Using AI safely
How to spot AI-generated images, video, voices and documents
A few years ago you could spot fake images by counting fingers. Not any more. AI can now produce convincing photos, videos, voices and documents in seconds. The most reliable defence isn’t a sharper eye. It’s checking the source and verifying anything that matters through a separate channel.
The short version
What actually works
- Check where it came from before judging how it looks.
- Verify any request for money or access through a known channel.
- Use reverse image search and look for content credentials.
- Slow down when content is designed to shock or rush you.
Why looking harder isn’t enough
The visual tells people relied on, like strange hands, blurry text and odd backgrounds, are disappearing as tools improve. Fakes are also often viewed small, on a phone, compressed by social media. Some clues still help, but they’re not reliable on their own.
Clues that sometimes still appear:
- Text, logos or signs that don’t quite make sense
- Lighting and shadows that don’t match
- Jewellery, glasses or teeth that blur or change between frames
- Voices with unnatural pauses or flat emotion
- Lip movements slightly out of sync with speech
Checks that still work
Check the source
Who posted it first? Is it from an official account or a reputable news outlet? A dramatic video from an account created last week deserves suspicion.
Look for other reporting
If something genuinely happened, other credible sources will be reporting it.
Reverse image search
Use Google Lens or TinEye to see where an image has appeared before, and whether it’s been altered or taken out of context.
Look for content credentials
Some cameras, editing tools and AI generators attach content credentials (based on the C2PA standard) that record how an image was made or edited. Some platforms show a small “CR” label or “AI info” tag. Absence doesn’t prove something is fake, but credentials can confirm origin.
Verify through a different channel
If a video call, voice message or document asks you to pay, share data or grant access, confirm it by calling the person on a number you already have.
Deepfakes in business fraud
Criminals have used AI-generated voices and video to impersonate executives and suppliers, sometimes convincing staff to make large payments. Protect your business with process, not detection:
- Payment changes and urgent payments are always verified by calling a known number.
- Two people approve large or unusual payments.
- No exceptions for seniority. A request from “the CEO” follows the same rules.
- Agreed verification phrases for executives and finance staff, if the risk is high.
See our invoice scams guide.
Fake documents too
AI makes it easy to produce realistic invoices, payslips, bank statements and ID images. Verify important documents with the issuer directly, especially in hiring, lending and supplier onboarding.
If you’re the target of a deepfake
If someone has created fake intimate images or videos of you, or is using AI fakes to harass you, the eSafety Commissioner can help get content removed. If you’re in danger, call Triple Zero (000).
Common questions
Are AI detection tools reliable?
Not reliable enough to depend on. They miss fakes and flag real content. Use them as one clue, not a verdict.
How can I tell if a phone call voice is cloned?
Often you can’t. That’s why a callback on a known number or a family code word works better. See family emergency and voice scams.
Should I share something if I’m not sure it’s real?
No. If you can’t verify it, don’t pass it on.
Sources
This guide is general information, not legal or financial advice. If you think you’ve lost money or personal information, contact your bank and the services named above straight away.
Keep reading
Related help
Talk to NovaSoft
Want a hand with this at work?
These guides are free to use. If your business wants help putting the checks, accounts and devices in place, our team can do it with you.