“Do not use Metformin or insulin to treat diabetes—it will kill you.” Seated in front of the red and white flag, the man wearing a black suit and red tie looking sharply at the camera is former Indonesian health minister Dr. Terawan Agus Putranto.
“The root cause of diabetes has nothing to do with carbohydrates, sugar, poor diet, age, or genetics,” the figure of “Terawan” in the video appeared to say. Claiming that “none of the options available to the corrupt pharmaceutical industry” can address this supposed real “root cause,” the fake Terawan then describes a “revolutionary solution” supposedly invented by Asian scientists, alleging it can eliminate the disease within three days.
“Terawan” then warned of various horrors — including blindness, amputation and death — that diabetes patients could suffer if they didn’t act immediately.
But beneath the dramatic promises and warnings lies a troubling reality: while the figure in the video was really Terawan, the voice making the exaggerated claims is not him.
The video used lip sync technology, which employs artificial intelligence and a complex interaction of neural networks to evaluate audio input and generate the accompanying lip motions. The video, which was posted by the Facebook page Gomedico, has been fact-checked by Cek Fakta Tempo. Yet despite its deceptive use of deepfake technology and baseless medical assertions, the video remains active on Facebook. As of writing, the Terawan deepfake mentioned above has racked up 2.7 million views, 7,900 likes, 459 comments, and 654 shares.
Its widespread reach highlights the dangers of unchecked health misinformation, especially when combined with emerging technologies capable of manipulating trust. It raises urgent questions about the role of social media platforms in combating such harmful content.
Dr. Terawan is not the only public figure being digitally impersonated in these deceptive posts. Similar deepfake tactics have been used to fabricate endorsements from two other former health ministers, amplifying the illusion of credibility.
What is most concerning is that almost all of the pages that have been circulating these fakes do not disclose any information about the identity of the persons or organizations operating them. This is true in the case of Gomedico, the page that posted the Terawan deepfake, mentioned above.
Clearly, without the platform’s help, it would be virtually impossible for victims of these deepfakes to hold those behind them accountable.
Paying Facebook for traffic, engagement
It is unclear how the Terawan video got so many views and engagement. The page only has 120 likes and 1.9k followers. With very little following of its own, other ways it could have reached over 2.7 million views could be through the following: (1) going organically viral; (2) getting suggested by Facebook’s “suggested for you algorithm.”
The level of engagement the video reached, however, indicates that it has been boosted. While Meta’s Ads Library does not list down any ads boosted by Gomedico, the fact that the post is not visible on Gomedico’s wall supports this conclusion. There have been instances when active ads on the platform do not show up on the Ads Library.
In the process of working on this story, this author found that many of these deepfakes gained traction by paying Facebook for traffic: either through normal Facebook ads or boosted posts.
To check on the extent of boosted circulation of deepfakes that promote dubious products, this author reviewed data extracted from Meta’s Ads Library platform and checked for ads from February 23, 2024 to October 30, 2024.
For this period, this author found 2,454 ads for the keyword “hypertension cure.” Out of these, 999 ads featured deepfakes which impersonated former health ministers. These ads were boosting 10 unique deepfakes earlier in 2024. Among these ads is a deepfake posted by the Facebook page “Abiding Mist” which impersonated another former Indonesian health minister, Dr. Nila Djuwita Faried Anfasa Moeloek. The deepfaked Moeloek in this ad claimed that Artinormee can cure hypertension “in one week.”
“Abiding Mist” ran a total of 81 ads, all using this video content and text. These ads varied in campaign parameters such as dates, locations, and budgets.
Promoting counterfeits?
Deepfakes often have exaggerated claims of efficacy and cures, although they fall under the category of Jamu (traditional herbal medicine) with misleading narration.
Artinormee is manufactured by Esa Mulia Arakasta, a company located in Bekasi City, West Java.
When confronted about deepfake videos advertising his products, Dede Leswanta, the owner of Esa Mulia Arakasta, claimed that he is not aware of the deepfaked ads. He also said he does not directly control the marketing of his product.
Leswanta told this author that he only handles production and licensing, while two distributors are responsible for marketing and branding themselves. “I can’t sell the products under the distributors’ brands,” he said.
Leswanta said his distributors employ a multi-layered marketing approach that relies heavily on third-party affiliates and online platforms. From what he knows, the distributors utilize affiliation marketing, where individuals promote their products online in exchange for commissions or other incentives.
This approach allows for widespread product promotion without the distributors directly managing the marketers.
Leswanta said that the deepfake videos are likely created and spread by other third-party affiliates without his or his distributors’ knowledge. However, he expressed concern about the deceptive nature of the videos selling fake products. “The problem is that my product is often counterfeited,” he said. He added that it could be the fake ones that are being sold using deepfakes.
In addition, Leswanta told the author that on May 10, 2023, Indonesia’s National Agency of Drug and Food Control (BPOM) raided an unlicensed drug manufacturer in Cibinong, West Java. The operation uncovered counterfeit versions of two Esa Mulia Arakasta products being sold through an online marketplace. Leswanta emphasized that he is cooperating with the BPOM to guarantee the authenticity of his products.
However, Leswanta pointed out that he had accepted the reality of counterfeiting, trying to sell his products at prices that could be three times lower. Many counterfeit products exploit his company’s name for products that were never manufactured. “We have reached a point where we just give up reporting counterfeiting in the marketplace to the authorities.”
