Stock photo sites are getting flooded with ai images and its getting hard to tell

Image Authenticity
1

i buy a lot of stock photos for marketing campaigns and landing pages. probably 50-100 images a month across various projects

over the last few months ive noticed a massive increase in images that look… off. not obviously ai generated but something about them feels weird. slightly too perfect lighting, skin that looks smoothed in a way that isnt just retouching, backgrounds that are almost right but not quite

the thing is some of these are on legitimate stock platforms mixed in with real photography. and theres no label or tag saying its ai generated

tried reporting a few that i was pretty sure about and the platform basically said they cant verify either way. which is wild because theyre selling these as authentic stock photography

anyone else noticing this? how are you handling image sourcing now

2

Absolutely noticing this. i run a few niche content sites and weve had to completely change our image sourcing workflow. we now reverse search every stock image we buy and check for common ai artifacts before using it

the hands are still a giveaway sometimes but flux and the latest midjourney versions have mostly fixed that. what i look for now: text in the background thats almost readable but not quite, jewelry that doesnt make physical sense, and reflections that dont match the light source

3

this is a real problem for anyone who cares about content authenticity. i write for a magazine and our photo editor told me they rejected about 40% of contributor submissions last quarter because they suspected ai generation. but they dont have a reliable way to verify so its mostly gut feeling

the platforms need to require provenance metadata. camera exif data, content credentials, something. right now its the wild west

4

there’s an interesting study from the Reuters Institute that found consumers generally cannot distinguish AI-generated stock imagery from authentic photography when viewing at typical web resolution. The problem is compounded by the fact that detection tools for images lag significantly behind text detection tools in accuracy.

What concerns me from a research perspective is the erosion of visual evidence. If we can’t trust that an image is real, photojournalism and documentary photography lose their evidentiary value.

5

@NovaJunkie88 good tips. The text in backgrounds thing is one ive noticed too. @EchoDust_7 40% rejection rate is insane