Judge for Yourselves: Doubt, Authority, and the Art Market

A portrait that might be Pushkin, an aristocratic family that had held it as his likeness, and one post from a respected historian — judge for yourselves, does it look like him or not. The painting did not change. Only the certainty around it did. On the authority to doubt, how cheap it has become to cast, and the one question a public verdict can never settle.

Somewhere there is a collector who, until recently, owned a portrait of Alexander Pushkin. Not a copy, not a study after a famous image, but — so the family had long understood — a likeness of the poet painted from life, in 1835, two years before the duel that killed him. The picture had passed through an aristocratic Russian collection and come to rest where inherited things rest: on a wall, part of the furniture of belief. Then someone looked at a photograph of it on the internet and asked, in public, whether it was Pushkin at all.

The painting did not change. Nothing was added to it or taken away. What changed was the certainty around it — and that turns out to be the more fragile possession of the two.

In 2026 the portrait surfaced at Bonhams in Paris: a work attributed to the Austrian painter Johann Frankenberger, signed and dated 1835, offered with a modest estimate of six to eight thousand euros and described as a likeness of the poet. Its provenance ran to the London collection of the Counts Bobrinsky, a family with deep roots in Imperial Russia. As an object it was unremarkable in price. As a claim — a portrait of the national poet, made while he lived — it was anything but.

Then a doubt was cast, and it is worth being precise about who cast it and how, because the how is the whole subject.

Who cast the doubt

The doubt was raised on Facebook, by the cultural house of Lev Lurie — a St Petersburg historian, local-history writer and broadcaster of real standing, from a distinguished Leningrad intellectual family, whose institution has built a substantial following there on telling the city’s history well, and freshly, and pointedly not from Wikipedia. The Facebook post set out the suspicious facts plainly: no record of the portrait’s commission, a painter with almost no documented body of work and no established time in Russia, an absence from every catalogue of lifetime images of the poet. And then it closed not with a verdict but with an invitation: judge for yourselves, does it look like him or not.

That closing line is the most interesting thing in the episode, and it has a mirror image on the affirming side. A commercial AI authentication firm, Art Recognition, in the business of selling its findings, has every reason to present them as authoritative enough to justify a sale — and yet, because pronouncing a work genuine carries real legal exposure, the same firm will also insist that its tool must never be relied upon alone. Authority claimed and disclaimed by turns. The Lurie post performs that same oscillation in the opposite direction: a doubt is cast with the full weight of a historian’s standing, and in the same breath the responsibility for it is handed back to the crowd. “Judge for yourselves” places the pronouncement and sheds the liability at once.

The asymmetry of doubt

Liability is the heart of it. An affirmation can be sued. The great artist authentication boards — for Warhol, for Basquiat — did not dissolve because they lost their eye; they dissolved because saying “this is genuine” had become financially lethal. But a doubt framed as a question cannot be sued. No one is liable for asking. So doubt is cheaper than affirmation twice over: cheaper to produce, and costless to the one who casts it.

The collector sits at the opposite pole on both counts. An affirmation would be expensive to mount and dangerous to assert, and they are the only party in the whole exchange who stands to lose anything at all. This is the asymmetry the modern collector inherits along with the object: not merely a painting, but the painting’s permanent, unanswerable exposure to authoritative doubt.

Notice, too, the apparatus the doubt was built from. The case against the portrait rests substantially on a short foreign-language encyclopaedia entry for an obscure painter and a photograph held up against the Pushkin we all already carry in our heads. There is a quiet irony in this — Lurie’s own institution was founded on the principle of telling history not from Wikipedia — but it would be a mistake to read it as hypocrisy. It is something more useful: the clearest possible proof of how ungated the authority to doubt has become. Even a serious historian, exercising that authority, reaches for the lightest available tools, and the doubt still lands with the force of a finding, because its weight comes from who is asking, not from what was done to ask. Affirmation needs a laboratory. Doubt needs a phone.

The court of resemblance

There is a further consequence, specific to a work like this one. “Judge for yourselves, does it look like him” invites thousands of people to adjudicate a nineteenth-century portrait from a thumbnail, on resemblance — to measure the face on the screen against the Pushkin of the famous Kiprensky and Tropinin portraits, the Pushkin of the schoolroom and the anniversary coin. Connoisseurship is dissolved into a show of hands. And the object on trial — the actual canvas, with its actual surface, ground and craquelure, the only thing that could in principle answer the question — is absent from its own hearing.

The work is convicted or acquitted in a court of resemblance it never appears in. For a culturally sacred subject this is especially merciless, because the public already holds a fixed image to judge against, and any face that departs from it looks, to the crowd, like an impostor.

