Fstoppers has published an essay that asks a question an increasing number of working photographers face as they consider licensing or training on their own archives: if a generative model trained on your photographs produces images that look unmistakably like your work, are those images photography?

The piece, titled "If AI Is Trained on Your Photographs, Is the Result Still Photography?", opens with a concrete scenario. You spend twenty years building a photographic archive, then train a generative model on it. The model begins producing images in your signature style. The question of authorship and category follows immediately.

The author splits that single question into two separate ones, and the distinction matters for anyone whose income depends on how photographic work is defined, licensed and sold. First: are the generated images themselves photographs? Second: can the practice that produced them still be called photographic?

The framing rests on a historical observation that photographers who work in experimental and cameraless traditions will recognize. As the piece puts it, photography "has worked without a camera and without shooting" before. Generative images, the author argues, test what remains of the medium after those elements are removed — and, implicitly, what remains of the professional identity attached to it.

For photographers weighing whether to train models on their own catalogs — or to license their archives for third-party training — the essay offers a lens rather than a verdict. It does not answer the two questions it raises. Instead it positions them as the core definitional problem that generative tools have created for the field: the output can resemble a photographer's work closely enough to be mistaken for it, while the process that made it shares nothing with the act of photographing.

That gap between output and process is where the business implications concentrate. If clients buy "photography" and cannot distinguish a generated image from a photograph made from a scene the photographer stood in front of, the value proposition of the archive, the style and the photographer's name shifts. The essay's scenario — a model producing images that "look unmistakably like your work" — describes a product that competes directly with new commissions from that same photographer, or with licensing revenue from the existing archive, depending on who controls the model.

The piece also implicitly raises the ownership question that follows from the definitional one. The model's output derives entirely from the photographer's twenty years of work, yet whether that derivation makes the output photography, and whether it makes it the photographer's, remain separate issues. The essay frames the first; the second sits behind it.

Fstoppers does not cite specific tools, court cases or licensing deals in the excerpt, so photographers reading it as a business signal should treat it as a definitional argument rather than a legal or market analysis. What it does establish is that the question is now being asked in practical terms — not as philosophy, but as a scenario a working photographer with a substantial archive can actually execute today.

The essay is available in full on Fstoppers, and anyone whose revenue depends on licensing an established body of work will find the two questions it poses worth answering before signing any training-data agreement.