How to Spot a Deepfake: Tips for Detecting Fake Media from an expert




All Images are AI generated

We have collected some key tips from a recent Guardian article containing an interview with DeepFake-o-meter creator Siwei Lyu.

As we navigate the digital landscape, our ability to discern what is real from what isn’t becomes increasingly pivotal. Deepfakes, hyper-realistic representations generated by artificial intelligence, are redefining the boundaries of authenticity.

Even though there exist various deepfake detectors, relying on a blend of personal awareness and technology to identify potential deepfakes is recommended. Always approach media with a discerning eye, looking for any incongruity or sign that could signal a manipulation. It’s a dynamic field; as detection methods evolve, so does the technology that creates these deceptions.

The prevalence of deepfakes, particularly on social media, raises valid concerns across various spheres such as identity security and the integrity of information.

Despite the sophisticated AI powering deepfake creation, these synthetic fabrications often leave subtle clues in their wake. Detecting tools in the vein of the DeepFake-o-meter, put forth by security researchers, collate algorithms to aid in verification processes.

Vigilance, coupled with the right tools and an understanding of AI’s capabilities and limitations, offers a defense against the potential deception these forgeries bring into our lives.

Key Takeaways

  • Deepfake detection benefits from a human-AI partnership as technology alone may not suffice.
  • Deepfake detectors use varying algorithms and provide different levels of accuracy.
  • Awareness of AI advancements and detection tools is essential in combating misinformation.


When you receive a call urging you not to participate in important events such as elections, stay vigilant as it might be a piece of falsified audio. Recently, voters were targeted by a robocall with a voice claiming to be President Joe Biden. Analysis of the call revealed varied results, indicating the challenges in detecting synthetic audio. In one instance, a test showed only a single algorithm had a high certainty of the voice being artificial.

Identifying Audio Deepfakes:

  1. Tonal Quality: Listen for a monotonous timbre. Contrary to natural speech, which ebbs and flows with emotion, deepfake audio might lack these variations.
  2. Conversational Cues: Notice the absence of natural conversation markers, for example, irregular breathing patterns or lack of breath sounds before speaking.
  3. Background Sound: Be aware of the ambient sounds. An absence or an excess of background noise, often added to simulate authenticity, can actually be a red flag.

The sophistication of these artificial voices can be attributed to the ample availability of voice data for public figures, allowing for more realistic simulations. However, the task of distinguishing them relies heavily on your auditory perception.

Remember, deepfakes are not solely a visual challenge. The sounds you hear might not always align with reality. If you’re ever in doubt, trust your instincts – if something sounds too uniform or lacks the natural nuances of speech, it might well be an engineered deception.

Examining Imagery for Authenticity

When you’re poring over photographs, pay attention to the minute intricacies. Sometimes, elements that defy reality, like architectural structures sporting unlikely angles or people having one too many fingers, can signal a fake.

In representations of hair, mouths, and even shadows, you might discover hints that indicate an image’s authenticity. Advancements in technology make it harder to spot these flaws, but they’re not entirely gone. A keen eye can still catch them, especially when something feels just too perfect, akin to a work of art.

Performing a reverse image search can be a practical step in verifying an image’s source. This tool helps see if the photograph appears elsewhere or if it’s perhaps been altered. You might be surprised to find inconsistencies across various instances of the same image.

Remember the reports of AI-generated images that slipped past the watchful eyes of some detection tools? They showed what purported to be varied individuals, yet upon a closer look, aspects around the neck or chin might have seemed off. Details like teeth or slight webbing between fingers were giveaways.

Images that carry an unnatural sheen or resemble a painting too closely should raise your skepticism—real life isn’t always glossy and flawless. So, while examining images, rely on your sense of normalcy, as AI occasionally misses the mark on the subtle textures and imperfections that make up our reality.

Identifying Authenticity in Digital Footage

When examining videos of individuals, it’s essential to look for hints that might reveal whether you’re seeing a genuine recording or a deepfake video. Deepfake technology has reached a point where it can create quite convincing fake videos. However, by paying close attention, you can spot inconsistencies that suggest manipulation.

Start by observing the mouth closely. If you notice the lips forming sounds for letters such as “B” or “P” without fully closing, this could be a red flag. Authentic human speech requires precise movements that AI often struggles to imitate perfectly.

Focus on the lighting and shadows too. In videos manipulated by AI, the lighting might not look natural or consistent with the environment. Also, skin tone can appear unnatural—look for sudden changes that don’t match the lighting conditions.

Be aware of how often a person blinks. If blinking seems too frequent or infrequent, or if the eyes show signs of pixelation, these could be signs of digital tampering. A person’s eyes, particularly the eyelids’ movements, are incredibly complex and can be a challenge for deepfake technologies to replicate accurately.

In the age of social media platforms like TikTok, it’s crucial to be vigilant. There have been instances where deepfake videos have been used to spread misinformation. Always consider the source of a video and question its authenticity, especially if it depicts significant political statements or confessions.

Detection Algorithms – While some algorithms are adept at sniffing out fakes by analyzing these elements, they’re not foolproof. Some might return high confidence scores in detecting AI-generated content, while others could be less certain. Therefore, trusting your instincts and looking for the tell-tale signs of manipulation in videos remains valuable.

Share this content

AI News

TikTok's AI Tool Sparks Outrage After Spouting Hitler References
TikTok’s new AI tool, designed to create AI avatars for businesses, has been pulled after it was discovered that the …
Apple logo with EU flag and regulatory symbols overlay
The delay affects millions of iPhone users in Europe … warns about AI-driven travel scams this summer
The company reports a 500-900% increase in phishing attacks, driven by advancements in generative AI technologies like ChatGPT …
Mira Murati
Dartmouth Engineering recently hosted an exclusive conversation with Mira Murati, the Chief Technology Officer at OpenAI, moderated by Dartmouth Trustee …
Hackers exposing AI model vulnerabilities in global effort
This global effort involves ethical hackers and cybersecurity experts, with companies like OpenAI, Meta, and Google continuously working to improve …
Snapchat AI tools enhance augmented reality features.
Learn about the new features and how they aim to compete with other social media platforms …
Humans and robots collaborating in a modern office, representing AI's impact on the workforce transformation.
AI is rapidly automating tasks traditionally performed by humans, transforming the workforce …
Futuristic robot with quill pen, digital code background, glowing Claude 3.5 logo, representing AI innovation.
The model introduces a new feature called Artifacts for enhanced collaboration and content editing …

Latest posts