Spotting AI images
AI image generation has come a long way, but it still leaves fingerprints. The trick is knowing where to look.
Start with your gut. If something feels "off" but you can't immediately say why, that instinct is usually right. AI images trigger this uncanny valley response where everything looks almost correct, but not quite. Trust that feeling.
Eyes are often the first giveaway. They might look lifeless, lacking that natural depth and wetness you see in real photos. The pupils could be asymmetric, or the reflections inside them might show something that makes no physical sense. Smiles can feel pasted on, not reaching the eyes the way genuine expressions do. And sometimes there's just... nothing there. An empty stare where emotion should be.
Then there's the "too perfect" problem. Skin that's unnaturally smooth, with no pores or texture. Lighting that has no source, or seems to come from everywhere at once. Colors that feel dreamlike and oversaturated. No grain or noise even in what should be a low-light shot. Backgrounds that are suspiciously uniform. Real photos have imperfections. AI often forgets to add them.
You might also notice repetitive patterns if you look at enough AI content. The same poses keep showing up. Body proportions feel weirdly standardized. There's this recurring "AI face" that you start recognizing. If you get a "I've seen this exact thing before" feeling, you probably have, just with different hair or clothes.
Fine details are where AI really struggles. Hands are the classic example: extra fingers, missing fingers, fingers that bend the wrong way or grip things impossibly. But it goes beyond that. Jewelry can fuse into skin or hang in ways that defy physics. Clothing has seams that lead nowhere, buttons in random places, fabric that doesn't fold right. Hair tends to melt into backgrounds or turn into texture soup at the edges.
And then there's text. Any text in an AI image is almost always a dead giveaway. Signs with unreadable gibberish. Shirts with letters that almost form words but don't. Fake watermarks and signatures that look like someone had a stroke while writing. Background objects that you can't identify because they don't actually exist. Architecture that makes no structural sense: doors to nowhere, windows at wrong angles, stairs that would be impossible to walk on.
Real Examples
The skin texture here is unnaturally smooth, almost plastic-like. While some pores are visible, they can look stamped on rather than naturally distributed, and the specular highlights feel too even across the face. The lighting and shadow gradients also read slightly "airbrushed", with fewer tiny imperfections than you'd expect in a real close-up. Pay close attention to where the eye meets the eyelid: AI often blurs the boundary at the base of the eyelashes, and the pupil isn't perfectly round, or the reflection inside it makes no sense. Also check small transitions like the nostril edges, lips, and hairline: AI can make these boundaries either too crisp or strangely soft, which adds to the waxy impression.
This one shows similar issues: the face has that telltale "too clean" look. Notice how the lighting creates shadows that feel slightly off, as if the light source exists in multiple places at once. Skin tone and texture can look uniformly smoothed, with fewer pores and tiny irregularities than a real portrait at this distance. Another common tell is the background bokeh: the circles can feel a bit too perfect and evenly shaped, and edge transitions around hair or the jaw can show slight halos or unnatural blending. The eyes, while detailed, can still lack the subtle wetness and depth you usually see in real photos.
Here's where things get interesting. Look at where the hair strands meet the background: they tend to "melt" into it, or stray hairs appear from anatomically impossible locations. Jewelry like earrings and necklaces are often asymmetric or simply fuse into the skin or clothing. In this example, check the wispy hairs around the jaw and ear: some strands fade abruptly into the background, and the overlap between hair and the hoop earring can look oddly blended, as if the edges can't decide which layer is in front. Also watch for "too clean" boundaries (like straps or clothing edges) and a background that feels like a perfectly smooth studio gradient, both of which can be hints of synthetic generation.
Continuing with fine details: watch how accessories interact with hair and skin. The AI struggles with these boundaries, creating impossible overlaps or objects that seem to phase through each other. These small inconsistencies are easy to miss at first glance but become obvious once you know what to look for.
Fingers and hands are the classic giveaway. AI often struggles with bone structure and occlusion, resulting in fingers that are too long, bend unnaturally, or come in the wrong quantity. But anatomy issues go beyond hands. Look at the musculature: AI tends to either over-smooth muscles into a plastic-looking surface, or create bizarrely defined abs and arms that don't match the person's overall build. Knees are another weak spot—they often look like simple hinges without the complex structure of real kneecaps, tendons, and the subtle bulges of a real joint. In this example the hands are tucked into pockets, so the tell is indirect: look at the pocket openings and the fabric tension. The hoodie and shorts should show clear, believable folds where knuckles and thumbs push against the fabric, but instead the pocket area can look oddly smooth. Also check the legs: the knee area might lack the natural definition you'd expect, or the calf muscles don't flow naturally into the ankle.
