Deepfake: How It Is Used in Marketing and How to Recognize It
Deepfake is artificially created videos, images, or audio recordings generated using artificial intelligence (AI) technologies. The term comes from combining two words: "deep learning" and "fake."
Bruce Willis' deepfake in the ad
Below, we will explain in more detail what deepfake is, how it is used, and how to protect against it.
The history of deepfakes
The origins of AI-generated content date back to the 1990s. In 1997, Video Rewrite synchronized human lip movements with computer-generated sound, achieving highly realistic speech mimics. The term "deepfake" itself emerged in 2017 when a Reddit user under the pseudonym Deepfakes created videos using AI and posted them. In these videos, celebrity faces were artificially superimposed onto the bodies of actors.
Not too long ago, many AI videos looked like this. Source: YouTube channel Robot Named Roy
Since the late 2010s, deepfakes have been gaining popularity. Gradually, the quality of such materials has improved. The development of AI technologies opened up new possibilities: creating realistic images, voice processing, and integrating computer graphics with real video. Now, artificial intelligence is accessible not only to specialists but also to regular users thanks to professional software and user-friendly applications.
Nowadays, deepfakes can fool anyone
The scope of AI-generated content is constantly expanding: advertising, marketing, gaming, science, education, fashion, design, journalism, and art.
The many uses of deepfakes: good and bad
Deepfake technology is used in various fields. In cinema, it creates digital replicas of stars, saving budgets on filming. However, this use of AI is considered controversial by many, as actors and other creative workers lose income.
Deepfake is also used to create high-quality voice imitation, which can improve movie dubbing in some cases. In museums and galleries, this technology is an interesting tool for creating interactive exhibits.
Unfortunately, deepfake technology is often used for harm rather than good. For example, malicious actors create deepfakes featuring well-known people, such as politicians, celebrities, or business leaders, to manipulate public opinion or commit crimes (e.g., luring people to fraudulent websites).
The unauthorized use of digital likenesses is considered an infringement on personal or brand rights. There have been cases where fraudsters used deepfake technology to imitate the voice and appearance of a department head, misleading company employees. Under the guise of following the manager's instructions, employees transferred funds to the criminals' accounts.
There have also been attempts to post fake messages on behalf of well-known entrepreneurs, where deepfakes were used to advertise nonexistent financial opportunities, directing victims to malicious websites for personal gain.
Deepfake technology is now considered controversial by many and often sparks conflict. For example:
In 2023, U.S. screenwriters went on strike due to disagreements with producers. The strike lasted 148 days, becoming the second longest in WGA history. In September of that year, the parties reached an agreement. One of the main points of contention was the impact of artificial intelligence on the industry. Writers wanted AI to be used only as an auxiliary tool, not as a replacement for creative workers.
In May 2024, actress Scarlett Johansson accused OpenAI of using her voice without permission to create a voice assistant. The actress claimed that the assistant, Sky, sounded suspiciously like her, and OpenAI’s founder even hinted at this on Twitter. Her lawyers demanded that OpenAI disclose details about the development of this voice.
“The founder of OpenAi posted this. Many thought it was a reference to the movie “Her”, in which Johansson starred
Deepfakes and AI in marketing
AI-generated content solves the same business tasks as traditional marketing methods: it showcases products, attracts new users or customers, makes your company more recognizable, and increases sales. Here’s what else this tool can do:
- Lowers advertising costs. Creating AI-generated videos and photos does not require hiring people for shoots (which means there’s no need to pay them). Deepfakes save companies significant amounts on advertising and can easily create images of any people, even celebrities. However, consent is necessary, and many stars and rights holders are against AI. If you use someone’s likeness, voice, or copyrighted work without permission, you risk facing large fines. AI-generated videos of others also cannot be used without permission.
Shah Rukh Khan in Cadbury ad. Small companies in India have been allowed to use his deepfake to promote their brands
- Accelerate content adaptation. With AI technologies, brands can quickly tailor their advertising to different conditions. Creating content for various platforms becomes simpler and faster: whether you need a video for social media or an image for outdoor advertising, AI can help. For example, deepfakes can create multilingual videos or change backgrounds without the need for real shoots. This way, companies avoid significant expenses.
Brand Zalando used AI in its video — actress Cara Delevingne spoke multiple languages
- AI enables creativity. Brands can bring bold ideas to life and surprise their audience. For example, some companies offer the option to virtually try on glasses and see how they will look.
