Cape Ann
Auction
Preview Auctions

Cape Ann Auction Journal

Antiques & AI

A robot holding an ornate lamp in an antiques shop

I've been buying, selling, and appraising antiques since 1979. In that time I've watched this business change its tools more than once: handwritten catalogs gave way to typewriters, typewriters to computers, film cameras to phones, and live in-person auctions to online bidding. Every one of those changes had the same two groups — people who said it would ruin the business, and people who quietly learned the new tool and got better at the work.

Artificial intelligence is the newest tool on the bench. It is not going away, and it is not magic. Used well, it can make a good dealer, collector, or auctioneer faster and sharper. Used carelessly, it can make all of us look foolish — or worse, dishonest. Here's how I see it dovetailing with the world of antiques, honestly, challenges included.

The Challenges

For resellers, the biggest risk is speed without knowledge. AI can write a listing in ten seconds, complete with a maker, a date, and a confident story — and every word of it can be wrong. Tools like this will happily "identify" a reproduction as period, invent a maker's mark history, and price an item off listings that were themselves wrong. If you copy and paste that into your shop, your name goes on the mistake, not the software's. We're also starting to see marketplaces fill up with AI-polished photos and generic descriptions, which makes honest listings harder to spot and real expertise harder to prove.

For collectors, the danger is deception, and some of it isn't accidental. AI can clean up a photo until cracks, repairs, and replaced parts disappear. It can generate a warm "found in my grandmother's attic" provenance for an item made last year. It can produce entirely fake scenarios — a "rare" variant that never existed, photographed in a setting that never happened. The old rule matters more than ever: buy the object, not the story, and never buy the photograph alone.

For auctioneers, the challenge is trust at scale. AI can help catalog hundreds of lots, but an auction house stakes its reputation on every description. If the software calls a plated piece sterling, or a print a painting, the bidder who trusted the catalog remembers. There's also a quieter problem: as more pricing "data" online comes from AI-generated listings and asking prices that never sold, the comps we all rely on get polluted. Garbage in, garbage out — at industrial speed.

Where AI Genuinely Helps

Now the other side, because it's real:

Research speed. Forty years ago, identifying an unfamiliar mark meant a shelf of reference books and a phone call to another dealer. AI can point you at the right book, the right maker, and the right questions in minutes. It doesn't replace the books — it tells you which page to open.

Your own numbers. This is the one most dealers sleep on. AI can chew through years of your own sold results in an afternoon and show you patterns you'd never see lot by lot: which categories actually carry your business, what your average lot really brings, and how much of your income comes from a small share of top pieces. That's not theory — that's sourcing strategy. It tells you what to hunt, what to lot together, and what to stop spending time on.

Cataloging and marketing. First drafts of descriptions, keyword lists, email announcements, social posts, video scripts — AI is a tireless junior assistant for the writing every small operation struggles to keep up with. The key phrase is first draft. You edit, you correct, you add the condition notes only a human who held the piece can write.

For collectors, AI is a learning accelerator: translating foreign marks and inscriptions, organizing a collection, flagging new listings that match what you collect, and summarizing what's known about a maker before you bid. A collector who uses it to learn asks better questions — and better questions protect your wallet.

Sourcing. Set up right, AI can watch more estate sales, auctions, and listings than any one person can, and surface the ones in your specialty. In a business where the money is made at the buy, that early alert is a genuine edge.

Blending the Two: Ideas Worth Trying

Here's where I think this gets interesting — not AI or experience, but AI and experience, each doing what it's good at.

The two-pass rule. Let AI take the first pass — research, comps, a draft description. Then a human who has handled the object takes the second pass: weight in the hand, wear where wear should be, tool marks, smell, sound, the underside, the inside. AI has never held anything. You have held ten thousand things. That second pass is the business.

Label your judgment calls. When a price or an attribution rests on sold evidence, say so. When it's experience talking, say that too. Buyers trust dealers who show their work, and in a market filling with confident nonsense, honesty is a competitive advantage.

Disclose your photos. Cleaning up a background is fine. Erasing a flaw is fraud, whether a person or a program does it. A trade that adopts simple disclosure habits now will be glad it did.

Build a shared fake file. Dealers already warn each other about reproductions at shows and in shops. Imagine that same network, organized: a shared, searchable record of known fakes, fantasy pieces, and the AI-generated scam listings now circulating — what to look for, side by side with the real thing. No single dealer sees enough to build that alone. Together, we see everything.

Teach with it. Some of the best content a dealer can make right now is "real or render?" — walk your audience through an authentic piece and an AI-produced image or story, and show them the tells. You'll educate collectors, build trust, and sharpen your own eye in public. I believe courses and communities built around exactly that — verifying real objects and real experience against AI-produced items and fake scenarios — are where this trade is heading.

The Bottom Line

An antique is real or it isn't. It was made when someone says it was made, by whom they say made it, or it wasn't. No software changes that, and no software can stand behind a description the way a dealer with 46 years and a reputation can.

What AI can do is give that dealer more time at the bench and less time at the keyboard, better questions before the sale, and better numbers after it. The dovetail joint — and I choose the word on purpose — only holds when both pieces are cut true. Let the machine do the fast work. You do the true work.

— Walt Kolenda, Cape Ann Auction

← Back to the journal