That photo of the Pope in a white puffer jacket fooled millions — including celebrities who shared it without a second thought. It was never real. It was made in minutes by an AI image generator. Could you have spotted it? By the end of this article, yes — and one small detail in that very photo gives it away.
Short answer: check the hands, the text, the skin, the light, and the background — then confirm with a free tool. No single sign proves an image is AI-made, but seven checks together catch most fakes in under a minute.

The AI-generated "Pope in a puffer jacket" image from March 2023 fooled millions. Image: Midjourney / Wikimedia Commons, public domain.
Here is the whole checklist at a glance:
| Sign | What to check |
|---|---|
| 1. Hands | Extra or fused digits, twisted joints, rings melting into skin |
| 2. Text | Garbled letters, misspellings, fonts that change mid-word |
| 3. Skin and faces | Too-perfect, waxy skin; mismatched eyes; odd teeth |
| 4. Light and shadows | Shadows in different directions, reflections that don't match |
| 5. Backgrounds | Warped lines, cloned faces, objects melting together |
| 6. Small details | Hair fading into skin, jewelry merging with the body |
| 7. Provenance | Watermarks, Content Credentials, reverse image search |
How Can You Tell If an Image Is AI-Generated by Looking at the Hands?
Start with the hands. Early generators were terrible at them: six fingers, fused knuckles, joints bending the wrong way. Newer models are much better, so clean hands no longer prove a photo is real. But hands are still the hardest thing for AI to draw, especially holding something. A hand around a coffee cup has to get the grip, the fingers, and the physics right all at once — and AI often gets close without getting it correct.
Zoom in and count. Trace each finger from base to tip. Look for extra digits, twisted joints, and nails that blur or vanish. Check whether both hands match in size. Then look at what the hands touch: rings blending into skin, a cup handle passing through a finger, a pen merging with a palm. Real hands are messy in consistent ways. AI hands are clean in impossible ones.
Now the payoff I promised. In the Pope photo above, the hand holding the cup is smeared — the fingers do not really grip anything. The glasses are wrong too: the frames and the shadow they cast do not match. Millions missed all three because the image felt real. Feeling real is exactly what AI is good at. Zooming in is what catches it.
Why Does Text in AI Images Look Wrong?
Text is still the weakest link — and the fastest check you can do. Image generators treat letters as shapes, not language. They can draw something that looks like a word from across the room, but up close the letters fall apart. Zoom in on every piece of writing: street signs, shop fronts, T-shirts, book spines, labels, license plates. Real text is consistent. AI text drifts: misspellings, letters changing shape mid-word, fonts shifting for no reason. A sign that reads almost right, one letter off, is one of the strongest tells there is.
Logos tell the same story. A brand logo in an AI image often looks familiar at first glance and wrong on the second: a letter missing, a color slightly off, a shape no designer would approve. If the text in an image cannot be read, treat the image as unverified until something else confirms it.
What Does AI-Generated Skin Look Like?
AI people often look too good to be real — and that is the tell. Real skin has pores, blemishes, uneven tone, and fine lines, even in a good photo. AI skin tends to be smooth, waxy, and oddly even, like a plastic doll. Cheeks with no texture, foreheads with no lines, skin glowing a little too evenly.
Then look at the eyes and teeth. AI eyes sometimes have pupils that are not quite round, or catchlights — those little reflections — that do not match between the two eyes. Teeth are another weak spot: too many teeth, teeth merging into one another, a smile slightly too wide. None of this is proof alone — a beauty filter can smooth real skin. But plastic-perfect skin and mismatched eyes and painted-on teeth add up to a pattern. Researchers at Northwestern University sort the visual signs into five groups, and flawless skin sits in the stylistic group: the image looks like an idea of a person, not a photograph of one.
How Do You Spot Fake Lighting and Shadows in AI Images?
Light follows rules. AI sometimes forgets them. In a real photo, every shadow points away from the same light source. In an AI image, shadows often disagree: one person's shadow falls left while another's falls right, or a shadow appears where no object could have cast it. Reflections are just as revealing — a mirror, window, or puddle showing something different from the scene, or a reflection missing when it should be there.
Also watch for lighting that is too perfect. Real light is uneven: one side of a face brighter than the other, a harsh highlight, a corner falling into shade. AI images often light everything evenly, like a studio shoot with no studio. That alone is not proof, but with a shadow pointing the wrong way, it is a strong signal.
Why Do Backgrounds in AI Images Look Strange?
Most people stare at the subject. The background is where AI gets sloppy, because that is where nobody is looking. Zoom into the edges: door frames, window edges, railings, and table legs that warp or vanish halfway. Objects melting into each other — a chair leg becoming a table leg, a plant stem with a gap in it, a building dissolving into the sky.
