Tools have launched that claim to show how much AI was used in an image — and, separately, to stop generative AI models from 'training' on photographers' pictures without payment or permission. Both promises target working shooters' two loudest complaints about the AI era: provenance and uncompensated scraping.

For professionals, the stakes are straightforward. If a client can't tell whether an image was captured or generated, the value of commissioned photography erodes. And if model makers can ingest portfolio images for free, licensing revenue leaks away before a contract is even signed.

What can these tools actually do?

One class of tool analyzes an image and reports how much of it is AI-generated or AI-edited. That's the pitch, at least. Detection of generative content remains an unsettled science; no vendor in this space publishes error rates in a form buyers can audit, and false positives — flagging a genuine photograph as synthetic — carry real commercial risk for a working photographer's reputation.

The second class of service addresses training. The claim: opt out of generative AI training on your pictures. Here the mechanics matter more than the marketing. An opt-out only binds companies that agree to honor it, and enforcement depends on crawlers respecting signals rather than legal obligation. What the announcements leave unsaid is which model providers have actually committed to compliance, and what recourse exists when they don't.

Why photographers should hold off on relief

Neither tool, as described, restores revenue already lost to scraping. And neither substitutes for the contractual levers that still work: explicit AI clauses in licensing agreements, watermarked previews, and registration that supports infringement claims.

The pattern is familiar. Platform responses to AI have tended to arrive as opt-outs and dashboards — features that shift the labor of protection onto the photographer — rather than as changes to the defaults that enabled mass ingestion in the first place.

What to watch next

Independent testing of detection accuracy against real-world, partially edited images — composites where a sky was swapped or a subject retouched by AI — will tell us whether these tools are workflow-ready or conversation pieces. Watch too for any named AI companies publicly agreeing to honor the training opt-out. Without that list, the promise is a fence with no posted property.

Photographers integrating detection into client delivery should treat current results as advisory, not evidentiary. The technology is moving quickly enough that verification practices will need revisiting on a quarterly basis.