How to Spot AI-Generated Images and Video

The short answer: your eyes are no longer enough. AI image and video tools have gotten good enough that the old giveaways (six fingers, melted text) show up far less often. The most reliable way to judge a viral image isn't staring at the pixels. It's checking where it came from.

Here's the checklist we use.

Step 1: Check the source before the pixels

Ask three questions first:

1. Who posted it first? An account created last week with no history is a red flag. A news organization's official account is not.
2. Is anyone credible reporting it? If a photo shows something huge (an explosion, an arrest, a celebrity doing something wild) and no established newsroom has it, be suspicious.
3. Does the caption match the image? Plenty of "fakes" are real photos with a false caption, an old photo passed off as new, or a photo from a different country.

Step 2: Run a reverse image search

Drop the image into Google Lens, Bing Visual Search or TinEye. You're looking for:

• Older copies. If the "breaking" photo appeared years ago, it's recycled.
• The original. Often you'll find the uncropped version, or the artist who posted it as AI art in the first place.
• Fact-checks. Many viral fakes have already been debunked by fact-checking outlets.

Step 3: Look for visual tells (they still happen)

AI generators have improved, but zoom in and check:

• Hands, teeth and ears: extra or merged fingers, too many teeth, earrings that don't match.
• Text and signs: garbled letters, nonsense words, logos that are almost right.
• Background people: faces that melt or repeat, limbs that don't connect.
• Physics: shadows going different directions, reflections that don't match, hair that blends into clothing.
• The "too perfect" look: glossy skin, dreamy lighting, flawless symmetry.

For video, also watch for: lips slightly out of sync with speech, unnatural blinking, flickering edges around the face or hair, jewelry or glasses that warp when the head turns, and voices that sound flat or oddly paced.

None of these prove anything alone. Real photos have weird shadows too. But a few tells together, plus no credible source, should stop you from sharing.

Step 4: Check for labels and credentials

• Platform labels. Major social platforms label some AI-generated content, either because the creator disclosed it or because the file carried metadata saying so. A missing label proves nothing.
• Content Credentials. Some cameras, editing apps and AI tools attach tamper-evident Content Credentials (the C2PA standard) that record how a file was made. Tools like the Content Credentials verify site can read them. Most images online don't have them yet, and screenshots strip them.
• Watermarks. Some AI tools add visible or invisible watermarks. Visible ones are easy to crop out; invisible ones need the maker's own detector.

Step 5: Treat AI detectors as a hint, not a verdict

Detection tools can flag likely AI images, but they get things wrong in both directions, and they get worse when an image has been compressed, cropped or screenshotted (which is how most viral images travel). Use one as a single clue alongside the steps above.

Before you share: the 10-second rule

If an image makes you feel a strong emotion (outrage, shock, "I have to send this to everyone"), pause for ten seconds. That feeling is exactly what fakes are built to trigger. Check the source, run a quick reverse search, and if you still can't confirm it, don't pass it on.

The bottom line

Spotting AI images is less about having a perfect eye and more about habits: source first, reverse search second, visual tells third. That's the same order DarkWire uses before a story runs. If it can't be confirmed, it doesn't get posted.