Home Tech Can We Still Trust What We See Online? The Challenge of AI-Generated...

Can We Still Trust What We See Online? The Challenge of AI-Generated Reality

2
0

As AI makes fake photos, videos and voices harder to spot, this article explains why checking where content came from now matters more than judging how it looks, and gives everyday habits for avoiding scams and misinformation.

Not long ago, a fake photo usually gave itself away. Faces looked a bit off, a hand might have six fingers and computer voices had that flat, robotic sound. You didn’t have to know much about technology to feel something was wrong.

That instinct doesn’t work as well anymore.

Generative AI can now create photos, videos and audio clips realistic enough to sit alongside normal posts on social media, in messaging apps and in search results. A lot of it is harmless: memes, ads, people playing with new tools… Some of it isn’t. A clip gets shared without its context or it’s made from the start to fool people.

This matters a lot in India, where news, family updates and everyday information mostly move through WhatsApp groups and social platforms.

The bigger issue isn’t only that fakes look more real. It’s that the shortcuts we’ve always used to decide what to believe are wearing out. “I saw it with my own eyes” doesn’t carry the weight it used to.

The Internet Now Requires a Different Kind of Attention

Nobody checks every photo or video they scroll past and nobody should have to. We go on context. We know the person who posted it. The video looks professional. The voice sounds familiar. The event seems believable.

Synthetic media can fake every one of those signals.

That’s not a reason to distrust everything. It’s a reason to pay closer attention when a piece of content wants you to do something that matters, like sending money, handing over personal details, downloading a file or forwarding a shocking story right away.

Basic security habits still count, too. Keep your software updated, be careful with personal information and know what your privacy tools can and can’t do. If you often connect to networks you don’t know, you might download the CyberGhost VPN app for your device so your traffic is encrypted on public Wi-Fi.

But keep in mind what a VPN is for. It protects your data while it travels. It has no way of telling you whether a viral video or voice note is real. Part of being digitally literate now is knowing which jobs your tools can do and which ones still fall to you.

Why Spotting the Obvious Mistakes Is No Longer Enough

Early AI images gave people an informal checklist: count the fingers, look at the teeth, check the signs in the background, watch for ears that look strange or glasses that blend into a face.

Those checks still catch sloppy fakes, but leaning on them is getting risky. The models keep getting better and the familiar mistakes keep getting rarer. Meanwhile, real photos can have odd lighting, motion blur or compression artefacts that make them look fake.

Audio is harder still. When a voice you know seems to be asking for help, you tend to react with emotion first and think later.

That’s why context is often more useful than staring at pixels. Instead of only asking “does this look fake?”, ask where it came from.

If a video claims to show a major public event, is any established news outlet covering it? If a well-known figure seems to be backing an investment scheme you’ve never heard of, is it on their verified accounts? If a relative suddenly asks for money, can you check through a different channel?

More and more, the story around a piece of content tells you more than the content does.

When Synthetic Media Meets Familiar Scams

The danger gets concrete when AI is added to scams that already work.

Online fraud has always relied on urgency, impersonation and emotion. Your bank account is about to be frozen. An “official” wants action today. A family member says it’s an emergency. The goal is to keep you from stopping to think.

AI doesn’t need to invent a new kind of fraud. It just makes the old ones more convincing. In March 2025, Reuters examined how digital scams are evolving across India, including the use of AI and impersonation to trick victims.

And it isn’t only people who aren’t comfortable with technology who get caught. A convincing voice clip or video call works because of the trust we naturally give family, colleagues and familiar institutions.

The advice is the same in Jammu and Kashmir as anywhere else in the country. Any unexpected request involving money, passwords, OTPs or sensitive personal details should be confirmed separately. Call the person on the number already saved in your phone. Open your bank’s official app yourself. Go to the organisation’s real website. Any of these will tell you more than minutes spent squinting at whether a face looks generated.

Usually the safest move is the simplest one: don’t act until you’ve verified the request.

