Archive · 31 Aug 2026
AI-Enabled Travel-Photo Analysis Gives Scammers Precise Clues
Scammers can use artificial intelligence to study holiday photos posted online.
They may notice signs, buildings, plants or other small clues in a picture.
By combining these clues, AI can sometimes work out where the photo was taken.
McAfee tested two freely available AI models on thousands of travel photos.
One identified the city and country correctly 87 percent of the time, and the other did so 91 percent of the time.
Scammers could use this information to make fake bank, hotel or airline messages seem believable.
Turning off location tags and removing photo metadata can still help protect privacy.
However, people should also think about what the picture itself shows.
If a message creates urgency, users should check directly through the company’s official app or website instead of clicking its link.
McAfee Labs found that AI can often infer a travel photo’s city and country from visual details alone.
In tests, Gemma3 27B reached 87% accuracy, while Qwen3 VL 30B reached 91%.
The research analyzed 21,236 public-domain travel images and separately reviewed 102 additional images.
AI can combine clues such as architecture, vegetation, mountain shapes, menus, signs and streetscapes.
McAfee advises avoiding urgent links and verifying travel, banking or booking messages through official apps and websites.
- Who
- McAfee Labs, including McAfee chief technology officer Steve Grobman, and two freely available AI vision models.
- What
- A McAfee study found that AI can infer travel-photo locations and potentially help scammers personalize phishing messages.
- When
- The articles do not specify when the testing was conducted.
- Where
- The images depicted travel locations, including examples from Lisbon, Mumbai and Goa; the testing used images in the public domain.
- Why
- AI can combine many visual clues in photos to identify locations and create more convincing scam messages.
- Research organization
- McAfee Labs, McAfee’s threat research and intelligence division
- Primary dataset
- 21,236 travel images in the public domain
- Additional review
- 102 images in a separate controlled review
- Model accuracy: Gemma3 27B
- 87% for identifying a photo’s city and country
- Model accuracy: Qwen3 VL 30B
- 91% for identifying a photo’s city and country
- Potential scam context
- Messages may falsely cite unusual activity, card transactions or booking problems during travel
- Recommended response
- Verify messages through the official app or website of the bank, hotel, airline or travel company