Crowds are the biggest giveaway of all. AI struggles to invent many distinct people at once, so background faces often repeat: the same nose, the same jawline, the same expression on three different people. A real crowd is full of distinct, imperfect people. An AI crowd is full of copies.
What Should You Check in Hair, Jewelry, and Small Details?
The last visual check is the smallest one — and often the most decisive. AI is good at the big picture and bad at the fiddly bits. Real hair has individual strands that connect to a hairline. AI hair often fades into the forehead, blurs into double lines, or merges with a hat or the background. Look at the edges where hair meets skin: if the boundary is smeared or glowing, be suspicious.
Jewelry does the same thing: earrings that are not a matching pair, a necklace disappearing into the neck, a watch face with no hands or the wrong numbers. The same goes for glasses, buttons, and zippers — small, structured objects that must follow exact rules. AI approximates them. Reality does not. The newest models are fixing these flaws fast, so no single visual tell is dependable anymore — which is why the seventh sign matters most.
What Is the Fastest Way to Check If an Image Is AI-Generated?
If your eyes cannot decide, let a tool decide. Three free checks take less than a minute, and together they beat any visual tell.
1. Check for a watermark or provenance record. Google's SynthID is an invisible watermark in the pixels of images made or edited with Google's AI tools, including Gemini's image generator. Check for it in the Gemini app — Google said in May 2026 that the check is expanding to Search and Chrome. Separately, the C2PA standard, Content Credentials, stores a signed record inside a file showing which tool created the image; inspect it at the free Content Credentials Verify site. The catch: these systems only flag content from tools that adopted them. Older or open-source generators leave no mark, so finding nothing proves nothing.
2. Reverse image search. Upload the image to Google Lens or TinEye and see where else it appears. A real photo of a real event usually has an original — an older copy, a news article, a photographer's page. An AI image often appears only where you found it, or in fact-check articles explaining that it is fake. If the image supposedly shows breaking news but no news outlet has it, that is your answer.
3. Try an AI detector — and distrust the result. Detectors return a probability score, but they are the least reliable tool here. A NewsGuard audit in May 2026 found that popular detectors flagged about 13 percent of real news photos as AI-generated on average — and one tool flagged 40 percent. A "likely AI" verdict is a reason to investigate, not a verdict. Use detectors as a second opinion, never as the judge.
Two more checks cost nothing: look for a platform label — Instagram, Facebook, and TikTok now label some AI-generated content — and check the source. A named photographer or a reputable outlet beats a brand-new account with no history. The EU's AI Act now requires AI-generated content to be labeled, in effect since August 2026.
What Should You Do When You Still Can't Tell?
Sometimes you will run every check and still not know. That is normal now. In one survey, 72 percent of Americans could not tell AI-generated images apart from real photos — and the models have only improved since. When you cannot tell, change the question: stop asking "is this image fake?" and ask "where did this come from, and what does it want me to do?"
Check the source before the pixels. A surprising image from an anonymous account, posted minutes ago, with an emotional caption and no other source, deserves skepticism whatever it shows. Ask what the image wants from you: urgency, money, outrage, or a share are warning signs in themselves. And if you are about to share it or act on it, verify it with a second source first. A thirty-second reverse image search beats a hundred confident opinions in the comments. The checks that matter most do not depend on your eyes: provenance, watermarks, and the source. Learn the seven signs, use them together, and never trust a single one alone.
The Bottom Line
Here is the one-minute version. Zoom in on the hands and count the fingers. Read every word of text. Look for skin that is too smooth and eyes that do not match. Check that the shadows agree. Scan the background for warped lines and cloned faces. Then confirm with a tool: a SynthID check, a Content Credentials lookup, or a reverse image search. One sign is a hint. Three signs are a pattern. And when nothing settles it, check the source — where an image comes from usually tells you more than how it looks.
AI image generators are only getting better, and the next viral fake will be harder to spot than the Pope in the puffer jacket. The defense is not a sharper eye. It is a slower thumb: pause, zoom, check, and then decide. The same technology is changing how we learn and work — OpenAI's GPT-6.2 Astra and Sol are built for exactly this kind of everyday AI work — from choosing the right AI chatbot for school to understanding agentic AI, the trend that lets AI tools act on their own.
Your 1-Minute Checklist Before You Share Any Image
- Zoom in on hands — count fingers, check joints, look at rings and grips
- Read all text in the image — signs, labels, logos, license plates
- Check skin, eyes, and teeth for the too-perfect look
- Make sure shadows and reflections agree on one light source
- Scan the background for warped lines and repeated faces
- Check SynthID or Content Credentials, then reverse image search
- Verify the source — and if it still feels off, don't share it
Comments
Post a Comment