2027: Google opens AI watermark detector to public

Google has opened its artificial intelligence watermark-detection platform to the public as concerns grow about manipulated images, videos and audio ahead of Nigeria’s 2027 general elections.

The company announced the expansion in a statement sent to PREMIUM TIMES, saying the SynthID Detector is now available globally in English.

The platform allows users to upload media files and check whether they contain invisible watermarks placed by supported AI-generation systems.

It can currently identify SynthID watermarks in content created using tools operated by Google and participating companies, including OpenAI, NVIDIA and Kakao. Apple is expected to join the initiative later.

Users can access the service with a Google, Apple or OpenAI account, according to the statement.

Google said the platform was designed for everyday users, including journalists, fact-checkers and members of the public who need additional information when examining suspicious online content.

Election relevance

The public rollout comes as political activity intensifies ahead of Nigeria’s 2027 elections and concerns increase about the possible use of artificial intelligence to manufacture political statements, impersonate candidates and manipulate public opinion.

AI tools can now produce realistic photographs, cloned voices and videos showing people appearing to make statements or perform actions that never occurred.

Such content can circulate rapidly on social-media platforms and closed messaging services, sometimes reaching large audiences before journalists, officials or fact-checkers can establish its origin.

The Independent National Electoral Commission has identified the weaponisation of AI-generated disinformation and social media volatility as threats to the credibility and security of Nigeria’s electoral process.

Google says the detector could therefore provide another verification tool for Nigerian journalists and voters facing questionable campaign footage, purported speeches, fabricated endorsements, or manipulated audio recordings.

How it works

SynthID is not a conventional visible label or logo placed on a photograph or video.

Instead, it embeds a digital signal directly into AI-generated content at creation. The watermark is designed to be imperceptible to people but identifiable by Google’s detection technology.

For images and videos, Google says the watermark can remain detectable after common alterations such as cropping, filtering, compression and changes to video frame rates.

For audio, the signal is embedded in a way that is inaudible to listeners and designed to withstand changes such as added noise, speed adjustments, and MP3 compression.

After a file is uploaded, the detector scans it for a recognised watermark and reports whether supported AI technology created or altered the content.

Google first introduced SynthID in 2023 and subsequently tested an early version of its detection portal with journalists and media professionals. The latest expansion makes the platform available to the wider public.

Not a universal AI detector

Despite its potential usefulness, the tool cannot determine the origin of every piece of AI-generated content online.

SynthID primarily checks for watermarks embedded by Google and participating technology companies. Content produced with an unsupported AI model or generated without a compatible watermark may not be identified.

Therefore, a result indicating that no SynthID watermark was found should not be interpreted as proof that the file is genuine or entirely human-created.

The detector also does not independently establish whether the events depicted in authentic footage have been correctly described. A genuine photograph or video can still be shared with a false date, location or caption.

Journalists and other users will consequently need to combine the tool with established verification methods, including checking the source, examining metadata, conducting reverse-image searches, comparing landmarks and contacting the people or organisations involved.

Scale of AI-generated material

Google says it has embedded SynthID watermarks in more than 180 billion AI-generated images and videos since introducing the technology.

It has also watermarked about 240,000 years of generated audio.

READ ALSO: 2027: Ex-INEC commissioner raises concerns over AI use in results management

The company’s other verification features are already integrated into Google Search, Gemini and Chrome. According to Google, those services collectively handle more than one million media-verification requests daily.

SynthID also supports watermarking for text generated through certain Gemini products, but Google’s publicly announced detector is currently presented primarily as a verification platform for images, videos, and audio files.

Google said the initiative is part of its effort to give users more context about the content they encounter online.

For Nigeria, Google says the platform offers an additional line of inquiry at a time when election-related misinformation is becoming increasingly sophisticated. It does not replace conventional fact-checking, but it may help users identify some synthetic media before sharing or acting on it.