue, not a confirmation. Run the actual code through the brand's checker as a separate step.
5. Weigh the AI's answer against the obvious red flags
Price is still the strongest signal available to you. If a $38 product is listed at $11, that gap matters more than anything a chatbot says about the font on the box. Combine what the AI flags with the seller's rating, the marketplace's authenticity guarantee (Sephora, Ulta, and Amazon all offer some version of one), and your own gut about the deal.
6. Get a second AI opinion before you act on the first one
Because these tools disagree with each other — and sometimes with themselves — I ran every test photo set through at least two of the three chatbots. When ChatGPT and Gemini agreed on a specific flaw, I trusted it more. When only one flagged something and the other saw nothing wrong, I treated it as inconclusive rather than as evidence either way.
Example prompts you can copy
These worked across ChatGPT, Gemini, and Claude with minor wording tweaks:
- "Here are photos of a [product name] I'm not sure is authentic, plus official product photos from the brand's site. List every visible difference between them — font, color, packaging material, layout — even small ones."
- "Read the ingredient list in this photo and tell me if any words look misspelled or if the formatting is inconsistent with standard INCI ingredient labeling."
- "This is a batch code from a cosmetics product: [code]. Does the format look consistent with how brands typically print batch codes, or does anything look off?"
- "I'm comparing two versions of the same [product]. Don't just tell me if it's fake — list what you're basing that on so I can check it myself."
The last prompt made the biggest difference in my testing. Forcing each model to show its reasoning turned a shaky yes/no into a checklist I could actually verify against the real product page myself.
Common mistakes to avoid
The mistake I made first: taking a single "this looks authentic" answer at face value because it was reassuring. One of my test chatbots cleared the resale-app blush on the first pass and only flagged the mismatched cap texture when I explicitly asked it to compare the cap to the reference photo. Second, photographing in bad light — glare off glossy packaging got misread as a printing defect by more than one model, which is the single most common false positive I saw across every tool I tested. Third, asking a vague "is this real?" instead of feeding the model an actual reference image to compare against; without something concrete to check against, you get a plausible-sounding guess dressed up as an answer. Fourth, skipping the manufacturer's own batch-code checker because the AI chat already gave an opinion — a chatbot cannot query a brand's production database, and treating its guess as equivalent to that lookup is the gap that matters most. Fifth, assuming a AI-powered authentication app exists for cosmetics the way one does for handbags — it doesn't yet, and searching for one wastes time you could spend on the checks that actually work.
Tools that make this easier
If you don't already have a multimodal chatbot set up, how to use ChatGPT for free covers what the no-cost tier includes and where its upload limits kick in. Google's tool is worth trying too — how to use Gemini AI walks through its free tier and the "Thinking" mode that does a more careful multi-image comparison. For a third opinion, how to use Claude AI covers setup and its more hedged, cite-your-reasoning style of answer, which I found genuinely useful for this specific task. If you're deciding which of the big two to lean on for photo comparisons generally, Claude vs ChatGPT has my side-by-side notes from testing both on the same prompts. Want to test all three without spending anything first? Free AI tools rounds up what each no-cost tier actually allows. And if you want to go deeper on using ChatGPT's image tools specifically — cropping, zooming, annotating a photo before you ask about it — how to use ChatGPT Image 2 is worth a look.
How the AI options compare
I picked these because they're the multimodal chat tools people already have access to — there's no dedicated consumer app built specifically to authenticate cosmetics, which is worth saying plainly before you go looking for one.
| Tool | Free tier photo uploads | Paid price | What it's actually good at here | Cosmetics-specific? |
|---|---|---|---|---|
| ChatGPT | Yes, capped per day | $20/mo (Plus) | Reading small print — ingredient list typos, label text | No — general vision model |
| Gemini | Yes, capped per day | $19.99/mo (Google AI Pro) | Comparing several photos at once in "Thinking" mode | No — general vision model |
| Claude | Yes, capped per day | $20/mo Pro (or ~$17/mo billed annually, per Claude’s pricing page) | Hedging clearly and showing its reasoning instead of a flat verdict | No — general vision model |
| Entrupy | Not available to individuals | From $139/mo, business accounts only | Microscopic image fingerprinting with a financial guarantee | No — bags, sneakers, and apparel only |
Entrupy is worth naming because it's the closest thing to a real "AI authentication" brand most people have heard of, and it's tempting to assume it covers cosmetics too. It doesn't, at least not for an individual checking one purchase — its consumer-facing service is closed to individuals entirely, and its supported categories are leather goods, footwear, and apparel, not beauty products. For cosmetics specifically, a general-purpose vision chatbot is genuinely the best tool available right now, which says more about the gap in the market than about how good these chatbots are.
My take
Counterfeit cosmetics are a real, well-documented problem, not just marketing copy for whichever tool claims to solve it — the OECD and EUIPO's most recent joint report on global fake-goods trade names perfumery and cosmetics among the product categories most likely to be counterfeited, and the health risk is not hypothetical. Treat an AI chatbot's read on your product photos as a genuinely useful first pass, especially for catching print and ingredient-list errors you'd otherwise miss. Don't treat it as a substitute for the brand's own batch-code checker, a retailer's authenticity guarantee, or basic skepticism about a price that's too good. In my testing, the tools were most useful together — running the same photos through two of the three and only trusting what both agreed on cut down the false alarms substantially. If a product turns out to have made you sick, that's a matter for the FDA’s cosmetic complaint process, not a follow-up chatbot question.
Frequently Asked Questions
Is identifying fake cosmetics using AI free?
Yes, for the basic version. ChatGPT, Gemini, and Claude all let you upload photos and ask questions on their free tiers, with a daily cap on how many images or messages you can send. None of them charge specifically for this use case — you're just using the general chat product.
How long does it take to identify fake cosmetics using AI?
Taking the photos and getting a first answer from one chatbot takes about five minutes. Getting a reliable answer — checking a second AI tool, running the batch code through the brand's own checker, and comparing prices — took closer to fifteen minutes in my testing, which is still faster than most authentication services.
What is the easiest way to do this?
Photograph the box, product, ingredient list, and batch code in good light, open the brand's real product page in another tab, and ask a chatbot to list differences between your photos and the official ones rather than just asking "is this real?" That single wording change produced noticeably more useful answers in my testing.
Can AI actually tell me for certain if a product is counterfeit?
No. Every tool I tested produced at least one wrong answer — usually a false alarm triggered by glare or a photo angle, not a missed genuine fake. Use it to flag things worth checking further, not as a final ruling.
Is there a dedicated AI app for authenticating cosmetics, like there is for handbags?
Not currently. Entrupy, the best-known AI authentication brand, doesn't support cosmetics or individual consumers — its service covers bags, shoes, and apparel and is sold to businesses. For beauty products, a general-purpose vision chatbot is the closest tool available.