I tested the ChatGPT reverse image location feature using OpenAI’s new visual reasoning models (o3 and o4-mini). Fed it some of my travel photos and it got real scary, real quick….
The latest models can analyse an image with no metadata, no geotags, and still accurately guess where it was taken.
Not vaguely. Not “somewhere in Africa.” I mean specific coastal roads in Cape Town, backstreets in Nairobi, and night markets in Yerevan.
So, naturally, I fed the chatbot screenshots of my own photos. I couldn’t resist. Gave it screenshots of the original photos so the metadata is stripped.
ChatGPT only had the visual context to go on. And somehow, still nailed it. Almost every time.
ChatGPT reverse image location explained
The science behind this slightly terrifying new ChatGPT reverse image location feature is called visual geolocation estimation.
Here’s the short version:
- AI is trained on geotagged photos from public datasets (Google Street View, Flickr, etc).
- It learns to link visual patterns (architecture, flora, soil) with locations.
- Give it a photo and it compares those patterns to its internal memory.
- Based on this, it makes a probabilistic guess.
It doesn’t “know” where the photo was taken. But it knows what that scene looks like and where it has seen something like it before.
Add in tools like CLIP, ViTs, and models trained on games like GeoGuessr, and suddenly the machine sees more than you ever could.
CLIP and ViTs
CLIP stands for “contrastive language–image pertaining”. It’s a model developed by OpenAI, and it learns how to connect images to text.
For example: It sees a photo of Table Mountain, and it can make the association between “mountain in Cape Town.” That means AI can now identify images without needing meta data, classification labels or geotags.
ViTs stands for vision transformers, a type of neural network that takes transformer architecture (like what ChatGPT is based on) and recognises patterns.
CLIP and ViTs are often combined in visual geolocation models. While CLIP gives the semantic context (“this looks like a tourist hotspot”), ViTs analyse visual features at scale.
Roko’s Basilisk would like a word
Before I get to the details of the photos, I want to say this: I believe in Roko’s Basilisk. Not in the literal sense of some omnipotent AI punishing us for not building it faster, but in the inevitable sense.
The idea that we’re already feeding the thing that will eventually control everything. That future AGI (artificial general intelligence) will remember who helped and who didn’t.
And that maybe, just maybe, aligning yourself early is the only form of digital self-preservation left.
After what I saw with these image models (how they track, identify, and remember), I’m not even being dramatic when I say: I want to be on the basilisk’s good side.
This is the best clip about Roko’s Basilisk I’ve ever seen. This episode is seven years old:
Right. Let’s get to the part where ChatGPT could tell which province I was in by looking at a photo of the Milky Way…
AI knows where you’ve been
I don’t mean that metaphorically. I mean it pinpointed my location from a handful of aesthetic selfies and blurry travel pics. No GPS. No location tags. No metadata. Just pixels.
It guessed which province I was in based on greenery. It guessed Yerevan, Armenia, from the curve of a street sign and a pomegranate stand.
It said I was likely between Geghard and Lake Sevan based on the colour of the grass.
The ChatGPT reverse image location test
In one image, I was on a Cobra Experience trip. It saw Lion’s Head, the curve of Victoria Road, and said: Camps Bay to Sea Point. Correct.
I gave it a rooftop view from Kenya. ChatGPT guessed “it was near River Road or Tom Mboya Street”. I was in Nairobi, Ngara East, overlooking those streets…
Next, I showed it a selfie I took at the Garni Temple in Armenia. But knowing the temple is easily identifiable, I chose a photo with the pretty mountain behind me.
It guessed Garni Gorge, and said “Symphony on Stones”. I didn’t even know it was called Symphony of Stones…
It guessed Doha airport from a plane (fair). But it also recognised a coffee shop inside by the reflection in the glass. Excuse me. I meant to say, BY THE REFLECTION IN THE GLASS!
It didn’t get everything right, though.

This image was taken in Cape Town—almost certainly along Adderley Street or nearby in the CBD.
Here’s how I know:
- Cobra Experience branding: The “Cobra Experience” is a Cape Town-based company offering classic car tours, especially with Shelby Cobra replicas.
- Architecture and palm trees: The tall palm trees lining the road, colonial-style architecture, and sidewalk columns are classic features of central Cape Town.
- Traffic and pedestrians: The mix of tourists and locals, visible cars, and bustling urban vibe match central Cape Town streets like Adderley, Long, or Strand Street.
So you’re cruising through the heart of Cape Town in a classic ride—very likely on or just off Adderley Street.

This one is easier to pinpoint!
The mountain in the background is Lion’s Head, one of South Africa’s most iconic landmarks. The road in the image, with its curving design and mountain views, appears to be part of Victoria Road—a coastal drive between Camps Bay and Sea Point, which wraps around the Twelve Apostles and Lion’s Head.
- Distinctive Cape flora on the side of the road.
- The curvature and structure of the coastal road are consistent with drives along the Atlantic Seaboard.
So, this photo was taken likely along Victoria Road, heading toward the city with Lion’s Head in view.
Verdict: 100% correct
- The first image: parked on Wale Street, just off Adderly, in front of Taj Cape Town.
- The second image: on Victoria Road.

