13 hours ago
AI in the Middle East: Fact-checkers fight fire with fire
Conflicts, crisis and information vacuums in the Middle East mean AI-generated disinformation is overflowing. But fact-checkers are turning Arabic-language AI into weapons on their mission to counter disinformation.
In 2014, the news about a miracle herb from Lebanon that could treat skin cancer was spread on social media. If rubbed on the skin, the herb promised to immediately cure all symptoms. It was that promise which made Moath Althaher's cousin stop her chemotherapy treatment for skin cancer. She passed away just a few months later.
For medical student Moah Althaher, this was the moment he decided that he needed to do something to fight health-related disinformation. He had just witnessed the power of such misinformation to cause severe damage firsthand. The idea for the fact-checking platform Fatabyyano was born, initially with a focus on health-related disinformation.
A real risk to social cohesion and regional stability
Misleading health claims are only a drop in the ocean in the vast disinformation landscape in the Middle East. The nature and content of disinformation in the MENA region vary significantly, depending on the specific country context. Ongoing political conflicts and instabilities such as the outbreak of the Israel-Hamas war or the recent fall of the Assad regime in Syria are the main drivers for the spread of toxic content, increasingly created by AI tools.
After witnessing the impact of health misinformation firsthand, Moath Althaher launched the fact-checking platform Fatabyyano Image: Yuan-Bin Zhao
Walid Al-Saqaf, a professor of journalism and AI at the Doha Institute, explains: "The MENA region's fragmented and government-controlled media, as well as its limited digital literacy, heighten vulnerability to AI-driven disinformation. AI has substantially accelerated the proliferation of disinformation in the MENA region, thereby posing serious risks to societal cohesion and regional stability."
Al-Saqaf also says that the development of AI technologies has exacerbated the spread of disinformation as more sophisticated deepfakes and AI-generated images and audio can be created with the help of easily available and cheap AI tools.
Next level disinformation
A single piece of disinformation spread in Arabic could potentially reach an audience of more than 400 million Arabic speakers. Fatabyyano founder Moath Althaher has noticed a change in both the character and quality of disinformation that has been spread over the years.
While in the early days of Fatabyyano, disinformation was often created with a silly or entertaining undertone which made it easy to detect, it now looks different. "Disinformation has gained a more serious tone and has become more sophisticated due to the development of AI tools," says Althaher. "Disinformation is now on another level and can even appear as deepfakes, using the voice or image of public figures."
A piece of disinformation that was fact-checked by the Fatabyyano team was a Facebook reel that appeared to show Ahmad Fakhouri, host of the BBC show "Trending," presenting a herbal spray that could help people quit smoking. While the host elaborates on the health benefits of the spray, instructions on how to buy it are provided right below the deepfake video. Fact-checking such a video now requires a multilayered approach: It needs to be checked for its claim, as well as the voice and images used.
Fatabyyano debunked a deepfake of BBC host Ahmad Fakhouri promoting a fake anti-smoking spray, showcasing the complexity of modern fact-checking Image: Fatabyyano
There has also been a shift in the motivation behind the spreading of disinformation. When Moath first started his fact-checking work, fake news were often being spread for entertainment or communication purposes. Now the main motivation is to deceive in order to generate profit - as the immediate buying instructions in the case of the herbal spray video demonstrates: It sells at 25 Euro for a pack of four.
Moath Althaher is part of a vibrant community of fact-checkers in the Middle East who are trying to keep up with the huge amounts of false and misleading information. AI tools seem like potential allies in an impossible task, however they come with their own set of challenges.
"The data that Arabic language models are trained on is of low quality in comparison to English language models, both in terms of quality and quantity," Althaher explains. "This leads to limited performance."
The complexity of Arabic, with its many dialects and writing styles, makes AI training and language processing especially challenging Image: Floris Van den Abbeele/DW
The complexity of the Arabic language further complicates the situation. With dozens of regional dialects, the creation and processing of datasets become increasingly difficult. Abed Khooli, an AI researcher based in Ramallah, explains: "Arabic also uses different accents that can alter the meaning of words, varying writing styles, and numerous regional and country dialects, in addition to Modern Standard Arabic (MSA), which complicates processing."
