As artificial intelligence (AI) technology advances, news organizations are intensifying efforts to verify photo and video content to maintain editorial integrity. The Globe and Mail has recently updated its verification protocols to address the growing sophistication of AI-generated images and videos, which have become increasingly difficult to detect.
Patrick Dell, the paper’s deputy head of photo and video specializing in visual verification, noted that AI-generated content has become pervasive and subtle, blending seamlessly into the digital landscape. A notable example occurred in July, when a doctored photo purportedly showing U.S. Senator Mitch McConnell on life support circulated widely before being debunked by fact-checkers. Such instances highlight the challenges newsrooms face in distinguishing between authentic journalism and manipulated visuals.
The Globe and Mail’s Editorial Code of Conduct mandates that news images be documentary in nature and prohibits manipulation, distortion, or compositing during or after capture. In response to the rise of AI content, the publication now employs a multi-layered verification process. This includes cross-referencing images taken at the same location and time, confirming the verified presence of subjects, and utilizing AI detection tools such as Hive AI Detector. Satellite imagery is also used to corroborate events and locations depicted in images.
While deepfakes have drawn significant attention, other forms of digital alteration have also been identified. For example, earlier this year, the paper’s visuals team detected unauthorized editing in real estate photos intended to make rental properties appear more attractive. One case involved the cloning of a flower bed to enhance a listing image, a practice common in marketing but at odds with editorial standards. When such manipulation was uncovered post-publication, the story was updated to exclude the altered image, and a correction was issued.
To prevent similar issues, The Globe and Mail has issued detailed AI photo policies to editorial contacts, particularly in sectors like real estate where visual enhancements are frequent. These guidelines explicitly forbid the use of generative AI for creating, editing, or enhancing visuals without disclosure. Prohibited actions include adding or removing image elements, compositing multiple photos, upscaling, generative fills, background extensions, and any adjustments that alter the image’s intended meaning.
Editors and reporters have noted cooperation from external contributors upon receiving these guidelines. However, challenges persist; a recent real estate photo was flagged for unrealistic clarity seen through a window, indicative of editing beyond natural optical limits. In another instance, a photographer unknowingly altered an image intended for editorial use, leading to reshoots after original files were no longer available.
The Globe distinguishes between editorial and non-editorial content, allowing more flexible editing for non-news assignments such as magazine portraits, with clear freelancing policies that prohibit generative AI use. The paper emphasizes transparency and rigorous fact-checking to uphold trust in news media.
As digital content proliferates across social media platforms without similar verification measures, readers are urged to approach visually striking material with caution. Verification resources, like the “Reporter’s Guide to Detecting AI-Generated Content,” are available to assist the public in identifying manipulated media. Dell underlines the importance of skepticism, noting, “Unless it’s a website that I know, I don’t take anything at face value.”
In an era where AI complicates the visual truth, the responsibility of verifying what audiences see increasingly falls to journalism.