You can try out different hair color options with Perfect365 Web App
How to create a deepfake
Deepfakes are used for harmless fun, such as seeing what you'd look like when older or as the opposite gender, as well as for more serious purposes.
Zao. In 2019, Zao became a sensation — it allowed users to insert their faces into scenes from popular films. However, the service quickly faced criticism due to privacy concerns. It's no longer available for download in Russia.
FaceSwap. This free app swaps faces in videos. The developers claim to adhere to strict ethical standards and aim to minimize the potential for misuse.
ToonMe. A simple app that processes photos and turns people into cartoon versions of themselves.
Source: toonme.com
DeepFaceLab. This application offers advanced tools for creating deepfakes, but it requires payment and can be difficult for beginners to master. However, its performance is impressive — the service replaces faces in videos with very realistic effects. DeepFaceLab not only swaps faces but also rejuvenates and syncs lip movements with speech.
An example of how the tool works. Still from Deepfakery video on YouTube
FaceApp. One of the most popular apps. It is available for both Android and iPhone. Want to see how you’d look in a different form? In this app, you can try on the appearance of the opposite gender, different hairstyles, glasses, or a beard. And if you'd like to correct some imperfections in your appearance, the app can remove flaws with a filter.
FaceApp
Avatarify. This app transforms static images by adding realistic facial expressions and articulation. It overlays sound recordings and favorite songs onto photos.
How this technology works
Ian Goodfellow developed Generative Adversarial Networks (GANs). These are two algorithms that work together to create AI-generated content, such as deepfakes. One algorithm studies real objects and generates a new image, while the other "criticizes" the generated content — checking how natural it looks. They work against each other and learn in the process, which leads to continuous improvement in the quality of deepfakes.
How to spot deepfakes
So how can you avoid being fooled? If you're unsure about the authenticity of a video, try looking for these oddities:
Frame rate. Pay attention to how often the frames change. Deepfakes usually "blink" less frequently than real videos, and sometimes this looks unnatural.
Sound. There are programs to alter voices, but many deepfakes only use the image. Be cautious if the video is silent or if the audio doesn’t match the visuals. Watch for lip sync issues.
Face and body. Creating realistic fakes of a whole person is challenging. Most often, only the face is swapped. Check for mismatched proportions between the face and body, skin tone, and whether facial expressions align with body movements.
Duration. High-quality video generation often requires significant time to train the algorithm, so fake clips tend to be short.
Mouth. AI technology struggles to accurately recreate teeth and tongue movements when someone is speaking. Blurriness inside the mouth is often a sign of a fake video.
Details. This is a weak spot for AI programs. Look for minor inconsistencies: unrealistic facial hair, overly smooth or wrinkled skin, fake moles, and unnatural lip colors.
Check out how neural networks make mistakes in the article «Neural Network Hallucinations: Types, Causes, and Mitigation.”
- Special software. For a more thorough check, use programs that detect deepfakes. This significantly reduces the risk. Here are a few options:
1. KaiCatch. An AI-based tool that detects unusual elements and manipulations in photos and videos.
2. Forensically. A free program for analyzing and verifying the authenticity of images.
3. Image Edited. A service for checking photos that works at the pixel level.
4. Sentinel. An AI tool used by governments, militaries, and companies to combat fakes. Upload a photo or video to the website, and the AI system will analyze its authenticity. If it’s a deepfake, Sentinel will show how it was created.
5. Ghiro. An automated service designed for forensic analysis of large volumes of images.
How to protect yourself from deepfakes
Deepfake technology will continue to evolve. If you want to protect yourself and your loved ones:
Limit access to your personal data. Avoid posting photos and videos on social media and other platforms that could be used to create deepfakes. If possible, make your accounts private.
Educate your loved ones on how to recognize deepfakes and the risks they pose. Show your elderly relatives what such photos and videos look like.
Verify information. If you have doubts about the authenticity of a video, try finding information from other sources. Use different channels to verify the information, for example, by calling the person whose image has been used in a deepfake. And if you need to confirm you’re talking to a real person, ask something that only you two would know.
Follow cybersecurity rules. Use strong passwords and change them regularly. Install antivirus protection and keep it updated. Regularly back up important data.
Conclusion
Deepfake is a powerful technology that can both entertain and mislead. On one hand, it offers opportunities for creativity and self-expression, while on the other, it requires the development of critical thinking and tools for verifying the authenticity of information. In the future, we can expect further advancements in deepfakes, making it essential to stay vigilant and recognize their potential for both positive and negative uses.