How good are LLMs at geoguessing?

Quite good, per Bellingcat [0] — Google Lens and ChatGPT could localize the majority of their test photos pretty specifically.

[0] https://www.bellingcat.com/resources/2025/08/14/llms-vs-geol...

▲

jillesvangurp 3 hours ago | parent | prev | next [-]

Pretty good. I played a bit with gpt-4 a year or so ago by feeding it random screenshots from Google street view. It will pick up a lot of subtle hints from what otherwise looks like generic streets. I imagine more recent models might be better at this now.

▲

firtoz 3 hours ago | parent | prev | next [-]

Pretty good. I test it every now and then from random photos. Sometimes spot on, sometimes gets very close, unless it's really ambiguous.

▲

embedding-shape 4 hours ago | parent | prev [-]

Basically all up to the training data, as things often are.

▲

xg15 3 hours ago | parent | next [-]

I wonder if that might be another reason to just completely disable this feature and not make it a permission: otherwise people could use it to build trainingsets for geoguesser models.

	▲	GRiMe2D 3 hours ago \| parent [-]
		People already uploaded tons of images and data while playing Pokemon GO. Probably model is already has been built and being tested right now

▲

eru 4 hours ago | parent | prev [-]

You still need some smarts, since the picture you just took won't be in the training data.