Google’s newly deployed AI image detection tool has been shown to be unable to determine with certainty whether a doctored photograph of a crying activist was created or modified by its own artificial intelligence, exposing the limitations of current safeguards as synthetic media proliferates online.

The incident, highlighted by user experiments and independent tests, underscores persistent challenges in verifying the origin of altered images even when detection features are in place.

The image in question, which circulated widely on social platforms, was run through Google’s detection system using its Gemini interface, which is designed to identify AI-generated content by searching for hidden digital markers known as SynthID and analysing visual clues embedded in images.

However, when presented with the doctored photo, the system returned an inconclusive result, reflecting the broader issue that detection tools are often optimized for their own models and may not reliably distinguish between authentic photography and synthetic manipulations created by complex AI workflows.

Google’s capabilities are strongest at recognising images marked with its own SynthID watermarks, but outside those parameters, confidence drops significantly.

Experts say the episode highlights a fundamental difficulty in the current landscape of AI image detection: while tools can signal when an image was generated or edited by a specific proprietary model, they struggle to assess content created by hybrid processes or multiple systems.

The design of Google’s detection framework, which focuses on its own embedded identifiers, means it may miss or remain uncertain about synthetic media originating from other sources or modified after initial creation.

This limitation raises concerns about the reliability of automated verification at a time when AI-assisted manipulation is rapidly advancing.

The inconclusive result from Google’s own detector comes amid broader industry debate about how to enhance media authenticity tools and promote trustworthy information online.

While Google and other tech companies have rolled out transparency features to help users identify AI involvement in images and videos, the case of the crying activist image illustrates that no singular solution currently exists to definitively authenticate all forms of visual media.

Observers say that improved standards, broader interoperability between detection systems and greater contextual metadata are necessary to strengthen the ability of platforms and the public to differentiate between genuine and AI-manipulated content as generative technologies continue to evolve.