Inside Luxury Authentication: Why AI Cannot Replace the Human Authenticator

“I handle these bags every day. I see them, touch them, smell them, examine them. I practically breathe Hermès.”

It may sound like an unusual way to describe the work of an authenticator, but for a member of our authentication team, it explains something important about authentication.

When you see, touch and handle luxury bags every day, knowledge develops beyond simply recognising a stamp, stitch or engraving in a photograph. You begin to understand how the leather should feel, how the hardware should feel and handle, how the construction should come together and, importantly, how all those characteristics should relate to one another.

Even with that daily exposure, experienced human authenticators can still encounter bags that require closer examination, microscopic magnification, or a second or third professional opinion.

So if authentication remains challenging for humans who have handled these bags every day for years, and even for authentication apps built specifically for this purpose, how can ChatGPT, Gemini or another broad AI tool reliably determine whether a high-value resale bag is authentic from just a few photographs?

We sat down with a member of our authentication team to ask her.

Why Is Hermès So Difficult to Authenticate, and What Does That Mean for AI?

Q: Is Hermès more difficult to authenticate than brands such as Chanel, Louis Vuitton or Dior?

A: I find Hermès much more challenging.

With some other brands, there can be more standardisation. Within a particular model or production period, you generally have a better idea of what you expect to see in the leather, texture, construction and other details.

Hermès can have much more variation. Even two authentic bags in the same model, colour code and leather may not look exactly alike.

There can be variations in the leather, stitching, hardware, engraving, finishing and even the colour itself. For example, the shade of Gold on one Birkin may look slightly different from the Gold on another Birkin, even when both are made in the same leather, such as Togo.

That’s what makes Hermès authentication so challenging. Something being different doesn’t automatically mean it’s wrong. You need experience to understand whether you’re looking at a legitimate variation or something that shouldn’t be there.

And that is important when we talk about AI. An AI tool may be able to identify that something looks different, but the much harder question is whether that difference is actually wrong.

What Does Experience Teach a Human Authenticator?

Q: When an Hermès bag comes in, where do you begin?

A: I don’t immediately look for one particular “tell.”

First, I look at the bag as a whole. After handling these bags continually, you develop an instinct for how a bag should look, feel and come together.

Then I break it down, looking at the leather, construction, stitching, hardware, proportions, stamping, engraving and smaller details.

I’m not simply asking whether each individual detail looks correct. I’m asking whether everything makes sense together.

Sometimes something doesn’t feel right before I can even explain exactly what it is. That’s when I know I need to examine the bag much more closely.

That’s the difference experience makes. After you’ve examined enough bags, you start to understand not only what something should look like, but how all the details should relate to one another.

Have Super Fakes Made Authentication More Difficult?

Q: What has surprised you most about counterfeit Hermès bags?

A: How good they can be.

I’ve seen counterfeits where the obvious details are extremely convincing. The stamp can look right, the stitching can look right and even the hardware can be very close.

A counterfeit can get many things right and still have one or two characteristics that don’t belong together. It could be the construction, proportions, leather, engraving or the way a component has been finished.

The better counterfeiters become, the less useful it is to look for one visual sign and assume you have your answer. A sophisticated counterfeit can deliberately reproduce the very features people have been taught to look for.

That’s why authentication can’t be reduced to a checklist.

If AI Can Analyse a Photograph, Why Isn’t That Enough?

Q: Why shouldn’t someone simply photograph their bag and ask ChatGPT or Gemini whether it’s authentic?

A: Because authentication isn’t purely visual.

Photographs can be blurred, edited, taken under different lighting or simply lack the resolution we need. Sometimes the differences are so small that we need much greater magnification or microscopic lenses to examine them.

And there are things a photograph cannot give us.

We can touch the leather and feel its texture. We can physically handle the hardware, feel its weight and temperature, examine its plating and inspect the construction. There are even characteristics such as smell that you become familiar with when you’re surrounded by these materials every day.

