AI in Numismatics 2026 vs. Ancient Coins: Slabbed Moderns, Raw Romans, and Why Tangibility Still Wins

How does collecting this relatively modern piece compare to holding a coin struck in the Roman Empire? I ask myself that question often. In my experience grading and handling ancient material for over two decades, it has never felt more urgent than it does in 2026. This year, artificial intelligence has shown up everywhere in coins. At government mints. Inside third-party grading companies. In dealer inventory systems. In patent filings. Even on a circulating coin. I have examined hundreds of Roman silver denarii, Greek tetradrachms, Byzantine solidi, and medieval hammered issues. They survived because someone, centuries ago, dropped them, buried them, or simply forgot them. I have also examined the newest generation of slabbed moderns. Fresh from the mint, sealed in plastic within weeks of striking. The contrast is not just age. It is philosophy. Supply. Trust. Preservation. And AI is now forcing us to explain why ancient collectors do things differently.

Historical Tangibility: Holding 1,800 Years in Your Hand vs. Holding Last Month

The first difference I feel, every single time, is tangibility. When I hold a silver denarius of Septimius Severus struck in Rome in 193-211 AD, I am holding something real. It passed through the hands of a legionary. A tax collector. A market vendor in Londinium or Lugdunum. The metal is roughly 50-60% silver, often debased. The weight standard is around 3.0-3.3 grams. The strike is imperfect. The flan is off-center. The portrait is worn smooth on the high points of the cheek and nose. That wear is not a flaw to be graded away. It is a biography.

Compare that to a modern commemorative or a modern Morgan dollar reissue. A 2026 proof. Or a South African 2 Rand Women’s March coin that the Reserve Bank reportedly used AI-assisted design for in 2026. The metal is .999 fine. The weight is exact to the milligram. The mint mark is laser-crisp. The design was rendered, reviewed, and approved digitally before a single planchet was struck. I’ve examined pieces like this in my experience grading modern material for resale. The experience is clinical. You admire the engineering. You do not feel history breathing through it. Every coin is in pure mint condition. And that is precisely the point.

Ancient coins are primary historical sources. A Roman provincial bronze from Alexandria tells you about local cults, imperial propaganda, and debasement under pressure. A Greek owl from Athens, struck circa 440 BC, tells you about civic pride and trade networks that reached Persia. A modern coin, even a beautifully designed one, tells you primarily about the mint’s marketing department in the year of issue. That is not a dismissal. It is a distinction in historical tangibility. It also affects numismatic value in a way no population report can capture.

Why Touch Matters for Authentication

As an ancient coin specialist, I rely on touch in ways no camera can replicate. I feel the edge of a genuine ancient flan. Irregular. Sometimes with casting sprue remnants. Sometimes with a slightly rounded edge from centuries in soil. I feel the weight distribution. I feel the difference between a struck coin and a cast copy. A cast copy often feels too light, too smooth, or too soapy. I feel the patina. The dark, earthy toning on a bronze. The iridescent horn silver on a Greek piece. The crystalline deposits that only form over centuries. That patina gives a coin its eye appeal. No photo does it justice.

AI image models in 2026 are trained on photographs. From Vardera’s identification and grading benchmark to CoinAILyzer’s app focused on Morgans, VAMs, and Greysheet pricing, to Coinaikon and Numista’s search-by-image. Photographs flatten. They cannot feel a flan. They cannot smell the faint metallic tang of a freshly cleaned ancient silver. Or the musty soil smell trapped in a toned bronze. For modern coins, where strike, luster, and surface are standardized, a two-dimensional image may be 90% of the story. For ancient coins, where every coin is an individual, that image is perhaps 50%.

Supply vs. Demand: Finite Survival vs. Manufactured Scarcity

The second philosophical divide is supply versus demand. It is the one that ultimately drives value. Ancient coin supply is finite and shrinking in quality. We are not making more genuine Athenian tetradrachms. The hoards that fed the market in the 19th and early 20th centuries are largely dispersed. What remains in the ground is increasingly protected by patrimony laws, export restrictions, and archaeological excavation. When a hoard of Severan denarii appears, I know my lifetime supply just got a little smaller, not larger.

Modern supply is manufactured. A mint can decide, based on pre-orders, to strike 5,000, 50,000, or 500,000 examples. Scarcity is created by mintage limits, privy marks, and marketing windows. In my experience grading modern issues for dealers, I have seen “low mintage” pieces languish because demand never materialized. And high mintage pieces spike because a grader population report created artificial urgency. The coin itself has not changed. The narrative has. That manufactured collectibility rarely holds the same way.

AI has entered this supply equation in 2026 in two opposite ways. At the top end, the Royal Mint issued a Request for Information for AI proof inspection. A government mint admitting that human eyes alone cannot keep up with production volume and defect rates. At the collector end, tools like Reliqra, SpotPro, StackerKit, and CounterMint are using AI to manage dealer inventory, generate spot quotes from slab photos, and automate listings. That increases liquidity for moderns. For ancients, AI cannot increase supply. It can only help us identify the finite supply we already have, faster.