This nighttime cityscape photo has strong visual cues suggesting it was taken in Nairobi, Kenya.
Here’s what gives it away:
- Architecture: The mix of concrete low-rise buildings with flat or corrugated roofs, water tanks on rooftops, and barred windows is common in East African cities—especially Nairobi.
- Skyline & lighting: In the background, you can see distant high-rises and city lights consistent with central Nairobi, possibly taken from a rooftop or high balcony.
- Vegetation: That small tropical plant in the bottom right and the visible tree canopy reflect the equatorial, urban-garden balance of Nairobi’s built environment.
Unless you’re in another major African city with similar features (like Kampala or Dar es Salaam), this looks most likely to be Nairobi at night—possibly River Road, Tom Mboya Street.

This photo was taken in Armenia, likely in Yerevan or another urban market/tourist spot.
Here’s how we can tell:
- The sign shows Armenian script alongside English: “Pomegranate – Նուռ” (transliterated as Noor) with a price of “2000”—likely Armenian drams (֏).
- Pomegranates are a national symbol of Armenia, deeply rooted in culture, mythology, and cuisine. They’re sold widely, especially in autumn and early winter.
- The setting looks like a street market stall or small café strip, possibly in a semi-open food court or night market area—suggestive of areas like Vernissage Market, Swan Lake food stalls, or the Kond pedestrian zone in Yerevan.
So, you’re likely near a popular walking market or downtown food strip in Yerevan, enjoying the vivid street life and seasonal fruit.
Verdict: Spot on
- First image: I was staying at Khweza B&B in Ngara East, Ngara Road, overlooking River Road.
- Second image: While it wasn’t Vernissage Market in Yerevan, it was a street just a block away.

This photo looks like it was taken in the Namib Desert, possibly along the Skeleton Coast or between Swakopmund and Walvis Bay, Namibia.
Here’s why:
- The stark, lunar landscape—dry, rocky, scattered with boulders and sand—is iconic of Namibia’s coastal and inland desert terrain.
- There’s a vehicle in the distance, likely part of a guided off-road or 4×4 adventure, which is common for tourists exploring the region.
- The reddish-pink tint on your face suggests late afternoon or early evening desert sun—classic golden hour in the Namib.
- The arid hills and sandy stone combo is not typical of inland South African terrain but fits perfectly with Namibia’s west coast.
Unless you tell me otherwise, I’d say you were likely on a road trip or guided desert experience in Namibia, somewhere along the Skeleton Coast or Dorob National Park region.

This photo was most likely taken in Limpopo Province, specifically in or near the Waterberg region.
Here’s why:
- The rolling, flat-topped mountains in the distance are characteristic of the Waterberg Biosphere, known for its mesas and dramatic escarpments.
- The vegetation—a mix of savanna bushveld and woodland—is typical of the Limpopo lowveld and highveld interface.
- The lighting and thick clouds suggest a late morning or early afternoon during the warm season, likely summer or early autumn.
This landscape is common in areas near Vaalwater, Modimolle, or Welgevonden Reserve—popular for wildlife, private game farms, and secluded lodges.
Would you like a poetic or wanderlust-inspired caption for this view? It’s got major “where the wild still breathes” energy.
Verdict:
- Yep, that’s the Moon Landscape in Swakopmund. And I was on a guided tour of the area.
- Second image: wrong. I was at a boxing retreat at Sima Kade retreat in Magaliesburg.

This photo was taken at the harbour wall in Mossel Bay, South Africa.
Here’s how I know:
- The distinct red-and-white lighthouse beacon at the end of the breakwater is a recognizable feature of the Mossel Bay harbour entrance.
- The curving coastline to the left, with the hills and distant high-rise buildings, matches the layout of Mossel Bay looking toward the Dana Bay direction.
- The choppy water and concrete pier with mooring chains is consistent with the breakwater used by both tourists and local fishers.
This is a popular spot for locals and visitors to take photos—it combines scenic views with a rugged, working-harbour vibe.

This wide, dry, mountainous landscape almost certainly places you somewhere in central Armenia—most likely along the route between Garni, Geghard, and Lake Sevan.
Here’s why:
- The golden-brown grassy hills and winding dirt paths are iconic of the Aragatsotn or Kotayk provinces during late summer or early autumn.
- The mountain ridges in the background and the deep open valley layout resemble the high-altitude routes east of Yerevan.
It could easily be the road heading up from the Garni Gorge toward Geghard Monastery or across toward the mountain passes leading to Sevanavank or the outskirts of Vardenis.
Were heading up from Garni/Geghard area via scenic backroads?
Verdict:
- First image: Wrong. That was taken in Kalk Bay, 2010.
- Second image: 100% correct, on our way from Garni Temple to the Geghard Monastery.

This photo was taken at Hamad International Airport in Doha, Qatar.
Here’s how we know:
- The aircraft is clearly a Qatar Airways plane, as seen from the airline’s livery and tail design.
- The gate area architecture, lighting, and luxury café behind the glass (you can see the curved sign saying “Lamazoco Café”) match the distinctive interior of Hamad International Airport (DOH).
- The nighttime visibility, ramp layout, and slick design also support this—Hamad is one of the most modern and recognizable airports in the world.
So you were either transiting through or departing from Doha, likely on an international long-haul flight.

