Simona Supekar's first book, Stock Photo (Bloomsbury, 2026), arrives this summer as the latest entry in the publisher's Object Lessons series, an inventory of an industry whose product now quietly feeds most AI-generated images.

The book is a working critic's audit of a banal object. Supekar argues that the stock photograph has become "the Rolodex of the imagination," a low-grade visual currency that confirms what audiences already expect to see, then loops back to confirm it again.

In Supekar's telling, the file you never noticed is doing the most cultural work.

What does the book actually argue?

The premise is unglamorous and uncomfortable. The most-circulated images on the public web are also among the least-questioned, and the bias inside them travels further than any single buyer intends.

Supekar's framing in the preface is worth quoting at length: "These sometimes inscrutable photos act as ciphers to clue us in to ourselves even as we are becoming ourselves. They can allow us to see what we value, and what we do not."

It is, in other words, a moral reading of metadata.

The book belongs to a series unafraid of small subjects. Object Lessons has covered dandelions, dust, traffic, lipstick; Supekar adds the corporate image bank to that list, with results that range from wry to grim.

How does the racial ledger add up?

Supekar's evidence reads like a controlled experiment with the reader's cursor. Run a stock search for "maternity" and "Indian" on one platform and the first page now returns mostly AI-generated results, "signifying that these filled a gap for something that did not exist in the first place."

That gap, she argues, was engineered over decades of editorial choice inside libraries such as Getty. Light-skinned Indian subjects are "vastly over-represented" in Getty's collection, she writes — a preference invisible to anyone who has not paused to compare.

A "Black businessman" query once offered, for years, "an image of a Black man sitting at a bar next to lines of cocaine." These are not anecdotes. Supekar treats them as evidence of curatorial decisions no editor ever had to defend.

She is unflinching about scale. Early digital image banks, in her reading, have served as the visual backdrop to almost all public discourse online — including the meme economy, where an unseen Getty frame can persist for a decade inside a single reaction image.

Why does this matter for what AI inherits?

The closing chapters turn to generative systems, and the stakes sharpen. The same libraries that shaped twenty years of clip art, news illustrations and memes now shape the weights of diffusion models.

A stock library that under-represented brown mothers in the 2010s produces, by 2026, a model that imports those gaps as defaults. Supekar does not call this theft outright; she calls it inheritance.

She is not optimistic about a clean fix. Stock Photo sits in a deliberately uncomfortable corner of the Object Lessons shelf, beside volumes on toothbrushes, dust, traffic and lipstick — objects chosen, like hers, because they were never meant to be looked at closely.

What now?

What Supekar offers is an instruction any working picture editor will recognise: type your own search, note what returns, and ask who chose it.

Modest, perhaps. Considering how much cultural weight a single stock frame can carry — and how many have been dragged, uncredited, into generative training sets — modest is what the field can still afford.