AI image generation has improved to the point where the old advice — count the fingers — no longer reliably works. Detection is getting harder, and it will keep getting harder.
That means the most valuable skill is not spotting artefacts in an image. It is verifying claims regardless of how convincing the image looks.
Visual clues that still sometimes work
These are becoming less reliable, but remain worth a glance:
- Text inside the image. Signboards, number plates, book covers and shop names often come out garbled or subtly nonsensical.
- Hands and teeth. Improved considerably, but still where errors cluster.
- Backgrounds. Look away from the subject. Background faces, patterns and repeated objects often break down.
- Physics of light. Shadows falling in inconsistent directions, or reflections that do not match the scene.
- Jewellery, glasses and straps. Thin objects that pass behind and in front of things often merge or vanish.
- An oddly perfect look. Skin without pores, lighting too even, everything slightly too polished.
Treat these as prompts to investigate further, never as proof either way.
The checks that actually work
1. Reverse image search
The single most useful tool. Upload the image to a reverse image search and see where else it appears. If a "breaking news" photo has been online for four years, you have your answer immediately.
This also catches the most common form of misinformation in India, which is not AI at all — it is real old photographs recycled with false captions.
2. Check whether real outlets are reporting it
If something dramatic genuinely happened, established news organisations will cover it. If a shocking claim appears only on social media and messaging forwards, that absence is itself the signal.
3. Look at the source account
Created recently? Posts constantly on one inflammatory theme? No identifiable person behind it? These patterns matter more than the image itself.
4. Use Indian fact-checking services
Several established fact-checkers cover Indian misinformation specifically, including claims circulating on WhatsApp. If something is going viral, it has often already been checked.
The WhatsApp forward problem
Most misinformation reaching Indian families does not arrive as a sophisticated AI image. It arrives as a forwarded message from a relative who trusts it because someone they trust sent it.
Two useful habits: check the "forwarded many times" label WhatsApp adds to widely circulated messages, and treat any message urging you to "share immediately" as suspect. Urgency to share is a manipulation technique, not a sign of importance.
Emotional manipulation is the real tell
The most reliable indicator has nothing to do with pixels. Misinformation is engineered to make you feel something strongly and immediately — outrage, fear, vindication.
If a piece of content makes you want to share it right now, that reaction is the designed effect. Pausing for two minutes to verify is the single most effective defence available, and it works even against content you cannot visually detect.
What to do about it
Verify before forwarding. Correct family members gently and privately rather than publicly. And accept that visual detection will keep getting harder — which is exactly why source verification, not image inspection, is the skill worth building.