The Problem Goes Beyond Believing Fakes

There’s a less obvious side effect: real material gets easier to dismiss.

Once everyone knows convincing fakes are possible, “that’s AI” becomes an easy way to brush off anything awkward or inconvenient. Genuine evidence gets doubted just because faking it is technically possible.

That leaves us in an uncomfortable spot. Trust too easily and you can be manipulated. Assume anything unusual is fake and you’re exposed in a different way. Sensible scepticism sits somewhere in between.

You don’t need an instant verdict on everything. Often “I don’t know yet” is the most honest answer, and it’s worth getting comfortable saying it, because social platforms push the other way: react fast, comment fast, share fast.

Our piece The World We Take for Granted Was Designed makes a related point about how digital systems shape what we see and how we read it. When a platform decides what reaches your screen, questioning it becomes more your job, not less.

Platforms Are Providing More Context, but Not All the Answers

Tech companies are trying out ways to flag synthetic content, such as labels, disclosures and data about where a file came from. These help, but no system catches every manipulated image or video.

Things get messier when content moves between platforms. A video might start out with an AI label, then get downloaded, cropped and forwarded to a group chat without it. A screenshot can drop the label entirely. A short clip can omit the exact context that explains what happened.

In a country where information moves across dozens of languages and regional communities, that gap matters. In September 2025, TheIndian Express reported on the growing sophistication of AI-enabled scams in India, including how voice cloning and impersonation exploit the ways people already communicate.

That’s why automatic detection can’t replace checking things yourself. A reverse image search for an earlier version, a look at the account that first posted a video, or a comparison with reliable reporting can turn up details your eyes would miss.

A clip doing the rounds on WhatsApp may be several forwards away from where it started. Tracking down that original source usually helps more than guessing whether AI made it.

A Useful Habit: Slow Down at the Point of Emotion

Nobody can fact-check their whole day, so the trick is to be selective about when you check.

Convenience plays a part here. Another Greater Kashmir piece, Prisoners of our own convenience, looked at how the habits that make digital life easier can also make us lean on technology more than we realise. Convenience itself isn’t the enemy. The trouble starts when we stop asking questions just because something arrived through an app we use every day.

A strong emotional reaction is a good cue to start asking again. Content built to make you instantly angry, scared, thrilled or shocked is exactly what gets shared before anyone checks it. That doesn’t make it false. It just means a short pause is worth it.

Say a dramatic video lands in your local group, supposedly showing an incident in Srinagar, some kind of disturbance or a surprise announcement from an authority. Before you forward it, see whether the place and date hold up and whether established local outlets are reporting it. If it’s old or has no reliable source, don’t pass it on as something happening now.

The same goes for money. Confirm an urgent voice note separately before you act on it. Check that the original page behind a screenshot actually exists. And if you can’t find reliable confirmation, waiting is a perfectly good choice.

These habits are basic on purpose. Digital literacy has to work in real life: on a commute, over lunch, mid-conversation, or scrolling at midnight. Nobody runs a forensic analysis in those moments, but most people can ask a couple of sensible questions.

Trust Is Becoming Something We Build, Not Something We See

None of this means photos, recordings and videos have stopped being useful evidence. It means their credibility depends more and more on where they came from, how they were made and whether independent sources back them up.

That shift probably matters more than memorising the latest signs of AI. Those signs keep changing, and a giveaway that’s obvious today might be gone after the next software update. What lasts longer is knowing when to hesitate.

This shouldn’t all fall on individual users, either. Schools, universities, newsrooms and platforms all have a part in helping people tell credible information from manipulation. For communities that increasingly rely on smartphones, practical media literacy deserves as much attention as getting people connected in the first place.

A photo can still tell you something real. A voice can still belong to the person you think it does. A video can still show exactly what happened.

But now that realistic media can also be made on demand, trusting it takes one more step: find out where it came from before you decide what it means.

Greater Kashmir