Actionable Takeaway on Supply

• For ancients: Rarity is intrinsic. A rare reverse type of a Roman emperor, a scarce mint, or a die variety does not need a population report to be rare. Verify with die studies, published corpora, and museum references — exactly what NumisVault Delphi is attempting with its AI image search built on museum reference data.
• For moderns: Rarity is often extrinsic. Check mintage figures, not just slab populations. Ask: was scarcity created by the mint, the grader, or the market?
• For investors: Ancient supply contracts over time as coins are lost, damaged, or institutionalized in museums. Modern supply expands the moment a mint sees profit. Price accordingly.

Slabbed vs. Raw: The Ancient Tradition That Refuses Plastic

This is where ancient collectors part ways most visibly with modern collectors. The modern world, in 2026, is a slabbed world. NGC ran an AI pilot letter in December 2025. Collectors, the parent company of PCGS, filed a patent on AI counterfeit detection. Vardera, by its founders’ own LinkedIn statements, is in talks with eBay about a potential partnership for authentication. The entire ecosystem assumes a coin should be photographed, graded, sealed, and sold inside plastic.

Ancient coins have resisted that tradition, and for good reason. I’ve examined more ancient coins in slabs in the last five years than in the previous fifteen. And I still prefer them raw. Here is why, from a specialist’s perspective:

Why Raw Remains King for Ancients

1. Surface is everything. Ancient grading is not Sheldon 70 versus 69. We describe coins by wear, strike, patina, and completeness. A slab’s holder obscures the edge, traps moisture if not perfectly sealed, and makes it impossible to rotate the coin under light to see hairline cleaning, tooling, or smoothing — the very things AI image models struggle to detect.
2. Weight and fabric matter. I weigh every ancient coin I consider buying. A genuine Roman silver should fall within expected tolerance. A slab prevents that. A cast copy inside a slab is a catastrophic failure of trust.
3. Patina is preservation. Ancient patina — stable, dark toning on silver, green-black patina on bronze — is historical preservation in action. Encapsulation can accelerate toning changes or hide active bronze disease until it is too late. Raw storage in a cabinet, with proper humidity control, lets me monitor a coin over years.
4. Handling is learning. I learned to grade ancients by handling them. By turning a worn Hadrianic denarius over and over, I learned what original luster in protected fields looks like versus a modern polish. No app teaches that.

That said, I am not anti-slab. For expensive, high-grade ancient gold — a superb aureus, a perfect Byzantine solidus — a slab from a reputable grader provides third-party verification for non-specialist buyers and for estate purposes. The issue is philosophical. Modern collecting has outsourced trust to plastic. Ancient collecting has traditionally kept trust in the collector’s hands, eyes, and scale.

AI is now trying to bridge that gap. The Boysel patent for two-sided reads and sorting, the Aizang patent for metal checks plus 3D wear analysis, and the Skoltech / Central University vision-language research all point to the same future. A machine that reads both sides, measures wear in three dimensions, and checks composition without touching the coin. As a specialist, I welcome that as a screening tool. I do not welcome it as a replacement for the moment I break a coin out of a holder to feel it.

Ancient coins were never meant to be sealed away from human touch. They were meant to circulate, to be counted, to be buried, to be found. Slabbing them is convenient. Handling them is historical.

Where AI Actually Helps Ancient Collectors in 2026 — And Where It Doesn’t

Let me be fair to the 19 developments catalogued in that forum thread. Accuracy percentage figures are each company’s own claims, not independently verified. And I treat them that way in my own appraisals. But several directions are genuinely useful for ancient specialists:

Useful AI Directions

• Image search against museum references: NumisVault Delphi and Numista’s search-by-image are the most promising for ancients. If I can photograph an unidentified provincial bronze and match it to a museum specimen with a published die number, that saves hours in printed corpora.
• Inventory management: Reliqra, SpotPro, and StackerKit are not grading my coins. They are organizing them. For a dealer with 5,000 ancient lots, AI-generated descriptions, pricing suggestions, and cross-referencing are a revolution — even if, as one forum poster noted, some auction lot descriptions now feel “fuzzy” and AI-generated.
• Counterfeit screening: The Collectors/PCGS patent and Vardera’s model are aimed at the modern market, but the same convolutional networks can flag the too-perfect surfaces, wrong die axes, and incorrect fabric of mass-produced ancient fakes from Bulgaria and elsewhere. As a first filter, that is valuable.

Where AI Falls Short for Ancients

• Die identification: Ancient numismatics lives and dies by die studies. Two coins can look identical in a photo but be from different obverse dies with a tiny letterform difference. Current vision-language models are not trained on that level of connoisseurship.
• Patina and cleaning detection: AI can learn that dark equals old, but it cannot reliably distinguish original earth patina from artificial toning, or a gently cleaned coin from a harshly polished one. I have examined coins where AI would likely grade them higher precisely because they had been cleaned — the opposite of what an ancient collector wants.
• Provenance and context: No AI can tell me if a coin was legally excavated, exported before 1970, or appeared in a documented hoard. That paper trail is often worth more than the coin’s grade.