This example is a close-up portrait, so the "anatomy" tells show up in small facial details. Look closely at the hairline and loose strands: some edges fade into the background in a way that feels painted, and a few strands have inconsistent thickness. Check the skin microtexture too: pores and fine lines can look selectively sharp in some areas but strangely smoothed in others, like a beauty filter with uneven masking. Also inspect the eyelids, lashes, and eyebrows: AI often makes them a bit too perfect, with repeating clumps or clean, uniform strokes that do not match natural variation. Finally, watch the mouth and nostril edges for tiny warping or overly crisp boundaries that add to the synthetic feel. Zooming out, the overall composition can feel staged, and the eyes may read as glassy or lifeless, which is a common trait across many generated images.
Now let's examine what's happening behind the subject. Background objects like buildings and plants often dissolve into an incomprehensible mess. The AI generates something "background-like," but if you try to identify any specific element, you can't tell what it actually is. It's visual noise that passes for detail only at a glance. Also pay attention to the gaze: even in a tight close-up, the eyes can feel oddly distant or slightly misaligned, as if the subject is staring past the camera. That "not quite looking at you" feeling is common in generated portraits, and it gets more noticeable when the background is already a soup of vague shapes and inconsistent depth cues.
The bokeh effect here doesn't follow optical laws. Real cameras create gradual blur based on distance from the focal plane. AI often blurs objects inconsistently: something close might be sharp while an object at the same distance is completely out of focus. The blur feels random rather than physically motivated. Look at the two black cats in this image: they appear to have melted into each other, their bodies merging where they shouldn't. And where are their hind legs? They simply vanish into the blurred mess. When AI doesn't understand the physical separation between subjects, it just... blends them together.
Clothing patterns are another weak point. Here you can see how the pattern illogically breaks or morphs into something else entirely. Buttons and seams lead nowhere or disappear into the fabric. These details would be consistent in a real photograph because they follow physical construction.
Finally, examine where different materials meet: skin touching fabric, fabric meeting jewelry, or any boundary between textures. The AI frequently fails here, with the fabric texture "floating" onto skin, or materials blending into each other in ways that would be impossible in reality. Once you start noticing these seams, they're everywhere.
At first glance this looks like a fun, cute scene, but let your eyes wander to the audience seated behind the main subject. The faces in those rows are the biggest tell: they're distorted, blurry, and almost featureless, as if the AI just "smudged" placeholder humans into the seats. Moving on to the dog itself, check out the gloved hands holding the popcorn. The fingers don't follow either a realistic paw structure or proper human anatomy; they look rubbery and strangely posed. And speaking of popcorn, zoom in on the bucket: the kernels have an oddly uniform texture, and in some spots they melt together into a single amorphous blob rather than individual pieces. Once you notice these details, the illusion falls apart.
This adorable bunny portrait hides several AI fingerprints if you look closely. Start with the glasses: where the black frames meet the fur, there's no real depth or shadow, no impression of the frame pressing down on the fur. It looks more like a 2D sticker pasted on top of the image. Then check the whiskers: AI often struggles with thin, delicate lines, and here you can see some whiskers passing straight through the frame without bending, or fading into the background unnaturally. Finally, look at the reflections in the eyes: instead of showing a coherent scene like a window or a light source, they're just abstract white blobs. Real eye reflections tell a story; these tell you it's generated.
This one is a classic case of AI struggling with fine metallic details. Take a close look at the necklaces: the chain links merge together, break apart randomly, or connect in ways that defy physics, almost like tangled metal spaghetti. Moving to the headphones, you'll notice the brand text is completely unreadable, just gibberish characters, and the cable attachment and curve feel unnatural, as if the cord doesn't know where it's supposed to go. And then there's the skin: it's almost supernaturally smooth and pore-free, with a slight plastic sheen that screams early generative model. Real skin has texture, tiny imperfections, and natural variation. Here, even makeup couldn't explain this level of airbrushed perfection.
Spotting AI videos
Video generation is newer tech, and honestly, it's often easier to spot than images. The flaws really show themselves once things start moving.