Digging deeper: Understanding the underlying narratives
The Fatabyyano team is currently developing the tool, which is an AI-powered instrument designed for real-time fact-checking and trend analysis in Arabic. This tool aims to keep pace with the immense amount of high-quality, AI-generated disinformation. goes beyond simple fact-checking: It helps to uncover and understand the underlying disinformation narratives, which is especially helpful in the health sector.
To overcome the impediment of low-quality data, the model has been trained on a decade's worth of fact-checking done by the Fatabyyano team, which presents a significant qualitative and quantitative dataset. The tool can respond to any requests by users based on this curated data, as well as draw on other trusted and specified sources. Furthermore, the team utilizes advanced AI tools across multiple trusted platforms to build a larger database. "A piece of misinformation is always the symptom of a narrative. If you understand the narrative, you can tackle the problem at its root," explains Althaher.
Palestinian Territories: Fact-checking in times of war
Narratives play an important role in the ongoing conflict between Palestinians and Israelis and shape the public discourse. "Decades of tension and violence between Palestinians and Israelis have led to an environment where misinformation and disinformation has flourished," explains Elliot Higgins from Bellingcat in a BBC podcast . People are generally less likely to question information that confirms their views. Following the October 7th terrorist attack on Israel by Hamas, the level of disinformation was at an unprecedented scale.
Journalists in Gaza work under extreme conditions, verifying information despite displacement and limited access to internet and electricity Image: Hamed Sbeata/Middle East Images/picture alliance
Fact-checkers from the Palestinian Territories do not only work in an environment fertile for disinformation, they also operate out of a war zone. Riham Abu Aita, based in Ramallah, founded the fact-checking organisation Kashif in 2019. The majority of fact-checking requests that the Kashif team currently receives are related to the situation in Gaza. One of Kashif's team members, Ahmad, is based in Gaza and is facing tough circumstances on the ground. He has been displaced more than three times and is currently with his family in Khan Younis. Access to connectivity or information are limited and Ahmad often relies on a wide network of people he knows on the ground and calls them to verify any information for fact-checks or reaches out to journalists or medical staff.
Some of the fact-checking requests that the Kashif team receives are related to a term coined "Pallywood," which refers to a disinformation campaign that accuses Palestinians of faking suffering and civilian deaths in their conflict with Israel.
One particularly disturbing aspect of these claims involves accusations that images of deceased children are staged and show dolls rather than children. In such cases, the Kashif team relies on their network to contact the children's families for testimonies or photos of the children — in order to prove that the accusations of staging are false.
"We were not able to respond to all those requests because we are a small team, that's why we decided to integrate a chatbot into our website," explains Riham Abu Aita. The "Kashif bot" responds to requests by users to fact-check specific information or answer questions. It functions as an interactive bot that is trained by the Kashif team and provides the users with verified and trusted information only. The bot makes sure that the users receive answers quickly, as the requests do not have to be responded to by individual team members. The Kashif bot also helps the team understand which topics are of interest to the users of Kashif.
Riham Abu Aita co-founded Kashif, a Palestinian platform promoting media literacy and fighting misinformation through fact-checking and ethical journalism Image: DW Akademie
"The process of building the chatbot was much more complicated than expected," recalls Riham Abu Aita. "It took us more than six months to make sure that the chatbot was treating our content properly. It was difficult to even find a chatbot provider, as most providers do not focus on Arabic language." So far, the performance of the chatbot has been satisfying; People are using it frequently and the team is continuing to train it.
However, Abu Aita acknowledges that her team's fact-checking capabilities are constrained by the environment in which they operate. "Sometimes the public asks us for fact-checks that would need security or military information," she explains. "But we do not have access to such information."
Lebanon: Battling disinformation on social media
Another hotspot where false and misleading information circulates is in Lebanon. There, disinformation is mostly shared and spread among friends or family members on WhatsApp groups, with content and interactions peaking during times of crisis or events.
