SecurityNot yet confirmed elsewhere1 publisher2 min readPublished
Graylark's Raven traced interior photos to named Airbnb, Vrbo and Zillow listings in its own demos
Graylark Technologies says its Raven tool can geolocate interior images across millions of short-term rental units, according to CEO Daniel Heinen. 404 Media previously reported that LAPD and Miami-Dade Sheriff's Office bought Raven access, and the interior matches it checked come from Graylark's own demo videos.
The Watch · Security desk
Map: the Los Angeles Police Department bought Raven access. Miami-Dade Sheriff's Office had a custom model trained for Miami-Dade County. Demos matched a Charlotte Airbnb listing, a Montecito Vrbo listing and San Francisco photos posted on Zillow. The interface says city-specific property AI.
- decision Los Angeles Police Dept. Bought access to Raven, which was previously called GeoSpy, claim 3
- capability Miami-Dade Sheriff's Office Emails showed it had access to a custom Raven model specifically trained for Miami-Dade County, claim 12
- exposure Charlotte Airbnb listing A washer/dryer photo from X returned a top result quoting this listing, whose photos show the same property, claim 4
- exposure Montecito Vrbo listing A bathroom photo from social media returned a top result that directly quotes this Vrbo listing, claim 5
- exposure San Francisco address An interior room photo was traced to it; the photos returned were exactly the same as those posted on Zillow, claim 9
- constraint Raven interface Reads "City-specific property AI to determine an address," implying a model built for each city, claim 11
| Who | How | Kind | Claim |
|---|---|---|---|
| Los Angeles Police Dept. | Bought access to Raven, which was previously called GeoSpy | decision | 3 |
| Miami-Dade Sheriff's Office | Emails showed it had access to a custom Raven model specifically trained for Miami-Dade County | capability | 12 |
| Charlotte Airbnb listing | A washer/dryer photo from X returned a top result quoting this listing, whose photos show the same property | exposure | 4 |
| Montecito Vrbo listing | A bathroom photo from social media returned a top result that directly quotes this Vrbo listing | exposure | 5 |
| San Francisco address | An interior room photo was traced to it; the photos returned were exactly the same as those posted on Zillow | exposure | 9 |
| Raven interface | Reads "City-specific property AI to determine an address," implying a model built for each city | constraint | 11 |
What happened
- In a demo video, a washer/dryer photo taken from X resolved to Charlotte, North Carolina, and the top result quoted a specific Airbnb listing whose photos showed the same property.
- A bathroom photo from social media returned a Montecito, California result titled "Montecito Miramar Beach House on the Sand," which 404 Media found quotes a Vrbo listing.
- A demo posted last year traced an interior room photo to a San Francisco address, and the interior photos returned matched those on Zillow exactly, according to 404 Media.
- Emails seen by 404 Media showed Miami-Dade Sheriff's Office had access to a custom Raven model specifically trained for Miami-Dade County.
Why it matters
- exposure A photo posted from inside a rental can be tied to its listing. The demo images came from X and social media, and 404 Media says officers could source theirs from anywhere.
- capability Stripping GPS and other metadata from a photo would not stop the match, because 404 Media says Raven finds the home without using any metadata attached to the image.
- contradiction Heinen's post says "worldwide." The interface text says "City-specific property AI," which implies a separate model per city, so coverage in practice may depend on where Graylark has trained.
- precedent Heinen wrote that in 2026 Graylark is "investing heavily in interior geolocation across U.S. hotels, short-term rentals, houses, and apartments," so the reference set is meant to grow past rentals.
Raven is not a reverse image search, according to 404 Media. A reverse image search finds an identical image posted elsewhere. Raven identifies unique characteristics in a photo and then finds what are likely other photos of the same property, even if those photos are not exactly the same or time has passed between them [7].
Heinen frames the purpose as casework on trafficking and exploitation. He wrote: "For investigators working trafficking and exploitation cases, that connection can turn an unidentified room into a location lead they can verify and pursue." [8]
The interior ability was not planned. "I'll be honest, we didn't train it to do that," Heinen told Indicator last year [14]. He said GeoSpy 1.0 could not do it, and the ability appeared as the company trained and tested GeoSpy 2.0 [15]. The model was trained on millions of images from around the world and first focused on markers such as architecture and soil [16].