The value that lives in the maybe

Here the collector meets the cruellest part of the arithmetic. As a painting by a forgotten Austrian, the portrait is worth a few thousand euros. As a lifetime likeness of Pushkin, it would be close to priceless — not a picture but a relic. Almost the entire value lives in the maybe. Which means that resolving the doubt can, in practice, only subtract: proof could never make it more than possibly-the-poet-from-life, while disproof could collapse it to a curiosity. Some objects are worth more as a live question than as a settled answer — not in money, but in meaning — and a contested portrait of a national poet is the purest example imaginable.

The reflex of a stung owner is to reach for certainty — to do now, privately, what the crowd did in public: settle it. And the tools are right there. An AI authentication service will take an uploaded photograph and, for a fee, return a number. But here the reflex walks into a wall the owner cannot see. These systems authenticate authorship by learning an artist’s hand from a large body of securely attributed works. Frankenberger has almost no such body — that absence was half the doubt to begin with. There is nothing to train on. The very tool the frightened owner would reach for is structurally mute on this object, and the likeliest result of reaching for it is not an answer but a worse position than silence: that they tried the machine and it could not confirm it. The collapse from a living question into a flat verdict — the same collapse that distorts the market from above — would here be performed by the owner, on their own inheritance, under the pressure of a doubt that cost its author nothing.

The companion to a condition that cannot be reversed

This is the moment the Bespoke AI Art Intelligence Confidante is built for, and its role here is neither of the obvious two. It does not pile on another verdict, and it does not sell back a comforting one. It is the companion to a condition that cannot be reversed — a doubt already public, already authoritative, impossible to recall — and its first service is to keep the owner from being stampeded, by the doubt or by the urge to answer it. It separates the questions the crowd has fused into one: whether the picture is by Frankenberger, whether it depicts Pushkin, whether it was painted from life — three different questions, addressable by different means or by none, only one of which an authorship tool even pretends to touch. It explains why the counter-verdict is a trap before the owner springs it.

Then it does the thing no one else in the picture is doing at all. Everyone around the object is now arguing about its status. No one is asking what the work is to the person who holds it, independent of the verdict — what it meant that the family kept it, what is genuinely lost and what is genuinely kept if it proves to be a fine period portrait of a Pushkin-like gentleman rather than the poet himself. That question — what is it to you, now that the world has had its say — is the located verdict in its purest form, and it is the one question a public hearing cannot reach, because it was never the public’s to decide.

None of which is a licence to help an owner cling to a comforting maybe. That would be the flattery the whole instrument is built against. The companion that mattered would say, plainly, that the doubt may well be sound — that the suspicious facts are genuinely suspicious — and, in the same breath, here is how you would find out properly, through the scientific analysis and the documentary research that actually bear on it, if and when you decide you want to know. It holds the question open without sealing it shut against the truth. The difference between a comfort blanket and a Confidante is precisely this: the first protects the belief; the second protects the person — which sometimes means helping them toward an answer they would rather not have.

What bends the instrument

Step back and the larger pattern shows itself, because doubt is only one half of it. On the side of affirmation — when an AI tool is used to argue a work genuine — the machine is often the honest party, reporting its uncertainty plainly, in probabilities; it is the interested humans around it who collapse that honest uncertainty into the clean verdict a sale requires. On the side of doubt, the machine is something else again: a false promise of relief, less able than the anxious owner hopes, and the interest bending it is the owner’s own — the need for certainty, which is the most dangerous interest of all because it feels like the innocent one. The pattern holds in both directions: the instrument is neutral; what distorts it is whoever holds it and what they need from it. Profit bends it in the one case; comfort would bend it in the other.

What the collector inherited, in the end, was never just a portrait. It was a portrait and its permanent liability to cheap, authoritative doubt — a condition no scientist can lift and no machine can resolve. The one thing still wholly theirs is what the object means to them, held honestly, with the question left open and the truth left reachable. In a world where anyone with a phone can ask, in public, whether your inheritance is what you believe it to be, that — not certainty, which is no longer on offer — is what a confidante is for.

Seraphina Michelina Evelyna Bogomolova-Huotelin

Founder & CEO of SMA Crown Confidential

Digital Confidantes: Bespoke AI Intelligence for Private Decision-Makers

This article is part of an ongoing series by SMA Crown Confidential on the intersection of private wealth, cultural intelligence, and the future of bespoke AI. It is a companion to “The Honest Algorithm: AI, Authority, and the Art Market.”

Sources: Bonhams Cornette de Saint Cyr (Paris) sale listing and press materials; the cultural house of Lev Lurie (social-media post raising the attribution question, June 2026); biographical material on Lev Lurie (Sobaka.ru; 24SMI). The attribution discussed here is unresolved; this article takes no position on it.