Have you seen viral videos showing dogs being rescued from train or subway tracks? Lead Stories has spotted a number of such videos recently and found that many of them were created using artificial intelligence. They do not show real people risking their lives to save their real pets.
This article is part of a series of stories examining various types of what is commonly called "AI-slop": short, vertical videos generated with artificial intelligence tools with the aim of going viral, often by exploiting the emotions or curiosity of the viewer with made-up content.
What these 'dog rescue' videos look like
One of the most common repeated scripts in such videos involved a man with a dog waiting for a train at what seems to be a New York City subway station. For no obvious reason, the pet suddenly jumps on the tracks, and its apparent owner jumps down too to return the dog to the platform. As the rescue attempt progresses, the fast approaching train appears in the background. However, both the pet and the owner manage to get back to safety at the last moment and get out of the ordeal competely intact.
In some cases, the people saving the dogs are train station workers. Women are also sometimes cast in the role of savior, though less frequently.
Occasionally, people are portrayed saving the dogs they accidentaly discovered chained to the tracks.
Some examples
Here is one example uploaded (archived here) on X on Feb. 21, 2026 by the @ShemouelNtabala account purported to be CCTV footage:
Oooh this man is a hero!!! He risked his life to save dog😳😨🙆
-- Samuela (@ShemouelNtabala) February 21, 2026
🏃Nkigendera pic.twitter.com/zN8QJKnX1D
Here is a variation (archived here) from TikTok where it was published on June 30, 2026 by @user7030585695653, and this video purported to show a real station worker:
@user7030585695653 MashAllah ☝🏻😇 Great Save Caught On Camera 😍 #allah#shorts #status #respect_SQ2HOMGuRiE♬ original sound - user7030585695653
And here is a version (archived here) from the @SecretDiariesTopVideos account on YouTube. In it, it is a woman who pulls the puppy back to the platform:
Keywords
These are some of the keywords and hashtags associated with the videos reviewed in this prebunk: man saves dog from train, heroic dog rescue, dog rescue from train/metro/subway tracks, train tracks rescue, subway rescue, metro rescue, hero saves pet, man saves dog, ,animal rescue, courage and bravery, heart-stopping rescue, dramatic rescue video, viral rescue video, heartwarming animal story, human-animal bond, interspecies compassion, dog trapped on tracks, pet safety awareness, dog hero, #Tiktok, #commentary, #Dog, #savinglives, #Hero, #DogRescue, #MetroRescue, #Bravery, #Humanity, #animalrescue, #shorts, #secondsfromdisaster, #TrainTracks, #CloseCall, #HeartAttack.
How to tell the videos are false
One of the most telling features were striking similarities between the storylines, angles and character's behavior in these videos, which suggested the use of the same prompt or a similar one.
Indeed, the dog rescue videos frequently contained the hallmarks of generative AI. One example is the distorted portrayal of space, as in the video below that unrealistically depicts the man and the dog being able to squeeze into the suddenly widened gap between the platform and moving train and escape unharmed, all while the ordeal completely erases the man's facial features:
(Image source: post by @newshub24live on TikTok.)
Another clip (archived here) features a fox that disappears and reappears in the frame as the scene unfolds.
Some of these videos originated from accounts that openly displayed labels disclosing that their posts was AI production:
(Image source: post by @paws.and.people on TikTok.)
In some cases, even accounts' names gave away that their content did not document real life:
(Image source: post by @aitrend.viral on TikTok.)
When Lead Stories ran these videos through online AI detectors, they confirmed a high likelihood of AI-generated content:
(Image source: Hive Moderation.)
(Image source: Sightengine.)
If you see videos like these on social media, here are some things you can try.
First, look for AI disclaimers added by the platform or the poster. On TikTok they might say "Contains AI-generated media" or "AI-generated" (archived here), on X they read "Made with AI" (archived here) and on Facebook/Instagram they often say "AI info" (archived here). Check the description of the video too: In some cases the creator might have added a note or a hashtag like #AI, #madewithAI or #fiction. Don't forget to check the main page of the account that posted the video, either: Maybe there is a disclaimer in the bio and in some cases AI use is really obvious when an account is posting dozens and dozens of variations of the same type of video.
Don't stop at the account that posted the video: Maybe they copied it from somewhere else. Use a tool to take a screenshot of the first frame of the video and run it through a reverse image searching tool to see where else on the internet it appears. It may have originated on an account that posts satire, AI creations or actual art.
If you are still not sure, try downloading the video itself. There are several AI detection tools that can tell you if there are watermarks or other technical characteristics present in the video that would indicate it was likely made with AI.
Finally, use common sense: If the video shows an event that would otherwise be newsworthy, use a news search engine to check if it has been reported on by a news service you trust. Also pay close attention to the video itself: look for physical impossibilities or glitches typical of AI-generated footage like:
- People or things appearing (or disappearing) out of thin air
- Objects behaving in physically impossible ways (heavy objects falling slowly, rigid objects bending ...)
- Garbled writing, oddly shaped letters or signs
- People or objects blending into or moving through each other
- Inconsistencies between different shots of the video (extra architectural elements in buildings, changing backgrounds, differences in clothing or hairstyle)
- An audio track that sounds strange: flat, unnatural speech, scripted-sounding yelling from bystanders ("Did you see that? OMG!"), sound effects being out of sync with events.
Unsure about a video? Email [email protected] and we will take a look!
These materials were developed in 2026 for the Prebunking at Scale project, with support from the European Fact-Checking Standards Network. If you share this on social media, use #prebunkingatscale.
