- 91% of respondents get election information from at least one social or messaging platform.
- 83% say false information poses a serious threat to South Africa’s democracy.
- 60% have encountered emotionally charged rumours designed to provoke fear or anger between groups.
- 71% believe influencers should legally disclose paid political promotion.
- South Africa’s new Electoral Code requires AI-generated political content to be labelled and false information to be corrected within 36 hours.
South African voters are relying heavily on social media and messaging platforms for election information despite having little trust in the content they encounter there, according to a new digital literacy survey ahead of the 2026 local government elections.
The Campaign on Digital Ethics (CODE) found that 91% of 771 respondents receive election-related information from at least one social or messaging platform, while only 11% expressed high trust in social media. At the same time, 83% said false information poses a serious threat to South Africa’s democracy.
The findings come as South Africa prepares to test its new Electoral Code of Conduct on Countering Disinformation during the local government elections on 4 November 2026.
The Code, which came into force on 4 September, imposes new obligations on political parties and candidates, including requirements to label AI-generated material and correct false information within 36 hours.
Social media part of election infrastructure
The report, The Algorithmic Ballot: Volume II – Democracy in the Feed, is based on a July 2026 survey of 771 people. CODE cautions that the sample is not nationally representative: respondents were predominantly urban, mobile-first and aged between 18 and 44, with Gauteng accounting for 45% of responses.
“Social media is no longer simply somewhere political conversations happen. It has become part of our electoral infrastructure,” said Kavisha Pillay, executive director of CODE.
“We are increasingly making political choices in information environments shaped by algorithms, influencers, political advertising and AI-generated content. Yet the systems determining what voters see remain largely opaque.”
Despite their low trust in social media, respondents demonstrated reasonable basic digital literacy. Eighty-three percent correctly identified the absence of an author, publication date and contact information as a warning sign that a news website may be untrustworthy, while 78% said they would consult legitimate IEC sources when confronted with an unverified WhatsApp warning about a changed voting station.
Political influencers
“One of the most important things this research tells us is that we cannot place the entire burden of protecting democracy on individual voters,” said Pillay.
“We can teach people to pause, check and verify but digital literacy cannot make an opaque advertising system transparent or tell you why an algorithm decided to amplify a particular piece of political content. We cannot fact-check our way out of a structurally broken information environment.”
But the ability to undertake more sophisticated verification was weaker. Reverse image searching was the clearest gap: 81% said they could not perform the task, despite 81% expressing confidence that they could recognise AI-generated media. The report warns that confidence does not necessarily translate into the ability to identify increasingly sophisticated manipulation.
The survey also points to the growing role of influencers and creators in political communication. Sixty percent had seen an influencer or content creator promote a political party or candidate, while 63% had encountered political promotion on social media that they suspected had not been clearly disclosed as advertising. Seventy-one percent believed influencers should be legally required to disclose paid political promotion.
The regulatory angle
The report says the new Disinformation Code is an important step, but its reach is narrower than the information environment it seeks to regulate.
It applies primarily to political parties and candidates, including people acting on their instruction, authority or for reward. It covers online and offline communication, paid and unpaid content. Parties and candidates must take reasonable steps to ensure information is checked before publication and must correct false or misleading information within 36 hours of becoming aware of it.
Political advertising must identify the relevant party or candidate, while AI-generated material must be labelled as synthetic content.
“The new Code is an important step forward, but this election will tell us whether it has teeth,” Pillay said. “A correction that nobody sees is not much of a correction. And we cannot meaningfully regulate digital campaigning while leaving the machinery of amplification outside the accountability framework.”
No direct obligations
However, the report argues that platforms including Meta, TikTok, X, YouTube and messaging services have no corresponding direct obligations under the Code. They are not required to identify, counter or slow the spread of electoral disinformation, maintain public political-ad repositories, disclose targeting information or provide researchers and regulators with data about content distribution.
This creates what CODE identifies as a major gap: political actors can be held responsible for publishing false claims, but the platforms’ recommendation systems — which determine how widely those claims travel — remain largely outside the Code.
The report says 71% of respondents support mandatory disclosure of paid political promotion by influencers, while 44% want clearer labels on political advertising and AI-generated material. It also argues that future regulation should require anyone receiving payment or another material benefit to promote political content to disclose who funded it, even where the funder is not a registered political party.
A choice
South Africans want greater transparency as political influence moves into influencer marketing and algorithmic feeds, while the new electoral disinformation rules place most of the responsibility on political actors rather than the platforms distributing their messages.
“A free vote is not only a ballot cast without intimidation,” Pillay said. “It is a choice made with access to information you can trust and without hidden manipulation determining what you see. If we care about protecting the vote, we have to protect the information environment in which that vote is formed.”
The report’s recommendation that platforms maintain accessible political advertising repositories, disclose who paid for advertisements and provide meaningful information about audience targeting is particularly relevant to advertisers, agencies, publishers and platforms.
