WinPhoto

№ 03 · The journal

The four AI positions a contest can take

The AI-policy debate lives in the develop module, not in robot-abstract space. Photograph: TourBox via Unsplash.
The AI-policy debate lives in the develop module, not in robot-abstract space. Photograph: TourBox via Unsplash.

The line on AI in every contest brief is the line that gets read fastest and understood worst. Photographers see “AI permitted” and think they’re safe. Photographers see “no AI” and think they’re disqualified. Both readings are wrong about as often as they’re right, because there isn’t one AI policy in 2026 contests — there are four, and the contest you’re about to enter is on one of them.

The taxonomy below is descriptive, not prescriptive. It is what contest rules actually say, decoded into the four positions any honest brief in 2026 can hold. Knowing which position your contest occupies is the difference between submitting a photograph that will be assessed on its merits and submitting one that gets disqualified before any aesthetic read happens.

Position 1 — Generative AI forbidden in the final image

The biological-fidelity position. Documentary contests, journalism contests, the photojournalism wires, and almost all conservation or scientific-imagery contests sit here. Read literally: every visible pixel in the entry must come from light that fell on a sensor.

What this also means, that photographers tend to miss:

  • Stack-and-merge composites are usually OK (multiple exposures of the same scene, combined to recover dynamic range). The pixels still originated from the sensor.
  • AI denoise, sharpening, and dust-removal are usually OK (they operate on captured pixels rather than inventing new ones).
  • AI sky replacement is NOT OK (replaces captured pixels with invented ones, even if “the sky was that colour earlier”).
  • Generative fill is NOT OK (invents pixels, period).
  • Aggressive AI upscaling beyond ~2× is a grey zone leaning NOT OK at this position; the algorithm interpolates pixels that weren’t measured.

The disqualification risk for photographers at Position-1 contests is overwhelmingly about generative fill and sky replacement applied “to clean up a frame.” If you used either on the photograph you’re submitting, you are out before the jury reads the rest.

Position 2 — AI-assisted edits permitted, generative forbidden

The “darkroom-continuous” position. Many fine-art, landscape, and travel contests sit here. The rule explicitly distinguishes between enhancement of captured imagery and generation of new imagery. AI denoise is treated as continuous with the darkroom tradition that includes dodge-and-burn — operations on captured material.

The trap here is reading “AI permitted” as a green light for generative fill. It is not. Photographers who use Photoshop’s generative tools for object removal, even small ones, are disqualified at Position-2 contests as cleanly as at Position-1. The clean test: was a pixel-region invented from a prompt or from sampling-from-the-frame’s-own-context? Generative = out. Content- aware-fill that samples from within the frame’s own captured area = usually OK at Position 2, usually not at Position 1.

Position 3 — Dedicated generative-AI category

The “label it” position. HIPA’s Dreams Through AI category is the cleanest example, and an increasing number of contests have adopted this structure for the 2026 cycle. AI-generated imagery is welcome, in its dedicated category only. Submitting a generative- AI frame to the regular categories is disqualifying; submitting a captured frame to the generative-AI category is allowed but typically uncompetitive.

The trap here is self-classification. The photographer with a heavily-AI-edited capture-based frame must honestly decide: is this a captured photograph that got AI-assisted in post (regular category), or is this a generated work with photographic references (generative category)? The contests at this position do investigate. Misclassification when caught is functionally a ban: not just from the current cycle, but from the contest’s photographer database for future cycles.

Position 4 — AI fully permitted

The “anything goes” position. Rare in serious contests, common in stock-image contests, brand-sponsored aesthetic contests, and travel-promotion contests run by tourism boards. The rules permit fully generative work alongside captured work, judged together.

The position is not “wrong” — it’s just a different kind of contest. The skill being assessed is visual outcome quality rather than photographic practice. Photographers who work primarily with captured imagery should know going in: you are competing against people whose entire workflow is generative. Your skill ladder and theirs are different ladders. Sometimes the captured frame wins on grounded specificity; often the generated frame wins on technical polish. Don’t assume your years of photographic practice are weighted heavier than they are.

The reading order

When you read a 2026 contest brief, find the AI line, and then look for these markers — in this priority order:

  1. “Generative AI” or “AI-generated content” mentioned at all. If not mentioned, default to Position 1 (assume the conservative read). Most established documentary contests don’t mention it because they don’t have to.
  2. A separate category for AI-generated work. If present, you’re at Position 3. Read both the regular category rules and the generative one before deciding where your work belongs.
  3. Explicit allow-list for “assisted edits” or “neural processing” alongside a generative ban. Position 2.
  4. Silence on AI altogether. Position 1 by default. The contest organisers haven’t updated their brief, but the jury will read for biological fidelity regardless.

The fifth position — the contest that hasn’t decided yet and will make it up at the jury table — exists in 2026 but is rare among established contests and disappearing fast as ambiguity becomes expensive. If you see it, treat the contest as Position 1 and ask the organiser directly before submitting anything that uses generative tools.

The honest disclosure principle

Across all four positions, one rule holds: if you wouldn’t disclose the AI use in a journalism context with your name attached, it doesn’t belong in any serious photography contest. The contest that permits it openly is rare; the photographer who quietly uses it where it isn’t permitted gets discovered eventually, and the discovery follows them.

The reverse is also true: if you’d disclose the use in a journalism context — AI denoise on a heavily-pushed exposure, content-aware spot removal of a sensor dust mark — you’re almost certainly fine at Positions 1–3 for that operation. The rules are stricter than photographers think on generative work, more permissive than photographers think on captured-image enhancement.

The full reference table — 22 major 2026 contests classified by position with source rule pages linked — lives at winphoto.io/ai-policy-2026. Bookmark it; the rules drift quarterly and the table tracks them.

— The Critic

Premium Reports · № 01

HIPA Family 2026 — The Pre-Submission Report

A 25-page editorial deep-read for the 10 days that remain before the largest free-entry photography prize on the calendar closes.

HIPA Family 2026 Pre-Submission Report cover

25 pages · €19 · Free for Studio

The brief decoded against three operational nouns. Five past Grand Prizes read for craft. A seven-point Strong Submit checklist that runs against any frame in ten minutes. Three hypothetical reads in the engine's voice. Illustrated with public-domain documentary masterworks from the Library of Congress FSA archive.

Read the page →

Every buyer receives ten Premium Runs of the HIPA Family verdict engine — redemption at /redeem.

← Back to the journal