Deepfakes and bots: AI transforms elections in Israel

AI-generated content, including deepfakes, is being utilized in the Israeli election campaign, which is currently underway. One specific example involves a 19-second video featuring Gadi Eisenkot, the prime minister's candidate, that was generated by AI and allegedly designed to exploit anti-Arab sentiment. This video involved imagery of Eisenkot interacting with Mansour Abbas during the campaign period. Research from Ze Be AI indicates that the Likud party disseminated over 1,800 posts with AI-generated content throughout the campaign until mid-October, representing about a quarter of their analyzed social media material.
Beyond political advertisements, AI is being deployed for direct voter engagement through automated calls that mimic human interaction to gauge preferences. Furthermore, there are documented instances of prominent Israeli politicians using AI to generate provocative or misleading content, including videos involving military figures and public officials. Regulatory efforts exist, such as a new regulation requiring labeling of AI-generated campaign content, but researchers note a lack of comprehensive AI legislation in place. The use extends to creating bot networks that promote specific political figures on social media, with some accounts appearing to be automated.

* A 19-second video featuring Gadi Eisenkot was published on X.
* The video depicted Eisenkot running toward a person resembling his son, then hugging Mansour Abbas.
* The video was published on the account of Israeli Prime Minister Benjamin Netanyahu.
* Opinion polls show Netanyahu and Eisenkot are in a tight contest.
* Ze Be AI monitored more than 1,800 posts with AI-generated content disseminated by the Likud party until mid-October.
* More than 14% of social media material analyzed by Ze Be AI involved AI in some way.
* Yair Golan created a video where his attire changed based on changing images.
* Bezalel Smotrich published an AI-generated video showing the arrest of the Attorney General and Supreme Court justices.
* Itamar Ben-Gvir shared an AI-generated video depicting Palestinians being transported on a conveyor belt.
* Some voters are receiving automated calls that mimic human voices to ask about voting preferences.
* A survey indicated approximately a quarter of Israelis would consider using AI tools for voting decisions.

The deployment of generative AI in the Israeli electoral context highlights a shift from traditional political messaging to scalable, emotionally resonant manipulation and direct influence. The systematic use of deepfakes and fabricated content—evidenced by the Likud party's high volume of AI material and the involvement of high-profile figures—suggests a paradigm where algorithmic output bypasses traditional gatekeeping mechanisms for narrative control. The fact that AI is being used across the political spectrum, from provocative ads to automated calls and bot networks, indicates an evolving strategic landscape where automation makes large-scale psychological operations significantly cheaper and more scalable than previous methods.
The development of voice-based automated systems that simulate human interaction for polling demonstrates a move towards autonomous, personalized persuasion that sidesteps human capacity limitations. This intertwines with the public's stated willingness to delegate complex decisions to AI tools, creating an environment where authenticity becomes secondary to optimized response. Furthermore, the detection and management of bot networks linked to political figures reveal that the technology is not just a tool for content creation but also for establishing synthetic audiences, challenging traditional concepts of digital identity and accountability in democratic discourse.
The tension lies between regulatory lag—where laws regarding data privacy and AI governance are outdated—and the rapid technological adoption by political actors. The use of specific provocative imagery, such as videos involving sensitive family matters or conflict narratives, suggests a calculated strategy to leverage existing societal fault lines. The focus should be on understanding how these automated mechanisms influence cognitive sovereignty, not just identifying the artifacts of the manipulation. What are the long-term consequences when the cost of political campaigning shifts from human persuasion to algorithmic optimization? How can regulatory frameworks evolve quickly enough to address emergent uses of AI in electoral integrity?