What Is Deepfake Technology, and How Can You Spot One?

FAPOHUNDA GRACE

Imagine receiving a video of a popular Nigerian celebrity promoting an investment platform and promising unusually high returns. The person's face looks familiar, their voice sounds convincing, and the video appears genuine.

But there is a problem: the person never made that video.

It could be a deepfake.

Deepfake technology uses artificial intelligence (AI) to create or manipulate images, videos and audio so that they appear to show or sound like real people doing or saying things that never happened.

As AI tools become more accessible, creating convincing fake content is becoming easier. These videos and voice recordings can be used for entertainment and legitimate creative projects, but they can also support financial scams, impersonation, misinformation and harassment.

So, what exactly is deepfake technology, how does it work, and how can you recognise manipulated content before believing or sharing it?

What Is Deepfake Technology?

Deepfake technology refers to the use of artificial intelligence and machine learning to generate or manipulate digital content, particularly images, videos and audio, to make it appear authentic.

The word deepfake combines two terms:

Deep learning: A type of machine learning that uses artificial neural networks to recognise patterns and generate content.
Fake: Content that has been altered or created to misrepresent reality.

For example, AI can be used to replace one person's face with another person's face in a video, recreate someone's voice or generate a realistic video of a person appearing to deliver a speech they never gave.

Deepfakes are not limited to videos. They can also appear as:

Fake images: AI-generated or manipulated photographs of people or events.
Fake videos: Clips showing someone appearing to speak, move or behave in ways that never happened.
Fake audio: Cloned voices used to make it sound as though someone said something they never said.

The technology is also used in legitimate applications, including film production, visual effects, entertainment and creative projects. The concern is when manipulated content is presented as genuine to mislead or harm people.

How Does Deepfake Technology Actually Work?

Deepfakes are created using AI models trained to recognise patterns in images, facial movements, speech and other forms of data.

The exact process depends on the type of deepfake being created.

Here are three common methods.

1. Face swapping

Face swapping involves replacing one person's face with another person's face in an image or video.

An AI model analyses facial features, expressions and movements before generating a replacement that follows the original person's head movements and expressions.

For example, a creator could take a video of one person speaking and digitally replace their face with that of a celebrity.

2. Voice cloning

Voice cloning uses AI to reproduce the sound and characteristics of a person's voice.

Depending on the system and available recordings, the technology can learn patterns such as:

Tone and pitch.
Pronunciation.
Speaking rhythm.
Accent and vocal characteristics.

The cloned voice can then be used to generate new speech.

This creates a serious impersonation risk because someone could receive an audio message that sounds like a relative, friend, business partner or public figure.

3. AI-generated video and facial reenactment

Some AI systems can generate a person's face or manipulate their facial expressions and mouth movements to match new speech or other movements.

This can make a video appear to show someone delivering a statement they never actually made.

The technology can also generate entirely artificial people and scenes rather than manipulating an existing recording.

8 Warning Signs That a Video, Image or Audio May Be a Deepfake

Not every deepfake contains obvious errors. Some are sophisticated enough to look convincing, especially when viewed on a small phone screen.

However, certain inconsistencies may help you identify suspicious content.

1. Unnatural facial movements

Pay attention to how the person's face moves.

Look for facial expressions that appear stiff, inconsistent or disconnected from the rest of the person's movements.

Some manipulated videos may also contain unusual blinking, awkward head movements or facial features that change unexpectedly.

2. Lip movements that do not match the speech

Watch the person's mouth carefully while listening to the audio.

If the lips appear to form different words from what you hear, the video may have been manipulated.

However, poor internet connections, compression and editing can also cause audio and video to become misaligned.

3. Strange lighting and shadows

Look at the lighting around the face, neck and background.

A manipulated face may have lighting that does not match the rest of the scene.

For example, the person's face may appear unusually bright while the surrounding environment is darker, or shadows may look inconsistent.

4. Unnatural skin texture or facial details

Some generated videos may contain unusual skin textures, blurred facial edges or inconsistent details around the eyes, ears, teeth and hair.

Pay attention to whether the person's appearance changes between frames.

These imperfections are not guaranteed to appear in modern deepfakes, but they can provide useful clues.

5. Unusual voice patterns

When listening to suspicious audio, pay attention to:

Unnatural pauses.
Unusual changes in pitch.
Robotic or overly smooth speech.
Strange pronunciation.
Background noise that does not match the recording.
A speaking rhythm that sounds different from the person's normal voice.

However, modern voice-cloning systems can reproduce speech convincingly, so a familiar-sounding voice is not enough to confirm someone's identity.

6. Inconsistent background details

Examine the environment surrounding the person.

Look for objects that appear distorted, disappear unexpectedly or change shape between frames.

Also check whether the lighting, reflections and background movements make sense.

7. Suspicious or sensational claims

A deepfake may be designed to make people react emotionally.

Be cautious when a video shows a public figure announcing an unbelievable investment opportunity, a celebrity endorsing a questionable product or an official making a shocking statement without any credible confirmation.

Ask yourself whether the claim has been reported by reliable sources or published through the person's verified channels.

8. No original source or reliable confirmation

A video circulating only through WhatsApp groups, anonymous social media accounts or forwarded messages deserves extra scrutiny.

Check who first published it, when it appeared and whether the person or organisation involved has confirmed it.

A video with thousands of views and shares is not automatically authentic.

How to Verify a Suspicious Video or Voice Recording

Spotting visual or audio inconsistencies is useful, but checking the source is often more reliable than relying on appearance alone.