The first thing you'll notice is that dreamlike quality. Everything feels slightly surreal, like you're watching through a filter that doesn't exist. The exposure stays weirdly consistent when it shouldn't. There's this soft, floaty atmosphere even in scenes that should feel grounded. Colors might shift unnaturally between frames.
Motion is where it really falls apart. Human movement has these tiny micro-jitters and imperfections, but AI motion is often too smooth, too clean. Gestures feel robotic, lacking that natural acceleration and deceleration we do without thinking. Faces can subtly morph between frames if you pay attention. Hair and clothing move wrong, sometimes clipping straight through the body. Limbs might stretch or bend in ways that would send you to the hospital in real life.
Watch for scene inconsistencies too. Objects appearing or disappearing between cuts. Backgrounds that warp when the camera moves. Shadows that don't match where the light should be coming from. Reflections showing something different than what's in the scene. Physics violations with water, fire, or smoke behaving in ways they never would in reality.
Audio tells are useful if the video has sound. AI voices have this robotic cadence with unnatural pauses. Lip sync is often slightly off, with mouth movements not quite matching what you hear. Ambient sounds that should exist in the scene are missing. Background noise stays perfectly uniform instead of shifting naturally.
Spotting AI writing
Text has its own fingerprints. Once you learn them, you'll start noticing AI writing everywhere: articles, comments, product descriptions, social media posts.
Large language models have favorite words they massively overuse. "Delve" is the famous one, but there's a whole family: crucial, pivotal, vital, tapestry, landscape, interplay, foster, enhance, showcase. Transitions like "Additionally," "Furthermore," and "Moreover" showing up at the start of sentences. Adjectives like "intricate," "vibrant," and "enduring" that sound sophisticated but say almost nothing. When you see several of these in one piece, that's a red flag.
AI text loves telling you how important everything is. It can't help itself. "Stands as a testament to..." is a classic phrase. Everything "plays a vital role" or "reflects broader trends." Even mundane topics get described as having an "enduring impact" or leaving an "indelible mark on history." This puffery is constant.
There's also this pattern of tacking on analysis-sounding phrases at the end of sentences, usually with -ing words. "...ensuring continued growth." "...highlighting its significance." "...contributing to the broader ecosystem." These sound thoughtful but add nothing meaningful.
The structure of AI writing is predictable. There's this "rule of three" overuse: "adjective, adjective, and adjective" or "phrase, phrase, and phrase." Negative parallelisms like "It's not just about X, it's about Y" show up constantly. Conclusions follow the "Despite challenges... continues to thrive" formula. Em dashes get scattered everywhere. And there's this thing called "elegant variation" where the AI refuses to repeat words, so "the city" becomes "the urban center," then "the metropolitan area," then "the municipality" all in one paragraph.
Watch out for vague attribution. "Experts argue..." but which experts? "Industry reports indicate..." but what reports? "Observers have noted..." but who, when, where? "Has been described as..." by whom? Often by the AI itself. This weasel wording makes claims sound authoritative without actually saying anything verifiable.
The promotional tone is another giveaway. AI writing often reads like marketing copy even when it shouldn't. "Nestled in the heart of..." is travel brochure language. Places apparently "boast" things constantly. Everything "continues to captivate audiences." Words like "groundbreaking" and "revolutionary" get thrown at completely ordinary subjects.
Structural patterns are telling too. AI loves Title Case For Every Heading. It uses excessive boldface like a textbook. Lists follow the "**Point:** Description" format everywhere. Articles end with "Challenges and Future Outlook" sections. Conclusions start with "In summary" or "In conclusion."
Sometimes AI leaves actual glitches. Markdown syntax bleeding through: asterisks for bold, hash symbols for headings. Curly quotes ("these") instead of straight ones ("these"), which is ChatGPT's signature. Placeholder text that was never filled in. Strange reference bugs like "citeturn0search0" or "[oai_citation:0]". Collaborative phrases like "I hope this helps!" or "Let me know if you'd like..." that were clearly meant for a chat, not published content. Knowledge cutoff disclaimers: "As of my last update..." or "While specific details are limited..."
And just like with images, there's an uncanny valley for text. It feels "off" before you can articulate why. Too smooth, no personality, no voice. Verbose but empty, using many words to say little. Universally positive, rarely taking a real stance or expressing genuine criticism. Grammatically perfect but somehow hollow. AI regresses to the mean, saying what's statistically most probable rather than what's most true or interesting.
The bottom line
If you have questions, want clarification, or want to dig deeper, let us know. We can expand on specific points, share examples, or help with related topics.