It’s going to be a big moment for South Africa. On 4 November 2026, the country heads to the polls for local government elections, and if past elections are anything to go by, the run-up will be messy in more ways than one. Alongside the usual political campaigning, policy debates, and voter education, there will be a flood of disinformation — some of it silly, some of it dangerous, and all of it designed to confuse, divide, or manipulate. The Independent Electoral Commission (IEC) has already flagged this. In fact, the IEC’s chairperson, Mosotho Moepya, has warned that generative AI is likely to supercharge “ward-specific deception” — meaning fake content tailored not just to South Africans in general, but to your specific suburb, your language, your party loyalties, and even your frustrations with service delivery. This isn’t a vague worry. It’s a prediction grounded in what we’ve seen before, and it’s one that every voter should take seriously. The goal of these falsehoods is to make you doubt the system, question the integrity of the election, and share something emotional before you’ve had time to think. But here’s the good news: a lot of this disinformation follows predictable patterns. Once you know how to spot the tricks, you’re far less likely to fall for them — and you can help stop them in their tracks.
Let’s start with the most common type: false voting instructions. These are the posts that sound incredibly helpful, often arriving as forwarded WhatsApp messages or well-meaning social media alerts. They tell you to bring your own black pen because the IEC’s ink will “evaporate.” They warn you that your voting station has moved, that you need to click a link to check your name, or that there’s a new rule you need to follow — usually something that sounds official but is completely made up. We saw this during the 2024 national and provincial elections, when a message spread like wildfire claiming that voters should bring their own pens because the ink provided by the IEC would magically disappear. The IEC had to step in and dismiss it. The tactic works because it uses urgency and practical detail. It feels like someone is looking out for you. But real voting instructions are never that mysterious. They come from the IEC website, verified party accounts, or reputable news outlets linking directly to official information. If you get a message that tells you to act fast, check your details on a strange link, or break a “new rule” that no official source has mentioned, slow down. That’s exactly what the people behind these messages want you not to do.
Then there are the recycled videos. Long before AI made it easy to generate completely fake footage, the oldest trick in the disinformation playbook was taking real video and changing the story around it. A protest, a police operation, an arrest, a violent clash, a crowd, even a pile of ballot papers — none of that has to be fake. The video itself might be 100% authentic. The lie comes from the caption, the date, or the location attached to it. A clip from a 2020 protest in Upington can be recirculated in 2026 as “evidence” of current attacks on foreign nationals. A video from another country’s election can be passed off as chaos at a South African voting station. This happened during Ghana’s 2024 election, when a video of men in buses was shared with the claim that they were “thugs from Burkina Faso” sent to disrupt the vote. Police investigated and found the men had been recruited locally to provide security at polling stations. Another video claimed ballot papers and a voters’ register had been found in a vehicle as part of a rigging plot — but fact-checkers found no evidence the booklets were even ballot papers. The lesson? If a video makes a dramatic claim, don’t take it at face value. Look at the road signs, the number plates, the shopfronts, the uniforms, the weather, the accents. Do a reverse image search on a screenshot. If the clip is old or from somewhere else, you’ll usually find it pretty quickly.
Now we get to the truly modern problem: AI-generated content. This is where things get really tricky, because a video of a celebrity, politician, or even a stranger can be manufactured out of thin air. In South Africa’s 2024 election, a deepfake showed rapper Eminem appearing to endorse the EFF. Another fake video showed former US President Donald Trump appearing to endorse the MK Party. Neither was real, but they looked convincing enough to make people pause. The same thing has happened elsewhere. In India’s 2024 election, deepfakes showed Bollywood actors criticising Narendra Modi and urging support for the opposition Congress party. In Ghana, fact-checkers identified AI-generated audio clips presented as private conversations involving presidential candidates. These fakes borrow authority from people we recognise. A celebrity, a foreign leader, a party official, or a candidate can be made to say almost anything — endorsing a party, attacking an opponent, or confessing to something damaging — without ever having spoken a word. The checks are actually pretty simple, even if they take a bit of effort. Trace the clip back to the original account. Look for the full recording. See whether the person shown has commented on it. Check whether credible reporters or fact-checkers have verified it. If all you have is a short clip with no source, no full version, and a few visual glitches, that’s weak evidence — no matter how convincing it looks on your screen.