Use these practical steps.

Step 1: Find the original video

Do not rely only on the version shared in a WhatsApp group or social media post.

Search for the original recording, full interview, official announcement or complete speech.

A short clip may have been taken out of context, edited or manipulated.

Step 2: Check the person's official accounts

If a video claims that a Nigerian celebrity, business owner, politician or public figure is promoting an investment scheme, check their verified social media accounts and official website.

Look for a matching announcement.

If the claim cannot be confirmed through reliable sources, treat it as unverified.

Step 3: Compare reports from credible news organisations

For major announcements, check whether reputable news organisations have independently reported the same information.

Do not rely on multiple accounts repeating the same original unverified post. Independent confirmation is more useful.

Step 4: Use reverse image search

If you suspect an image or video thumbnail has been manipulated, take a clear frame or screenshot and use reverse image search to look for earlier appearances of the image.

Tools such as Google Lens can help you find where an image has appeared online.

This may reveal whether an image was taken from an older event or reused in a misleading context.

However, reverse image search cannot independently prove that a video is genuine or fake.

Step 5: Verify unexpected voice messages independently

Imagine receiving a WhatsApp voice note that sounds like your friend asking you to send ₦100,000 urgently.

Do not send the money just because the voice sounds familiar.

Call the person using a number you already trust. If the request is unusual, confirm it through another channel.

Families and businesses can also agree on a private codeword or verification phrase for urgent requests.

Step 6: Check the date and context

A genuine video can still be misleading if it is old, edited or presented as a recent event.

Check when it was originally published, where it was recorded and what was happening at the time.

How Deepfakes Are Being Used in Scams and Online Fraud

Deepfake technology can make traditional online scams more convincing because criminals can imitate the appearance or voice of people victims already trust.

Here are some common ways it can be abused.

1. Fake celebrity investment endorsements

A manipulated video may show a popular celebrity or business personality promoting a cryptocurrency, investment platform or money-making opportunity.

The aim is to use the person's reputation to convince viewers to deposit money.

2. Voice-cloning emergencies

A criminal may use a cloned voice to impersonate a family member or friend and claim to be in an emergency.

The victim may be pressured to transfer money immediately without verifying the caller's identity.

3. Fake executive or business instructions

A manipulated voice or video may impersonate a company executive and instruct an employee to make a payment, share confidential information or change banking details.

This can be part of a business email compromise or impersonation scam.

4. Fake news and public misinformation

Deepfakes can make it appear that politicians, public officials or other influential people have made statements they never made.

Such content can be used to spread false information, damage reputations or create confusion during sensitive events.

5. Identity abuse and harassment

A person's face or voice may be inserted into manipulated sexual or embarrassing content without their consent.

This can cause serious emotional, reputational and privacy harm.

What Should You Do If Someone Uses Your Face or Voice in a Deepfake?

If you discover a fake video, image or audio recording using your identity, take the following steps:

Preserve evidence: Save the original link, screenshots, account name, publication date and relevant messages.
Report the content: Use the platform's reporting options for impersonation, manipulated media, harassment or non-consensual intimate content, as applicable.
Warn people who may be affected: Let friends, family, colleagues or customers know if someone is using your identity to request money or spread false information.
Secure your accounts: Change compromised passwords, enable two-factor authentication and review unfamiliar account sessions.
Report serious threats or fraud: If money has been stolen, contact your bank immediately. For threats, extortion or other suspected crimes, preserve evidence and report to the appropriate law-enforcement authorities.

Do not pay someone who threatens to release manipulated content without seeking help and reporting the incident.

Can AI Detection Tools Identify Every Deepfake?

No. AI detection tools can help analyse suspicious content, but they are not perfect.

Some tools examine facial movements, image patterns, audio characteristics and other technical features to estimate whether content may have been manipulated.

However, deepfake technology continues to improve, and detection systems can produce false positives or miss sophisticated manipulations.

The US Government Accountability Office has also highlighted the limitations of existing detection technologies and the difficulty of reliably identifying increasingly sophisticated deepfakes.

Digital watermarks and content provenance systems can provide additional information about how media was created or edited. However, not every image or video carries this information, and the absence of a watermark does not prove that content is fake.

The safest approach is to combine technical checks with source verification and independent confirmation.

7 Ways to Protect Yourself From Deepfake Scams

You do not need advanced cybersecurity knowledge to reduce your exposure to deepfake-related fraud.

Start with these habits:

Verify unexpected requests for money, even when they appear to come from someone you know.
Call people back using numbers you already trust.
Create private verification phrases with close family members.
Avoid sharing sensitive personal information with unverified callers.
Check public figures' official accounts before believing promotional videos.
Avoid forwarding sensational videos or voice notes without verifying their source.
Limit the personal photos, voice recordings and videos you share publicly, particularly where they reveal sensitive information.

For Nigerian users, be especially cautious of fake investment endorsements, celebrity giveaway videos, cloned voice calls and messages claiming that someone urgently needs money.

Conclusion

Deepfake technology is changing how images, videos and audio can be created and manipulated. While it has legitimate uses in entertainment, education and creative production, it can also be exploited for scams, misinformation, impersonation and harassment.

The most important lesson is that seeing someone's face or hearing a familiar voice is no longer enough to confirm that the person actually said or did something.

Before believing a suspicious video, sending money or sharing a dramatic voice recording, verify the source and confirm the information through an independent channel.

CYBERSECURITY

Pages

Start Up