In our view the reach follows what hosts and agents have already published. 404 Media checked demo outputs against listings on three sites: Airbnb, Vrbo and Zillow [18]. It concluded that Raven is pulling data from sites like Zillow and Airbnb [1]. Every interior match the report describes landed on a property with a public listing.
The report does not say how Graylark obtains the listing data, how often the top result is right, or whether either agency uses the interior feature. The Charlotte, Montecito and San Francisco results are demonstrations, and 404 Media's checks confirm that the output text and photos match real listings. They do not measure error.
What to watch
- Whether Graylark, Airbnb, Vrbo or Zillow explains how Raven obtains listing photos.
- Whether LAPD or Miami-Dade Sheriff's Office confirms it uses the interior geolocation feature, and whether other agencies buy city-specific models.
- Demos covering houses and apartments, Heinen's stated 2026 target, would show whether matches extend to homes that are not listed anywhere.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence55
- Adoption20
- Hype gap+25
- Incentives65
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
404 Media says Raven is clearly pulling data from sites like Zillow and Airbnb.
- [2]
Graylark CEO Daniel Heinen wrote in a LinkedIn post earlier this month that the company was introducing "the ability to geolocate interior images across millions of short-term rental units worldwide."
- [3]
404 Media previously reported that both the Los Angeles Police Department and Miami-Dade Sheriff's Office have bought access to Raven, which was previously called GeoSpy.
- [4]
In a Graylark demo video, a washer/dryer photo taken from X was uploaded to Raven, and the map zoomed to Charlotte, North Carolina after a couple of seconds; the top result, "Renovated! Walk Uptown/located in 2 Hot Spots!", directly quotes a specific Airbnb listing whose photos show the same property.
- [5]
In the same demo, a bathroom photo from social media returned an area of Montecito, California, with the top result "Montecito Miramar Beach House on the Sand"; 404 Media found this directly quotes a listing on the short-term rental site Vrbo.
- [6]
From a single photo, Raven appears to find the specific home it was taken in without using location data or other metadata attached to the image.
- [7]
Raven is not a reverse image search: it identifies unique characteristics in a photo, then finds what are likely other photos from the same property on sites like Airbnb, even if those photos are not exactly the same or time has passed between them being taken.
- [8]
Heinen wrote in another post: "For investigators working trafficking and exploitation cases, that connection can turn an unidentified room into a location lead they can verify and pursue."
- [9]
In a demo posted last year, Raven traced an interior room photo to a specific address in San Francisco; the interior photos it returned were exactly the same as those posted on Zillow, according to 404 Media, indicating Raven is at least pulling data from Zillow in some form.
- [10]
Another video posted last week shows Raven identifying a particular hotel.
- [11]
A part of the Raven interface in one video says "City-specific property AI to determine an address," which implies the property search is not necessarily worldwide but depends on an AI model built for each city.
- [12]
Emails seen by 404 Media showed Miami-Dade Sheriff's Office had access to a custom Raven model specifically trained for Miami-Dade County.
- [13]
Heinen wrote in his post: "In 2026, we're investing heavily in interior geolocation across U.S. hotels, short-term rentals, houses, and apartments."
- [14]
Heinen told the publication Indicator last year: "I'll be honest, we didn't train it to do that."
- [15]
Heinen said GeoSpy 1.0 could not geolocate building interiors; the ability appeared as the company trained and tested GeoSpy 2.0.
- [16]
Graylark trained Raven on millions of images from around the world; at launch the tool focused on distinct geographical markers like architecture and soil.
- [17]
The photos in the demonstrations came from X and social media; 404 Media says that in practice they could come from anywhere a law enforcement officer has obtained them.
- [18]
404 Media checked Raven demo outputs against listings on three sites: Airbnb (Charlotte), Vrbo (Montecito) and Zillow (San Francisco).
Sources
1 independent publisher whose own reporting we read for this story.
- 404media.coAI Company Lets Cops Pinpoint Specific Houses Based on a Single Interior Photo
1 article · October 12, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.