Last week, Google made a talking AI face available to businesses. It is called Live Avatar, and it runs on Google’s Gemini 3.8 Live model. It can look at you, listen, and answer out loud with lips that move in time with its words. It works in 97 languages.
That raises a question many people are now typing into Google: how can you tell if a video call is an AI avatar and not a real person? This guide explains what these avatars are, where you will meet them, and simple checks that still work in 2026.
What is Google’s Live Avatar, in plain words?
Live Avatar is a computer-made face that can hold a live video conversation. It hears what you say, thinks of a reply, and speaks it with a moving face and matching lip movements. Businesses can place it on a website, in an app, or on a screen in a store.
Google announced it was generally available on September 25, 2026. That means any business customer can now use it, not just testers. Google’s examples include:
- An insurance helper that watches a video of your damaged car while filling in your claim.
- A car-shopping assistant for Autotrader, run by Cox Automotive.
- Customer service inside Salesforce’s Agentforce tools.
Here is the key idea in one line: an AI avatar is a video “face” for a chatbot, so the chatbot can talk to you like a person on a call.
Will you actually meet AI avatars on video calls?
Yes, and mostly in harmless places. Expect them on company help pages, shopping sites, bank and insurance apps, and screens at airports or shops. These are honest uses. The business wants to answer your question faster. The risk comes from scammers who use similar tools to pretend to be someone you know.
Honest companies usually tell you the helper is AI. A label, a small badge, or a spoken line like “I’m a virtual assistant” is common. In the European Union, the EU AI Act includes a rule that people must be told when they are talking to an AI system. In other places, rules are still catching up.
Google also says it adds a hidden watermark, called SynthID, to all the sound and video Live Avatar makes. A watermark here means an invisible signal baked into the video that special software can detect later. You cannot see it with your eyes, though.
Why AI avatar video calls matter for scams
This is not a far-off worry. In early 2024, an employee at the engineering firm Arup in Hong Kong joined a video meeting with what looked like the company’s finance chief and other colleagues. Every other person on that call was fake. The employee made 15 transfers worth about US$25 million before anyone realised.
And most of us are bad at spotting fakes. In a 2025 study by iProov, 2,000 people in the UK and US were shown a mix of real and fake images and videos. Only 0.1% got every answer right. People were also 36% worse at spotting fake videos than fake photos.
A “deepfake” is a fake video or voice made by AI to look and sound like a real person. Google’s Live Avatar is not a deepfake tool. Making a custom face from a real person’s photo is locked behind an approval process. But other tools, some free, are far less careful.
The simple takeaway: your eyes alone are no longer a reliable test of who is on the other end of a video call.
How can you tell if a video call is an AI avatar?
The most reliable way is to test things a fake cannot easily copy. Ask the person to do something unexpected with their hands or camera. Ask a question only the real person could answer. Then check through a second channel, such as calling them back on a number you already have.
Here are the checks, from quickest to strongest.
1. Watch the mouth and face closely
- Do the lips match sounds like “p”, “b”, “f” and “v”? These need the lips to close or touch the teeth.
- Do the teeth, ears or glasses look blurry or “melt” slightly when the head turns?
- Is the blinking too regular, or missing for long stretches?
- Do the same small gestures repeat, like a head nod on a loop?
Be careful: newer avatars are very good at this. Passing these checks does not prove someone is real.
2. Ask for a surprise movement
Ask the person to wave a hand slowly in front of their face. Or ask them to turn fully sideways, or hold up today’s newspaper or a random object from their desk. Many fake faces glitch, blur or “smear” when something passes in front of them. Also ask them to turn the camera around and show the room.
3. Interrupt and change the topic
Cut in mid-sentence. Ask, “Sorry, what was the last word you said?” Then jump to a personal topic, like a joke you share or a recent family event. Some live fakes pause too long, or give a smooth but vague answer.
4. Use a family safe word
Agree on a secret word or phrase with family members that you never post online. If a “relative” calls in a panic asking for money, ask for the word. The FBI recommends this exact step to protect against AI-made voices and videos.
5. Hang up and call back
This is the strongest check of all. End the call and contact the person yourself, using a number or chat you already trust. Do not use a link or number the caller gives you. Arup’s loss might have been stopped by one call back to the real finance chief.
Warning signs that matter more than the face
Scammers often give themselves away by what they ask for, not how they look. Be extra careful if a video call includes any of these:
- Urgency: “This must happen in the next hour.”
- Secrecy: “Don’t tell anyone else about this yet.”
- Money or codes: requests for a bank transfer, gift cards, crypto, or a one-time password.
- Odd channel: your boss or relative suddenly using a new account or app.
- Camera “problems”: the video keeps freezing right when you ask them to move.
These red flags are the same ones we covered for AI voice scam calls. Adding a face just makes the lie more convincing. If you want to practise spotting fake faces in short videos, our guide on AI-generated doctors on TikTok shows the same clues.
What should businesses do?
If your company plans to use AI avatars, be open about it. Tell customers clearly that they are talking to an AI. Never let an avatar ask for passwords or payments. For staff, set a simple rule: no money moves because of a video call alone. A second check, like a call back or a sign-off from another person, must always happen.
Frequently Asked Questions
Is Google’s Live Avatar available to ordinary people?
Not directly. Google released it for businesses through Gemini Enterprise on September 25, 2026. You are most likely to meet it as a helper on a company’s website, app or in-store screen.
Can an AI avatar copy my face?
Google says custom avatars made from a real person’s photo and voice are only allowed for approved business customers. Other AI tools may be less strict, so be careful about sharing clear videos and voice clips of yourself publicly.
Can I spot an AI avatar just by looking?
Sometimes, but not reliably. A 2025 iProov study found only 0.1% of 2,000 people spotted every fake. Surprise movements, personal questions and calling the person back work much better than staring at the face.
What is the single best way to avoid a deepfake video call scam?
Hang up and call the person back on a number you already trust. Never send money or share codes based only on a video call, no matter how real the person looks.
