Israel’s coming election is being fought not only through speeches, rallies and conventional advertising, but inside the digital systems that shape what voters see, hear and believe. Bots are being offered to manipulate online polls. Fake digital personas can infiltrate closed groups and amplify political messages. Artificial intelligence is being used to identify voters, generate campaign material and conduct personalised phone calls. Advertising technology once associated with intelligence operations can now be used to track and target voters.
The scale and sophistication of these tools have changed dramatically since Israel’s last general election in November 2022, just a month before ChatGPT was launched. Benjamin Netanyahu’s government has remained in power during that period, but the technology available to political campaigns has evolved rapidly.
A Haaretz investigation has found a broad new arsenal of campaign and influence technologies, raising questions over whether laws designed to protect electoral integrity are capable of dealing with the digital systems now being deployed. Some of the technologies are illegal, according to the investigation, while none of the tools described is subject to dedicated election oversight.
Israel’s election laws were originally drawn up in the 1950s. They have been updated to address newer threats, including deepfakes and other forms of AI-generated campaigning. Yet the latest technologies extend far beyond manipulated videos. They operate through polling platforms, social networks, messaging services, voter databases, advertising systems and increasingly AI-powered chatbots.
One of the clearest examples is the emergence of polling bots. Bots are software programmes capable of carrying out tasks normally performed by people, automatically and at scale. They have long been used in digital marketing to inflate view counts or increase the apparent reach of online content. Political campaigns have also used them to generate comments, shares and engagement that can make a party’s support appear more organic than it is.
Now, according to the Haaretz investigation, campaigns can buy fleets of bots specifically designed to locate open online polls and vote repeatedly for a preferred candidate. More sophisticated digital avatars can even possess real phone numbers, allowing them to participate in polls distributed through text messages or large WhatsApp and Telegram groups.
The potential impact is significant even though such polls have no statistical credibility. An open poll can attract hundreds of thousands of views and more than 100,000 votes, as happened recently with a poll on an Instagram news account asking whether Naftali Bennett or Gadi Eisenkot was better suited to become prime minister. Such results can subsequently be amplified across social media, creating an impression of momentum.
For parties operating close to Israel’s 3.25 percent electoral threshold, manipulated online sentiment could be particularly consequential. An apparently strong showing might reassure uncertain supporters that a party is safely positioned to enter the Knesset, even when the underlying survey has no representative statistical basis.
Israeli authorities have long recognised the influence polls can have on voters. Polls are forbidden during the final days before an election, and this month the Central Elections Committee instructed media organisations to submit information about every published poll within 24 hours. But that requirement applies to surveys employing recognised statistical methods. Open internet polls remain outside the same framework.
The investigation found that although the law requires publishers of non-statistically significant polls to make clear that their findings do not represent public opinion, this is not yet consistently being done.
The manipulation of perception extends beyond polling. Haaretz found that at least two networks of digital avatars have been established ahead of the election, serving candidates across the political spectrum. Unlike fully automated bots, avatars are generally maintained partly by humans and designed to appear indistinguishable from genuine users. They can publish and share material, comment on articles and posts, amplify campaign messages and, in some cases, identify closed groups of potential voters before entering them and directing targeted influence at their members.
Building such networks requires infrastructure. A senior figure at a company working with one campaign told Haaretz that the process could not be created overnight. Earlier investigations by Haaretz and TheMarker had already exposed Israeli software capable of creating fake accounts at scale, alongside a database of tens of thousands of false identities used in an election interference and disinformation service.
The technology has also been deployed beyond Israel. In 2024, Haaretz reported that Israel had acquired a mass-avatar influence system following the digital backlash after October 7. The system was used to troll the pages of US lawmakers before Meta detected and removed the campaign.
But political influence is not confined to fictitious users. Campaigns are also organising real supporters into what one political consultant described as volunteer armies. Thousands of supporters can receive daily instructions containing messages, images, videos and links they are expected to amplify or comment on. Artificial intelligence can identify online discussion hubs for these volunteers and generate the messages they post, transforming organised political activity into something that can appear spontaneous.