Let’s not forget the xenophobic angle. Election disinformation in South Africa often leans heavily on anti-migrant sentiment, and 2026 will almost certainly be no different. Expect claims about migrants voting illegally, foreign nationals standing as candidates, fake IDs, corrupt Home Affairs officials, and political parties “importing” voters from other countries. We’ve seen this before. Ahead of the 2024 election, a post that’s still online falsely claimed the EFF was making fake identity documents for migrants so they could vote. The images were actually from an earlier police raid on an unrelated forgery operation. In 2026, Africa Check found false Facebook posts claiming an “illegal immigrant” was contesting a by-election in Gqeberha. The IEC has also warned about fake posters falsely claiming foreign nationals were standing as local election candidates. These stories are dangerous because they feed on real fears and real frustrations. They also distract from the actual issues. And then there are the fake results. Before the IEC has declared any final ward or council results, you’re likely to see “leaked” tallies, screenshots of results dashboards, and claims that ballot boxes or votes have disappeared. Some of these are wholly invented. Others take a real early figure and present a later update as “proof” that votes were added or removed. During counting, results can shift as more voting districts come in — that’s not fraud, that’s just how elections work. After the 2024 national and provincial elections, one of the biggest claims was that 9.3 million votes were missing. The IEC’s chief electoral officer, Sy Mamabolo, called it “patently false” in court papers. Remember: early results are provisional. The only result that truly counts is the one declared by the electoral authority.
Finally, there are two newer twists that are worth understanding. The first is when real footage or audio is dismissed as AI. This is the flip side of the deepfake problem. A genuine recording surfaces, and the people it implicates immediately claim it’s fake. Indian politics has already shown how messy this can get. In one case, a politician denied that a leaked audio recording was real, saying it was AI-generated. Independent experts later found that one clip may have been tampered with, while another appeared authentic. The answer to this is the same as everything else: ask where the material came from, whether the full version exists, who captured it, and whether credible reporters or forensic analysts have checked it. The second twist is the use of AI chatbots as “proof.” You’re going to see screenshots of Grok, ChatGPT, or Gemini answers being used as evidence for all sorts of election claims. It looks convincing, especially on X, where users can ask Grok to comment on a post. But a chatbot answer is not a fact-check. It depends on the prompt, the sources it can access, and the way the question is framed. A user can often push a chatbot toward the answer they want, then screenshot it as if it were independent verification. Grok has already produced false election information in the United States, wrongly claiming that ballot deadlines had passed after Joe Biden withdrew from the presidential race. Google’s Gemini caused an uproar in India after a post revealed how the AI chatbot said some experts had characterised Narendra Modi as fascist. Treat AI answers as a lead, not proof. If the only evidence for a claim is a screenshot of a chatbot agreeing with someone, that claim has not been verified.
So what can you do about all of this? The most important thing is to slow down. Before you share anything, take a breath. Find the original source. Check the date and place. Look for the full context. Compare it with official information from the IEC, named officials, reputable newsrooms, or established fact-checkers. If something makes you angry or excited, that’s a red flag — disinformation thrives on emotion. And if you see a suspicious election claim, don’t help it travel further. Screenshot it, save the link, note where you saw it, and send it to News24’s Disinformation Desk at [email protected]. That’s what they’re there for. Elections are supposed to be about hope, choices, and the power of your voice. Don’t let a pile of lies take that away from you. Stay curious, stay skeptical, and remember: facts matter. This article was produced by the News24 Fact Check Desk and supported by Truth First. If there’s something you’d like us to check out, debunk, or uncover, send an email to [email protected].