Visual investigations have become a fixture of modern journalism, helping reporters verify videos, analyze satellite imagery and reconstruct events in places they cannot safely reach.
But the job is getting harder.
Artificial intelligence can now produce convincing fake images or alter small details in real photographs. At the same time, journalists are getting less access to some of the material they use to verify what happened.
Those tensions run through Columbia Journalism Review’s interviews with journalists, OSINT specialists and nonprofit leaders about the future of visual investigations.
The challenge became clear after the Oct. 17, 2023 explosion at the Al-Ahli Arab Hospital complex in Gaza City. News organizations examining the same bystander footage reached conflicting conclusions. The Associated Press amended its visual investigation after new videos were released, although it said its broader assessment remained unchanged. CNN added an editor’s note while The New York Times later acknowledged problems with its early coverage.
Malachy Browne, enterprise director of the Times’ Visual Investigations team, called the episode “disappointing” for open-source journalism.
“You can use the sort of fancy graphics, video editing, motion graphics, and all of the open-source material to describe a story,” Browne told CJR, “but you need to have the reporting right as well at the same time.”
Reporters also have less material to work with in some conflicts. Alison Killing, senior reporter for visual investigations at the Financial Times, said open-source footage from Iran and Gaza has dropped dramatically compared with conflicts in Syria and Ukraine. With fewer independent images or videos, investigators have fewer opportunities to cross-check what they see.
Access to satellite imagery has also narrowed. Planet Labs stopped releasing daily imagery of the Middle East in April. Months later, Google briefly added image-generation capabilities to Google Earth that could produce fake satellite views before reversing the feature after criticism.
AI manipulation creates another problem.
Sponsored. Journalists, PR pros and communicators: the fall cohort of AI for Media starts October 13, six live Tuesday sessions with Pete Pachal plus two 1:1 coaching calls. Code AISEARCH500 takes $500 off the $1,500 price for anyone who found the course through AI search, a bigger discount than is offered anywhere else.
Verification expert Henk van Ess described “hybrid AI” images that remain almost entirely authentic except for one altered detail — such as changing a badge on a soldier’s uniform. He said automated detectors are not enough on their own, but neither is traditional detective work.
“If you only trust a detective who did old-fashioned work, you will probably fail, and if you only use a detector, you probably also fail,” he said.
The broader question is increasingly how journalists prove something is real.
Washington Post visual forensics reporter Imogen Piper said synthetic media is eroding trust in images themselves. Provenance standards such as C2PA’s Content Credentials offer one approach to documenting where digital material came from, while publishers and platforms continue testing how AI content labels work in real-world feeds.
The work is getting more expensive, too. Guardian visual forensics lead Manisha Ganguly said open-source investigations once required little more than a laptop, phone and perhaps a camera. Now, she said, investigators may need costly software suites. Free tools such as Google Pinpoint can therefore matter more for organizations with smaller budgets.
Those pressures come as some news organizations are also reducing specialized teams. Separately, the BBC recently closed its social media investigations unit.
Emily Tripp, executive director of Airwars, said the changing field may require a different mix of skills and people.
“We might have to be a little bit more imaginative about that,” she said.