At the centre of the campaign machinery is voter data. Parties use voter registries supplied by the Interior Ministry and enrich individual records with contact details and political indicators. AI enables them to integrate larger and more diverse datasets, including information developed for the mobile advertising industry.
This is where advertising intelligence, or ADINT, enters the electoral process. Haaretz investigations have previously reported on technology that converts data collected from applications and smartphones into geolocation information. Digital advertising identifiers that are supposedly anonymous can, through de-anonymisation capabilities, be connected to telephone numbers. Technologies once used for capabilities such as tracking nuclear scientists in Iran are now available to commercial clients.
Campaigns combine this information with voter-management systems and applications such as Elector. They can identify people following a party on Facebook, joining politically aligned groups or responding to targeted surveys. Some parties buy databases, while sources acknowledge that the legality of the methods used to acquire some of the information varies.
These voter-management systems also operate on election day, when observers can report which targeted voters have already cast ballots, allowing campaign headquarters to identify those who have not yet voted. This practice came under renewed scrutiny this week when Justice Noam Sohlberg, head of the Central Elections Committee, barred observers from reporting real-time voting information to their parties, saying that doing so “constitutes a violation of privacy”. The ruling angered ultra-Orthodox parties and Likud, the main users of such systems.
Once voters have been identified, campaigns can reach them through advertising, social media, SMS, WhatsApp and increasingly sophisticated phone calls. Robocalls have existed since the 1990s, but artificial intelligence is changing their character. Industry sources describe systems capable of tailoring calls to individual recipients and responding to questions using pre-prepared answers, creating conversations that sound real.
The newest frontier is even more difficult to see. Campaigns are attempting to influence not only voters but the answers generated by AI systems themselves. The strategy, described as GEO, or generative-engine optimisation, involves creating online material that AI engines may absorb when answering questions about political candidates.
A Haaretz test found that chatbots avoided directly answering questions about which candidate was most suitable to become prime minister. But when questions became more specific, their answers became less cautious. Asked which candidate was most likely to rehabilitate Israel’s standing around the world, ChatGPT listed Yair Lapid, Bennett and Eisenkot as having potential, while omitting Avigdor Lieberman and stressing that there was no professional or public consensus that any one candidate was clearly the most suitable.
The central question is whether any of this is adequately regulated. Attorney and activist Shahar Ben Meir has argued that Israel’s regulatory framework leaves substantial gaps between privacy law and election law. He described elections as “a mass casualty event for privacy” and said political campaigns engage in activities that significantly infringe privacy, with considerable doubt over their legality.
Ben Meir argues that the absence of effective regulation is not accidental, blaming Prime Minister Netanyahu and the party he leads for preferring an environment governed by outdated laws. He also cited Ofer Golan, the former head of Likud’s campaign headquarters, who said in a 2019 interview: “We have no red lines. The goal is to win – we’ll do damage control later.”
The most consequential challenge, however, may emerge after voting ends. Tehilla Shwartz Altshuler, head of the Democracy in the Information Age programme at the Israel Democracy Institute, warned that the same technologies used to promote candidates, damage rivals and influence turnout could also generate a real-time narrative that an election had been stolen.
In that scenario, she said, the authorities would have to respond swiftly and decisively. Once public trust in the election results is lost in the crucial hours after polls close, restoring it through a legal ruling days later could prove exceptionally difficult.
The technology now surrounding Israel’s election has therefore moved well beyond traditional political campaigning. Bots, avatars, voter databases, advertising intelligence, AI-generated calls and chatbots have created an ecosystem in which the boundaries between genuine public opinion, organised persuasion and artificial influence are increasingly difficult to distinguish. The election laws may have evolved since the 1950s, but the campaign technology described by Haaretz is moving at a considerably faster pace.