That’s what I mean when I say we practically breathe these bags. We see them, touch them, smell them and handle them constantly.

AI cannot feel the leather for us. It cannot smell the bag. It cannot physically handle the hardware. And if an important detail hasn’t been captured in the photographs, it cannot examine something it hasn’t been shown.

What About Authentication Apps?

Q: Authentication apps also use technology and AI. How are they different from ChatGPT or Gemini?

A: Authentication apps are built specifically for authentication.

They can work with large amounts of relevant reference data and images, using technology to compare the details being examined. Human authenticators may also be involved.

But even authentication apps can encounter difficult cases.

That’s particularly important with Hermès because there are so many legitimate variations. Human judgement may still be necessary to decide whether something is an acceptable variation or a reason to investigate further.

So if an authentication app built specifically for luxury authentication can still require human expertise, how can we expect ChatGPT or Gemini to give a definitive authentication from a few photographs?

I think AI can help an authenticator. That’s very different from AI being the authenticator.

“But ChatGPT Says My Bag Is Fake”

Q: What do you say when a customer tells you ChatGPT has examined their bag, engraving or receipt and said it isn’t authentic?

A: I would tell them not to treat an AI response as authentication.

There isn’t one magic stamp, stitch, engraving or hardware detail that proves whether an entire bag is authentic.

The stamp is one part of the examination. What matters is whether that stamp makes sense in the context of the particular bag, its model, period, materials, construction and other characteristics.

The same applies to a receipt.

If someone is capable of counterfeiting an entire luxury bag, including its leather, stitching, hardware and construction, a piece of paper can also be reproduced.

At the same time, genuine receipts can vary between countries and periods. Brands can change the paper, printing, layout and other details.

So beautiful photographs, packaging, receipts and convincing stamps are not enough on their own.

You have to look at the story the entire bag is telling.

Can an Experienced Authenticator Still Get Stuck?

Q: With enough experience, do you eventually reach a point where you can confidently authenticate every bag?

A: No. Authentication is a lifelong journey.

Brands keep changing. Materials change. Details change. New bags appear and counterfeiters continue to improve.

We have to keep learning and adapting.

There will still be difficult bags, particularly those that are rare, vintage or unusual. Sometimes you’re not completely confident and another authenticator needs to look at the bag.

A second or even third opinion can be the responsible thing to do.

I think that’s important when we’re talking about AI. If someone who handles these bags continually can still say, “I’m not completely sure, I want another opinion,” why should we accept a definitive answer from ChatGPT based on a few photographs?

Can AI Ever Replace the Human Authenticator?

Q: So do you see any role for AI in luxury authentication?

A: Absolutely. I see AI as a tool, not a replacement for an experienced authenticator.

It can help with image comparison, flag unusual details and potentially make parts of the examination more efficient. I would certainly use technology to support authentication.

But for a high-value luxury bag, particularly something rare, vintage or unusual, I still want human judgement involved.

AI can assist the authenticator. It shouldn’t replace the authenticator.

The Takeaway

Perhaps the biggest misconception about luxury authentication is that there is one detail that can prove a bag is real or fake.

There isn’t.

Authentication is about consistency. Does the leather make sense? Does the construction make sense? Does the hardware make sense? Does the stamp make sense? And most importantly, do all of those things make sense together?

For buyers in the resale market, the message is simple, don’t let confidence replace due diligence.

ChatGPT, Gemini and other AI tools can provide information about luxury bags, but an answer generated from photographs of a bag, receipt, stamp or engraving should not be relied upon as professional authentication.

The strongest approach combines experienced human authenticators with specialist knowledge of the particular brand, physical examination of the bag, authentication apps and technology where useful, and additional professional opinions when a difficult case requires them.

Because an authenticator doesn’t simply look at a bag.

They handle it, feel it, examine it, question it and continuously learn from it.

And that human experience cannot be captured in a handful of photographs.