Reality is a legacy feature. We’re living in the era of the visual deep-fry, where every scroll through a feed involves a subconscious Turing test you didn’t ask to take. If you feel like you’re losing your mind, don’t worry. You’re just witnessing the final collapse of "seeing is believing."

For years, we poked fun at the six-fingered horrors and the spaghetti-eating nightmares. But the models got better. They got faster. Now, for the low price of a $20 monthly subscription, any bored teenager or political operative can generate high-fidelity friction. We’re being flooded with high-gloss sludge, and the platforms are doing nothing to stop it because engagement is engagement, even if it’s fake.

If you want to keep your grip on the truth, you have to look for the glitches in the matrix. Here are seven ways to spot the rot before you accidentally share a photo of a politician doing something that never actually happened.

First, look at the hands. It’s the classic tell for a reason. AI doesn’t know what a hand is; it just knows what a hand looks like from ten thousand different angles. The result is often a fleshy pile of knuckles or a thumb that emerges from the wrist. If the subject has their hands tucked into their pockets or hidden behind a back, be suspicious. The prompt-engineer was probably trying to hide the model’s inability to count to five.

Second, read the signs. Not the metaphorical ones—the literal ones. Look at the background text, the labels on bottles, or the street signs. AI handles the English alphabet like a drunk toddler. It catches the vibe of the letters but misses the actual semantics. If a "Stop" sign looks like it was written in a lost Mesopotamian dialect, you’re looking at a hallucination.

Third, check the "plastic surgery" sheen. Midjourney and its cousins love a smooth finish. Real human skin has pores, scars, uneven peach fuzz, and oily patches. AI skin looks like it was buffed with a bowling ball polisher. It’s too perfect, too luminous, and utterly devoid of the messy reality of being a biological organism. If everyone in the photo looks like they’ve had eighteen rounds of Botox and a high-end chemical peel, it’s probably a bot.

Fourth, follow the light. Shadows are hard. Physics is harder. AI doesn't understand that a light source should cast a shadow in a consistent direction across every object in a frame. You’ll often see a shadow falling to the left of a chair while the person sitting in it casts a shadow to the right. It’s a subtle discordance that your brain registers as "wrong" even if you can't immediately say why.

Fifth, watch the accessories. Earrings are a death trap for generative models. One might be a classic gold hoop while the other is a dangling pearl that seems to be fused directly into the wearer’s jawbone. Glasses are another disaster zone. The frames will often melt into the temples or have asymmetrical lenses. AI struggles with the concept of pairs.

Sixth, look at the background extras. The main subject might look great, but the people in the "nosebleed seats" of the image are usually nightmare fuel. Look for faces that lack features, limbs that merge into the scenery, or people who seem to be three feet tall compared to the person next to them. The model spends all its "compute" on the center of the frame and leaves the edges to rot.

Seventh, look for the "dream logic." Does the scene actually make sense? You’ll see a man wearing a winter coat while standing in a tropical surf, or a reflection in a window that shows a completely different street than the one in the foreground. It’s a digital collage built on probability, not logic.

The friction here isn't just the difficulty of spotting these errors; it’s the cost of the search. We’re being forced to do the unpaid labor of fact-checking our own eyeballs. Tech giants like Meta and Google talk a big game about invisible watermarking and "Content Credentials," but that’s mostly PR theater. Metadata can be stripped in seconds with a simple screenshot. The tech to create the fake is always six months ahead of the tech to catch it.

So, we spend our days squinting at pixels, trying to find the sixth finger or the melted earring. We’re training ourselves to be forensic analysts just to get through a Tuesday afternoon without being duped by a bot. It’s an exhausting way to live, but that’s the trade-off for "democratizing" creativity.

We’ve finally reached the point where the image is no longer a record of a moment. It’s just a suggestion. If everything can be fake, does it even matter when something is real?