In my experience grading, the best use of AI in 2026 is as an assistant, not an authority. Let it suggest a ruler, a mint, a date range. Then I, as the specialist, verify with weight, diameter, die axis, style, and fabric. The forum joke about “Numismatics Anonymous” fitting the AI era is apt. We are all anonymous data points feeding models. Ancient collectors must remain named, accountable experts.

Historical Preservation: AI as Tool, Not Replacement

Preservation is the fourth pillar of this comparison. And it is where ancient philosophy diverges most sharply from modern. A modern coin is preserved by being sealed. An ancient coin is preserved by being understood.

When I acquire a Roman coin, my preservation protocol is not encapsulation. It is documentation. I record weight to two decimal places, diameter, die axis, edge characteristics, and a full photographic record under raking light. I store it in a low-humidity cabinet, away from PVC, away from sulphur. I monitor for bronze disease — the powdery green spots that indicate active corrosion. I do not clean it. I do not polish it. I let the patina that formed over 1,800 years remain as a protective skin.

AI can assist preservation by creating digital twins. A high-resolution 3D scan, as suggested by the Aizang patent’s 3D wear concept, could preserve a coin’s exact surface for study even if the physical coin is later lost. That is powerful for museums and universities. But a digital twin is not preservation of the artifact. It is preservation of the data.

The Royal Mint’s interest in AI proof inspection and South Africa’s AI-assisted coin design show the modern mint’s philosophy. Preserve perfection through automation. The ancient collector’s philosophy is the opposite. Preserve imperfection through restraint. We do not want perfect coins. We want honest coins.

Preservation Checklist I Give Every Ancient Buyer

1. Never clean an ancient coin. Cleaning destroys patina, value, and historical evidence.
2. Store raw coins individually in inert flips or cabinet trays, not in bulk.
3. Keep relative humidity between 35-45% to prevent bronze disease and silver toning acceleration.
4. Photograph obverse, reverse, and edge before any handling or sale. AI search tools are only as good as your images.
5. Maintain provenance paperwork. In 2026, with AI-generated listings flooding eBay, documented provenance is the ultimate authentication.

What This Means for Buyers, Sellers, and Investors in 2026

So where does this leave us? I have seen forum members suspect, correctly I believe, that major auction houses are already using AI to generate filler lot descriptions. eBay has offered AI listing generation for some time. It is trivially easy for a large firm to automate first drafts. That means buyers must read more critically, not less. Fuzzy verbiage is a red flag, whether written by a human or a model.

For sellers, AI tools like CoinAILyzer, which claims expertise in Morgans, VAMs, and Greysheet pricing, and CounterMint, which turns a slab photo into spot quotes, will make modern selling faster. For ancient sellers, speed is not the goal. Accuracy is. A misattributed ancient coin, amplified by AI to thousands of listings, becomes a misattributed ancient coin forever.

For investors, the lesson is clear. Modern coins offer liquidity, grading transparency, and manufactured scarcity. Ancient coins offer finite supply, historical tangibility, and a collector base that values knowledge over plastic. In my experience, ancient coins weather downturns better because their value is not tied to a population report. A great Roman aureus was desirable in 1926, in 1976, and it will be desirable in 2076, slabbed or not.

Conclusion: The Philosophy That Endures

How does collecting a relatively modern, AI-designed, AI-inspected, AI-graded piece compare to holding a coin struck in the Roman Empire? It compares as a product compares to a relic. The modern piece is designed to be perfect, graded, sealed, and sold. The ancient piece survived despite imperfection, and its imperfection is its value.

AI will continue to transform numismatics in 2026 and beyond. It will identify coins faster, grade moderns more consistently, manage inventories more efficiently, and perhaps one day read ancient dies with specialist-level accuracy. I welcome every one of those developments as tools in my cabinet.

But I will still pick up a raw denarius, feel its weight, turn it under the light, and read its wear as a story of empire, economy, and everyday life. That moment — tangible, historical, and human — is something no model, no matter how accurate its own claimed percentage, can replicate. And for the ancient coin specialist, that moment is the entire point of collecting.

Related Resources

You might also find these related articles helpful:

• What is the Real Value of a Split in Half 1946 Jefferson Nickel in Today’s Market? – Determining the true value of this piece means looking past the book price. I have to understand current market demand. …
• The Global Market: International Demand for AI & Numismatics in 2026 — Foreign Buyers, Repatriation & Cross-Border Auctions – The Market Isn’t Just Local. Let’s Look at How Overseas Collectors and Repatriation Trends Are Affecting Its…
• The Buyer’s Mindset: Why Collectors Overpay for AI in Numismatics in 2026 — Completionism, FOMO, and the Thrill of the Hunt – What Drives a Collector to Pay a Massive Premium for a Tiny Piece of Metal? What drives a collector to pay a massive pre…