Free Press Election Night Photo Sparks SynthID Detection Dispute

According to Vanity Fair, a photo of an election night event in Michigan posted by The Free Press reporter Olivia Reingold was suspected of being AI-generated. After reporter Scott Nover reproduced the check with Google SynthID and got a positive result, the news industry quickly erupted in questions about the image's authenticity. Reingold responded that she had only used ChatGPT to "brighten and sharpen" the photo, not generate it; however, the two original images she shared both came back without a detected watermark under OpenAI's SynthID check.

Confirmed

• Scott Nover's SynthID test was positive, but he later clarified: a positive SynthID result only means the image was processed with OpenAI tools in some way, not that the entire image was AI-generated from scratch—exactly which parts were generated or altered remains unknown.
• Reingold admitted using ChatGPT to brighten and sharpen the photo, arguing that afterward it was "still a real woman in Valentino shoes carrying a Chanel bag."
• Both originals Reingold shared failed to trigger a warning under OpenAI's SynthID detection.
• Reporter Drew Harwell strongly advised colleagues against using AI tools to enhance news photos, recommending traditional methods like brightness adjustment instead.

Unconfirmed

• It remains unclear which parts of the photo, if any, were AI-generated or modified; the fact that the "originals" didn't trigger the watermark also calls SynthID's detection reliability into question.

Why it matters

• The episode shows that conclusions from AI watermark detection tools must be read carefully: a positive result doesn't mean the whole image was generated, and a negative result doesn't fully rule out AI processing.
• Scott Nover's reminder to journalists: even well-intentioned AI processing of news photos can undermine their credibility, and news organizations need clear policies on AI post-processing.

2026-11-06 ~ 2026-11-06 · 5 related posts

• [source] Vanity Fair flags Free Press election-night photo as AI-generated, confirmed via SynthID — ScottNover · 2026-11-06
• [source] SynthID hit only means an image was processed by OpenAI tools, reporter clarifies — ScottNover · 2026-11-06
• ChatGPT-sharpened photos dodge SynthID, exposing AI watermark detection gaps — ScottNover · 2026-11-06
• [source] Journalist's ChatGPT-Brightened Photo Sparked Authenticity Doubts — a Cautionary Tale — ScottNover · 2026-11-06
• Journalists warned: never AI-enhance news photos, adjust brightness instead — ShakeelHashim · 2026-11-06