Posting on social media? AI may help others track your location

A new study has found that AI models can track your location solely based on your social media posts as much as 9 out of 10 times.

In Short

• AI tools can detect your location from photos, study finds
• This may allow a scammer to track your city
• It may allow for targeted emails or messages

Social media has become a big part of our lives today. Be it making Reels or sharing snippets from a vacation, most of us don’t think twice before posting something online. But a new study has found that anyone may be able to track your location from these posts with AI tools.

A study by McAfee has found that AI tools can identify where many photos were taken even without GPS data, location tags or captions. The company says that AI was able to identify the correct city and country almost 9 out of 10 times.

The models used in the study were Google’s Gemma 3 27B and Alibaba’s Qwen 3 VL 30B. Both models are open-weight and can be used freely without needing any subscription.

How does AI detect your location?

For the study, the company would attach a photo and write a simple prompt to the AI, asking it to identify the location. No metadata, EXIF data or file names were used, and human reviewers checked edge cases where place names could be interpreted in more than one way.

Out of 21,236 travel images, Gemma 3 27B identified the city and country correctly 87 per cent. of the time, while Qwen 3 VL 30B reached 91 per cent. The models identified locations by reading details in the image itself, including architecture, signage, skylines, light, street markings, storefronts and even food stalls.

As per the study, images taken at well-known tourist sites or places with distinctive visual features were easier to place, while generic beaches, rural roads and hotel rooms were harder. Even then, the AI was often still able to identify the country.

Outside this study, employees uploaded personal travel photos, including some that had never been posted publicly, to tools such as ChatGPT, Claude and Copilot and asked where the images had been taken. In one example, AI identified a river scene with trees as Hastings-on-Hudson, a small village, in the US state of New York.

Can this help scammers?

The study says that this can help scammers move from generic phishing to more personalised fraud. A scammer could take public travel photos from platforms such as Instagram, Facebook or X, run them through an AI model, get the likely destination and timing, and then send messages referring to that place.

For instance, someone may send you warnings about unusual bank activity while you are on a trip, or alerts about an attempted login from that location. These messages do not need to be perfectly accurate, only plausible enough to lower suspicion.

At a time when cybercrime is on the rise, it may make things easier for scammers. According to the Ministry of Home Affairs, 28.15 lakh cases were recorded in 2025, a 24 per cent increase from 22.68 lakh in 2024. The ministry says that Indians lost Rs 22,495 crore from cybercrimes last year alone.

The next time you share photos on social media, you may want to keep this in mind. Though AI models may not be as good at identifying your exact location if the photo is not near a prominent landmark.

- Ends