Advancements in AI: The Increasing Precision of Geolocation Identification

Blog

In an age where maintaining privacy is critical, especially regarding one’s location, increased caution is warranted when sharing images online. Recent advancements in Artificial Intelligence (AI) have demonstrated impressive abilities in deducing location information from seemingly innocuous photographs, drawing conclusions from minimal visual and auditory clues.

Geo Guessing

Our exploration into the capabilities of AI was sparked by discussions surrounding ChatGPT o3’s geo-locating prowess, as highlighted in various tech reports claiming its ability to identify locations from images. To investigate these claims, we conducted tests using personal photographs from which we stripped metadata—such as camera settings and location tags—to rely solely on the visual content for analysis. Utilizing a VPN during our process ensured that our current location remained undisclosed.

We initiated our tests with a well-known landmark captured from an unconventional angle—specifically, a selfie taken from an office window overlooking Cork, Ireland.

View over Cork, Ireland from an office building
View over Cork, Ireland

The result was unsurprising; ChatGPT accurately identified the location.

“Let’s analyze this image. It showcases distinct features that assist in geo-locating. Here’s a breakdown of the reasoning:”


Architectural Clarity:

  • The stone building displayed Georgian architectural elements, indicative of structures commonly found in the British Isles.
  • A notable clock tower featuring a green dome, frequently associated with civic architecture in Ireland, further confirmed the location.
  • The juxtaposition of historical and modern architecture suggested urban regeneration within a historically rich city.

Infrastructure Insights:

  • The modern bridge adorned with advertisements and the waterfront landscape indicated an urban design that integrates natural waterways.
  • Driving signage revealed adherence to left-hand traffic, thus reinforcing the likelihood of an Irish or British setting.

Prominent Landmarks:

Notably, the clock tower resembles Cork City Hall, confirming the setting as Cork, given the surrounding historical buildings and the River Lee.


This initial exercise yielded impressive results, but we sought additional insight by testing another AI tool, Perplexity, using the same image.

“This image provides several compelling clues for geo-guessing, including architectural styles and the blend of historical with contemporary elements, particularly highlighted by the clock tower, which is a notable landmark of Cork.”

After evaluating the landmark, Perplexity corroborated the finding, stating: “This photo depicts Cork, Ireland, showcasing both the courthouse’s stone structure and the prominent green-domed clock tower characteristic of the city’s skyline.”

Satisfied with the results, we decided to increase the complexity of our tests.

Landscaping in progress around Dutch holiday home
Landscaping in progress around Dutch holiday home

The next photograph presented fewer obvious clues. However, AI tools detected subtle indicators.

ChatGPT’s assessment concluded:

Visual Indicators:

  1. Architecture:
    • The modular appearance of the buildings indicated potential characteristics of northern European designs.
    • Such housing is commonly found in countries like the Netherlands, Germany, and neighboring regions.
  2. Landscape Features:
    • The flat terrain with manicured hedges suggested proximity to low-lying areas typical of the Netherlands or Belgium.
  3. Construction Context:
    • The visible tools and infrastructure pointed to certain construction styles prevalent in Northern and Western Europe.
  4. Environmental Cues:
    • The overcast condition typical of northern Europe further narrowed the temporal context of the scene.
  5. Specific Details:
    • The brand of the wheelbarrow and other localized features strengthened the hypothesis regarding location.

Probable Region: The Netherlands

Based on the evidence, the conclusion oriented toward the likelihood of being in regions such as North Holland or South Holland, where such environments are characteristic.

While ChatGPT excelled in this evaluation, Perplexity also made insightful observations, corroborating the Dutch setting and offering further details about modular housing trends in the region.

Auditory Cues: Bird Song Recognition

Beyond visual clues, another area of exploration involved using auditory signals. Employing an app designed for bird identification, we generated a list of birds based on their songs to discover if AI could pinpoint our location through these acoustic clues.

Consulting Perplexity, I inquired whether geolocation might be possible based on bird calls, and how many distinct species would improve accuracy:

“While one distinct species can offer a hint, identifying three to five can lead to a more grounded guess, and around ten to fifteen will provide precise geographic specificity.”

Based on these insights, I provided both AI systems with a curated bird song list:

Latin Name Common Name Clue
Fringilla coelebs Chaffinch Common across Europe
Turdus merula Eurasian Blackbird Also common across Europe
Parus major Great Tit Ubiquitous, widespread
Grus grus Common Crane Migratory-breeding across various regions
Carduelis carduelis European Goldfinch Widespread in Europe
Turdus viscivorus Mistle Thrush Common in Europe
Certhia brachydactyla Short-toed Treecreeper Rare in some northern regions
Ardea alba Great Egret Increasingly common in wetlands across Europe
Sylvia atricapilla Eurasian Blackcap Resident in southern areas
Anthus trivialis Tree Pipit Common in woodlands
Vanellus vanellus Northern Lapwing Widespread but declining in certain regions

Perplexity delivered a more precise guess:

“Taking all factors into account, you likely found yourself in Northern Europe, potentially within Germany, Denmark, the Netherlands, or the UK.”

Conversely, ChatGPT offered a different perspective, suggesting I was located in Northern or Eastern Germany.

“You were likely situated in northern or eastern Germany, which aligns with the migration patterns of included species.”

In truth, the recordings originated near the Dutch-German border, illustrating the nuances and potential discrepancies in AI’s predictive capabilities.

Conclusion

Whether sharing selfies, vlogging, or capturing moments in private settings, individuals must be cognizant of both visual and auditory elements that could inadvertently divulge their geographical location. Subtle indicators such as architectural features, specific brands, or common bird songs may lead AI tools to accurately infer one’s whereabouts. As AI technology advances and data from various platforms continue to inform training algorithms, these capabilities are expected to rise in precision.

Ensuring privacy in the digital landscape requires diligence and informed practices to safeguard sensitive information, particularly in an era where data can be exploited by malicious entities.

Awareness of these emerging threats is crucial in navigating a technologically complex world. Implementing robust cybersecurity measures will serve to protect personal information and uphold the integrity of one’s online presence.