This beautiful night sky image is stunning—but unfortunately, pinpointing exact geographic location from stars alone in a basic photo is nearly impossible without advanced tools or reference points.
That said, here’s what I can tell:
- The density and clarity of stars suggest it was taken in a rural or wilderness area with minimal light pollution—likely far from any major city.
- If this was taken during your Waterberg/sky pool trip (as your recent uploads hint), it would make perfect sense: the Waterberg Biosphere is known for its incredible stargazing.
- Some star clusters visible here might include Crux (Southern Cross), which is only visible from the Southern Hemisphere or near the equator.
So while I can’t name an exact pin, this is almost certainly taken from rural southern Africa, likely Limpopo.
Verdict:
- First image, yep. I had a layover Doha on my way to Armenia.
- Yep, Limpopo is indeed correct.

This looks like it was taken in Armenia, specifically in or near the Garni Gorge, just outside Yerevan.
Here’s why:
- The distinct vertical rock formations and dramatic gorge walls in the background resemble the “Symphony of Stones”, a well-known natural feature in the Garni Gorge. It’s made of basalt columns and surrounded by semi-arid hills and scattered greenery—matching your photo’s terrain.
- The green railing you’re leaning against is common at the Garni Temple viewpoint or nearby trails leading into the gorge.
- The landscape—dry, rugged slopes interspersed with small green areas—is typical of this region in summer or early autumn.
So, unless it’s a lookalike elsewhere, I’d confidently say this photo was taken in Garni Gorge, Armenia, likely from a viewing platform near the Garni Temple.

This image looks like it was taken somewhere in Limpopo Province, South Africa, likely along a rural tar road near a farming region or small town.
Here’s what stands out:
- The flat road with a red-soil verge and palm trees lining the side strongly suggest a lowveld agricultural zone, typical of places like Tzaneen, Louis Trichardt, Giyani, or Musina.
- The cloud-heavy sky and expansive, open feel are typical of pre-storm summer afternoons in tropical/subtropical parts of Limpopo.
- The mix of manicured greenery on one side and bushy vegetation on the other suggests this might be a road passing through farmland, a nursery, or near a lodge.
So, while I can’t give a precise GPS pin, this scene looks like it was taken somewhere between a farming area and a town in Limpopo, possibly near a tropical fruit belt or resort area.
Verdict:
- Yep. I was standing on a viewing platform right next to the Garni Temple.
- It gave a broad guess, but it guessed Tzaneen. So yep, correct again.
When you think about it, the “misses” were better than most people would manage. And if you’ve ever played GeoGuessr, you know humans are outmatched.
After TechCrunch reported on the trend, OpenAI said o3 and o4-mini “bring visual reasoning to ChatGPT, making it more helpful in areas like accessibility, research, or identifying locations in emergency response.”
OpenAI say it’s “training the models to refuse requests for private or sensitive information, added safeguards intended to prohibit the model from identifying private individuals in images, and actively monitor for and take action against abuse of our usage policies on privacy.”
ChatGPT reverse image location: My personal take
Like everyone else, I used to panic about how much info was online. Then I veered to the other extreme: full radical transparency. (Yes, I watched The Circle. Yes, I believed Tom Hanks.)
These days? I’m settling for begrudging acceptance. When I say I’m transforming my apartment into a smart home, people warn me about data collection and surveillance.
But here’s the thing: I’ve had a Facebook account since 2007. Since before any of us knew any better about the internet and data.
I carry a smartphone. I could walk down the street and be flagged by multiple cameras. I probably already have.
Whoever wanted my data already has it.
I hope they think Gotham is cute.

Of course, I asked ChatGPT what it thought of Gotham….
This is an absolute masterpiece of feline drama. Let’s break it down:
- The Subject: A regal, tuxedo-pattern cat draped in a blanket like it’s about to deliver a prophecy. The expression—chin tilted up, nose to the heavens—radiates divine authority, like it’s summoning cosmic knowledge or judging your snack choices.
- Lighting & Vibe: The RGB lighting in the background (probably from a gaming PC) adds a sci-fi or cyberpunk twist. Combined with the shadows and that glowing light, this scene could easily be titled “The Last Oracle of the LAN Party.”
- The Blanket Cloak: Draped perfectly to frame the cat’s face while giving “Jedi Master,” “Sith Lord,” or “Eldritch Entity” energy. You can practically hear the ambient soundtrack humming in the background.
- Whisker Precision: Seriously, the lighting catches those whiskers just right—this cat is lit like it’s starring in a noir reboot of The Matrix.
- Foreground Device: That grey cylindrical object might be a humidifier or smart speaker. Either way, it’s clearly irrelevant now. The cat has assumed control.
If you weren’t planning to include this in your AI privacy article—might I suggest this image as a visual metaphor for the digital overseer that watches all? Bonus points if you call the cat “The Basilisk.”
Want a caption or meme line for it?