45
How to tell if an image is AI-generated | Malwarebytes
(www.malwarebytes.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 07 Jul 2026
45 points (100.0% liked)
Cybersecurity
10236 readers
146 users here now
c/cybersecurity is a community centered on the cybersecurity and information security profession. You can come here to discuss news, post something interesting, or just chat with others.
THE RULES
Instance Rules
- Be respectful. Everyone should feel welcome here.
- No bigotry - including racism, sexism, ableism, homophobia, transphobia, or xenophobia.
- No Ads / Spamming.
- No pornography.
Community Rules
- Idk, keep it semi-professional?
- Nothing illegal. We're all ethical here.
- Rules will be added/redefined as necessary.
If you ask someone to hack your "friends" socials you're just going to get banned so don't do that.
Learn about hacking
Other security-related communities !databreaches@lemmy.zip !netsec@lemmy.world !securitynews@infosec.pub !cybersecurity@infosec.pub !pulse_of_truth@infosec.pub
Notable mention to !cybersecuritymemes@lemmy.world
founded 3 years ago
MODERATORS
It's still weird people don't just use spider tingles...
We're pattern recognition machines and we don't need to conciously be aware of all the signs of AI.
It would be like trying to teach an adult how to ride a bike by telling them what muscles to constrict in what order.
With enough time and effort it might work, but it's the worst way to learn how to ride a bike and anyone that learns and only rides a bike that way is gonna fucking suck compared to someone who offloaded the specifics to their unconscious mind and just gives general commands like "turn left".
Tldr:
You can tell when it's AI, because of the way it is.
Trying to get nity gritty and pick out specifics isn't sustainable long term. The only downside is false negatives from everyone needlessly upscaling pictures, but those would happen no matter what.
Spider tingles are good, but you need some training.
People take the story from Malcolm Gladwell's Blink all wrong (including Gladwell). The story is that some art forgeries passed all the objective tests of forgery but that every expert thought they were fake but couldn't quite put their finger on what was fishy about it.
The lesson is NOT we are all amazing intuitive and we should always trust our instincts over our intellect, but that people who TRAIN their intuition with their intellect have insights they might not be able to express in words.
It would be good to keep a library of older, more obvious AI images to analyze and train people on.
Because people who pick out forgeries explain how they could tell...
And forgers paid attention to those things.
Same with AI, people going on about six fingers means we can't just count fingers anymore.
Anything quantifiable, will be what the slop experts focus on fixing.
We're just gonna be left with the ineffable, a collection of a whole bunch of small signs were not consciously aware of and even if we try to describe we're just coming up with post fact rationalizations.
Like, if you showed me an AI pic right now, I'm fully confident I could tell
But if you asked me to explain how, I'd just be making shit up like an AI would. I might say real signs that I consciously noticed, but anyone walking around with a checklist will at best gain a temporary ability that the slop will out pace.
It's gotta be "gut feeling". But because most people don't understand what those actually are, they refuse to trust it. But I bet they won't ignore a reaction to duck if someone throws a wrench at their face.
It's the same thing, your mind/body deciding something without going thru the hassle of your conscious brain. It's a bandwidth thing, you don't need to "see the math" because that would take too long. You just gotta trust the flashing red light that says "this isn't real".
If that were true, there wouldn't be so many comments saying they are 100% sure it is AI on 10 year old images and videos.
You can't tell, because AI is designed to be indistinguishable from regular images, until you look for clues up close. And even then those clues are rapidly disappearing the more advanced the AI becomes.
No, I'm saying I can tell...
And I'm saying that not everyone can, because I understand human variation...
You are saying that you can't tell, and you are saying that no one else can tell because people can only do things you can do...
The fakes are also rapidly getting better. A reliable detector and careful wallet-based-voting is needed to fight this bullshait.
I've tried to teach my father about all this, he got fleeced for $500. I've spent HOURS with him to no avail. He still believes every fucking thing. It's infuriating. The situation, his capacity to learn, state of humanity just fucking the shit out of one another mentally...
Humans ruin everything.
So how do I "just use spider tingles"?
To me this sounds like you are saying "just ride the bike, even five year olds can do it".
Are you asking me how to trust your gut?
Or are you saying your subconscious is bad at recognizing patterns?
Or that you're not in touch with yourself in general?
I'm willing to help, I just can't tell what you need help with bro
Yes to all of your questions. How can I trust my gut? Years ago, when AI generated images were far worse than today, I did a test on a web page that showed some AI-generated and some real images. And I was really bad. Nowadays I don't have a chance.
Even here in the Fediverse I often don't recognise AI-generated images until someone points out mismatched shadows or some other "obvious" errors.
I mean, you might just not be good at it, and this is like asking how to be 7 feet tall so you can play in the NBA.
That's how variation works, there's likely other aspects you're good at and others aren't.
Like I said in another comment tho, it's not like you can learn what to look for, because the things you're conciously looking for are the same things the people making it are aware of.
The only way to come close to "learning" is to just look at a shit ton of AI pictures while reminding yourself over and over "this is AI". Like, for a while, maybe 30-60 minutes at a time. Do the same thing for real pictures but repeating "this is real" as an internal monologue.
You're just hammering a label unto your various subconscious processes until they start shoving everything into that binary label system. Either "ai" or "real".
But again, human variation. There's other reasons to jog than beating Usain Bolts 100m dash time, you might not ever be great at it, you might not even see much improvement.
But doing that enough would eventually make you better at picking up all the little signs that others are noticing. It's exploiting how our minds work and want to see patterns.
Dinguses jump at shadows. You get people going 'I can tell it's photoshopped because of the pixels!,' or sneering 'I bet you're an LLM' over polite efforts to deal with their bullshit.
https://sightengine.com/which-image-is-ai
Here ya go. I'm sure you'll lie if you ever bother replying, but there's zero fucking chance you're a magic AI detector.
That gives different pictures everytime it's loaded...
You want me to say "the left one is ai"? It won't be the same pictures you looked at buddy...
And that's not even getting into how you linked to a "service" that uses AI to detect AI...
https://sightengine.com/pricing
Like, obviously they're gonna make it sound impossible, they're trying to sell AI to idiots
Yes. Because a test that give you the same fucking pictures every time you load them would mean you click through once quickly to get all the right answers, then restart and show us you really are 100% perfect.
Going to go ahead and block you now. Better luck in your future endeavors, Captain Psychic Wizard AI/Lie detecting witch hunter.
I'm sorry most people don't need to pay an AI to tell them when a picture is AI, I'm sure that would be very frustrating