To fight fire with fire, the fact-checking platform Sawab was initially founded as a WhatsApp group in 2022. During the 2023 earthquake in Turkey and Syria, disinformation spiked, and it became difficult for people to navigate, which brought the number of users in the Sawab WhatsApp group from a few hundred up to 2000 within only four or five months.
During the earthquake, the number of AI-generated images circulating on WhatsApp was striking, such as the image of a dog sitting on top of the rubble under which a child's hand protrudes. Such images were likely spread to evoke sympathy or to gain traction or views. Due to the high demand for fact-checks, the Sawab team soon explored other social media channels and started their website in 2024.
Just one of many images the Sawab team has identified as fake Image: Sawab
Sawab has recently welcomed a special new member to its team of 14 part-time members: a virtual AI bot named VERIFAI that supports the team around the clock.
The Sawab team was recently put to the test in the light of the high influx of disinformation emerging during the war in Gaza and the military escalation between Israel and Lebanon.
The team members who commit to Sawab alongside full-time jobs or study commitments were simply not able to keep up with the amount of content. The VERIFAI Sawab bot now supports them on their mission of verifying news and combating hate speech.
The VERIFAI bot assists the team in research, information gathering or specific editorial tasks while the team makes sure that the bot is given very precise instructions or reliable sources to use.
However, the bot's contribution so far remains limited. "Until now, our virtual team member is not a fact-checker," explains Sawab's 22-year-old co-founder Youssef Al-Amin, a student at Doha Institute. "It is difficult to teach him (VERIFAI) how to use reliable sources. He still needs time to learn."
The AI bot has been trained in Sawab's editorial style and platform-related specificities to be able to support its human team members in the best possible way. The bot was able to render some of the workflows more time efficient, which has helped other team members to free up and dedicate more of their time to fact-checking tasks.
Youssef Al Amin, a Lebanese journalist and co-founder of Sawab, brings his fact-checking expertise to the stage at a DW Akademie conference in Tunis Image: DW Akademie
The Sawab team has noticed that the bot still has significant limitations in its performance in Arabic language or its application in the Arabic context. "We still mostly use English prompts, as there are more mistakes made in Arabic. We also noticed that there is a lack of understanding on local realities on the ground, especially when the task concerns smaller communities or villages in Lebanon," recalls Al-Amin. One major caveat is that the bot would never admit to its own limitations or knowledge gaps, but would rather provide made-up answers, but with a strong degree of confidence.
AI tools to detect misinformation such as AI generated images, also lag behind their English-language counterparts. "They are still bad at detecting AI-generated images. Every time the tools used to generate AI content improve, the more difficult our mission becomes, as the gap widens," explains Al-Amin.
Election periods make it harder for users to spot disinformation online Image: Vahid Salemi/AP/picture alliance
With municipal elections taking place in Lebanon in 2025 and parliamentary elections in 2026, the Sawab team anticipates a high rise in disinformation, as well as its spread by chatbots and on coordinated accounts, and has already started to prepare for what is yet to come.
Fact-checkers from the Middle East operate in very specific country contexts, but there is one thing that unites them all: Giving up is not an option. They all do their best to respond to each request and deliver any fact-check demanded. Despite all the hardships and difficulties they face, they continue to do their work and follow ambitious aims.
"We do whatever we can. We hope that the bot could help to reduce hate speech, incitement as well as gender-based disinformation," says Riham from Kashif.
The Middle Eastern fact-checkers try to make the region a better and safer place through verified information: One step at a time — and now assisted by machine effort through their narrative detection tool, chatbots or even virtual team members.
Jenna Kleinwort is a project manager in the Middle East and North Africa Department at DW Akademie and a member of the DW Akademie's Community of Practice on Artificial Intelligence. She holds an MSc in Economics of the Middle East from Marburg University. She wrote her thesis on Female Entrepreneurs in the United Arab Emirates, where she also worked as a reporter. Fluent in Arabic, Jenna has spent over seven years in North Africa and the Gulf region. She is also a 2023 Rosalynn Carter Fellow for Mental Health Journalism.