Untraceable, unaccountable
Like Gomedico, “Abiding Mist,” the page that ran the Moeloek deepfake does not provide information about the entity or administrators responsible for the page. It also does not provide any information that will allow users to contact it or hold it accountable.
According to its Page Transparency section, Abiding Mist was created on May 27, 2022.
This appears to be a trend for many pages promoting fakes.
Some of the deepfake videos on Terawan were posted through pages which were named after him. There are quite a number of pages (at least 150 were spotted) bearing the former Indonesian health minister’s name on Facebook.
According to the Ads Library link for this specific “Dr. Terawan Agus Putranto” page linked to a deepfake video of Dr. Terawan alongside Kompas TV news anchor Tily Rheabela, denouncing conventional hypertension treatments and calling those who rely on medications or surgery “stupid.”
The page was not just promoting health products for hypertension. It also promotes other products for other illness treatments. When the author accessed it in January 2025, it turns out it ran an ad featuring and impersonating a religious public figure, Zaidul Akbar, promoting a “simple trick to heal prostate in 10 minutes”.
Foreign-operated?
Many of these pages appear to be managed by administrators from other countries even though they are clearly using Indonesian language.
For instance, “Abiding Mist,” the page that posted the Moeloek deepfake, appears to be initially located in Ukraine and the USA, then recently relocated to Ukraine and Kazakhstan.
The “Abiding Mist” ads linked to the site unbiasedharbor.site, which is no longer accessible. Domain registration details are obscured, complicating efforts to identify its operators.
It is a pattern that repeats over and over again. One of the dubious pages named after former health minister Terawan was managed by administrators in Ukraine and Poland.
Facebook page Oblong Handsome, which posted a deepfake impersonating Dr. Siti Fadilah Supari, another Indonesian former Health Minister, is managed by 4 administrators located in Ukraine and 1 administrator located in Indonesia. In the Oblong Handsome post, which was boosted in September 2024, Supari was shown promoting a dubious male virility product.
Yet another page, “Beautiful Casino,” which was similarly discovered using the “obat hipertensi” keyword, is managed by administrators based in Ukraine and the USA.The page ran 108 ads employing a single video and text combination across various campaigns. The deepfake video featured a prominent Indonesian news presenter and former health minister Terawan, endorsing Cardio Normin.
Cardio Normin is registered at the BPOM website as “traditionally used to help relieve mild symptoms of high blood pressure”.
Easy to create
Rapid advancements in artificial intelligence (AI) and the availability of open-source software contribute to the ease of creation and–consequently–the proliferation of deepfakes, tech experts say.
“Deepfake technology is unavoidable because AI is a vast and evolving field,” said Associate Professor of Computer Science, Institut Sains dan Teknologi Terpadu Surabaya (iSTTS), Esther Irawati.
It has come to a point when this already poses a significant threat to public health.
Unfortunately, while tech companies like Meta have community standards and advertising policies designed to curb the spread of misleading and harmful content, the enforcement of these rules often fall short.
Meta’s Community Standards explicitly prohibit “manipulated media” that is likely to mislead or harm. The company states that they will remove deepfake videos if they meet these criteria, especially when they misrepresent medical information.
Additionally, Meta’s Advertising Standard prohibits health-related discriminatory content, misleading claims, inappropriate content, data collection, and restricts weight control/management and body image, and COVID-19- and vaccine-related content.
In practice, however, enforcement is inconsistent. Many deepfake ads featuring fraudulent health claims continue to circulate for days or weeks before being flagged and removed—if they are detected at all. Platforms rely heavily on user reporting and automated systems to identify violations, but these measures often fail to keep up with the sheer volume of new content uploaded daily.
Since 2016, Meta has taken some steps to address the issue, such as partnering with independent third-party fact-checkers that are certified through the non-partisan International Fact-Checking Network (IFCN). However, Meta CEO Mark Zuckerberg recently announced the end of Facebook’s fact-checking program in the US recently and moving to a “community notes” system, similar to X’s.
Solutions
Irawati stressed the importance of responsible AI development and the need for “safe AI frameworks” to prevent the spread of misleading content. She noted, however, that while large corporations often implement responsible AI frameworks and safety measures, there are numerous open-source AI tools available without such limitations.
“This open access allows individuals with malicious intent to readily create and disseminate disinformation,” she said. Methods used for scams have changed and are are continually adapting with the constant evolution of technology, she added.
“In the past, this type of fraud used the telephone, SMS, then WhatsApp. Recently they use the APK invitation, PDF, and now telephone and video call,” Head of Natural Language Processing (NLP) Lab at Institut Sains dan Teknologi Terpadu Surabaya (iSTTS), Joan Santoso explained. This, according to him, is because technology continues to develop.
Santoso stressed the urgency of digital literacy. In the era of AI-generated content, he said that people need to be more critical of the information they consume, not rely solely on technology to determine the truthfulness of information.
Consumers of information, he said, must be extra-cautious about claims related to medicines and cures. “By being information literate, we can fight it.” – with Gemma B. Mendoza/ Rappler.com
Artika Farmita is Tempo’s journalist and fact-checker based in Surabaya, Indonesia, one of Rappler’s #FactsMatter Fellows for 2024.. She has a strong interest in health and science issues, contributing to Tempo’s 2015 investigative reportage on the collusion between pharmaceutical companies, doctors, and hospitals, and has covered science topics from tropical to stem cell research developments. Now, she focuses on raising awareness about online scams, health-related disinformation, and the ethical implications of AI.