"In the last weeks of 2025, as I write this prediction, my social media feeds are awash with a new set of pronouncements about the death of photographic truth. As a historian of photography, I'm familiar with obituaries of this sort, which tend to circulate in popular media at moments when new photographic or image-editing technologies arrive on the scene that force us to reevaluate photography's capacity to represent reality accurately."
"Nano Banana Pro has been described as creating images that are "hard to distinguish from the real thing" and, according to one tech reporter who couldn't resist the low-hanging fruit, "so realistic it's bananas." Online responses to its release have tended to combine awe at its seeming accuracy and convincingness in creating photographic-looking images with a raft of concerns about its potential for deceptive use."
"Like other means of AI image generation, Nano Banana Pro is a technology that is existentially distinct from photography. Rather than depicting people, things, or environments from life, as photography does, AI image generators fashion images by drawing on massive datasets. They are, we might say, regurgitative media. For this reason, most news institutions' embrace of AI-generated imagery will be limited in 2026 to those realms of journalism where images aren't expected to faithfully capture news events."
Social media in late 2025 is filled with proclamations about the death of photographic truth following the release of Nano Banana Pro. Such obituaries have historically coincided with the arrival of new photographic or image-editing technologies that force reevaluation of photography's capacity to represent reality, with Photoshop's 1990 launch as a notable example. Nano Banana Pro, Google's newest AI image-generation tool, produces images hard to distinguish from real photographs, provoking both awe and concerns about deceptive uses. AI-generated imagery fashions images from massive datasets rather than depicting life, making it regurgitative. Newsrooms will likely limit its use to illustrative domains where strict accuracy is not required.
Read at Nieman Lab
Unable to calculate read time
Collection
[
|
...
]