As voters increasingly turn to artificial intelligence for political information, the battle for municipal power could extend beyond posters, rallies and social media feeds into the answers generated by chatbots.

By Gcina Shandu | October 2026

When South Africans head to the polls on 4 November 2026, they will be deciding who should govern their municipalities, maintain local infrastructure and deliver essential services. But an emerging development in political communication raises another question: what happens when voters ask artificial intelligence which candidate is best equipped to run their city?

The question is particularly relevant in Johannesburg, where Democratic Alliance mayoral candidate Helen Zille and ActionSA leader Herman Mashaba are prominent figures in the city's political debate. Their political records, public statements and policy proposals are already part of the information environment from which voters and potentially AI systems form impressions of municipal leadership.

A new frontier in political campaigning may therefore be emerging: competing not only for the attention of voters, but also for how artificial intelligence describes, compares and evaluates candidates.

From search engine optimisation to AI visibility

A Reuters report published on 26 October, titled “US political candidates tailor messages for AI models used by voters”, highlights how some American political campaigns are exploring ways to influence the responses produced by AI assistants such as ChatGPT, Claude and Gemini.

The approach is known as generative engine optimisation (GEO). Whereas traditional search engine optimisation seeks to improve a website's visibility in search results, GEO seeks to make information about a candidate, organisation or policy more discoverable and usable by AI systems.

The techniques described in the Reuters report include publishing detailed policy pages, securing coverage on authoritative websites, creating online content and monitoring how chatbots answer questions about candidates.

The potential political significance is considerable. A voter who previously searched for a candidate's manifesto may now ask an AI assistant: Who can fix Johannesburg's water problems? Which candidate has a better plan for municipal finances? Who has the strongest record on service delivery?

Instead of navigating dozens of sources, the voter may receive a single answer summarising the available information. The sources selected, the claims emphasised and the omissions in that answer could influence how the voter understands the choices before them.

Johannesburg: Zille, Mashaba and the contest over political narratives

Johannesburg offers a useful case study because its municipal challenges are complex, its political competition is intense, and its leadership history provides voters with competing records to examine.

Helen Zille, the DA's mayoral candidate, has made municipal functionality, financial recovery, infrastructure maintenance, corruption and public safety central to her campaign messaging. In an October address outlining the party's priorities, she emphasised the need to repair municipal finances, improve administrative competence and restore essential services. Source: Mail & Guardian, 17 October 2026.

Herman Mashaba, meanwhile, brings a different political narrative. As a former Johannesburg mayor and leader of ActionSA, he is associated with the city's previous period of DA-led coalition government and with ActionSA's current appeal for voters seeking an alternative to established parties.

These contrasting political identities create an interesting test for AI-generated political information. If a voter asks which candidate can improve Johannesburg, what evidence should an AI system use to answer?

It should not simply repeat the most frequently published campaign slogans. It should distinguish between a candidate's promises, their record in public office, the performance of the administrations they led, and the practical constraints facing any future municipal government.

For Zille, that means examining the feasibility of her proposals and the evidence supporting her claims about municipal recovery. For Mashaba, it means assessing his previous administration's record alongside ActionSA's current commitments. The same standard should apply to every competing candidate.

Neither candidate's visibility in AI responses would, by itself, demonstrate superior governing ability. Nor is there evidence presented here that either has adopted the AI-influence strategy described in the American report. They illustrate how the phenomenon could become relevant to South African municipal politics.

Why local government is especially vulnerable to AI-driven narratives

National political debates often revolve around broad questions of ideology, economic policy and constitutional leadership. Municipal elections are more immediate. Residents want functioning traffic lights, reliable water supplies, electricity, refuse collection, maintained roads, safe public spaces and municipal accounts they can understand.

These are also precisely the kinds of questions voters can put to AI assistants.A resident might ask, “Which party has the most credible plan to fix Johannesburg's billing system?” Another might ask, “Who is responsible for the city's water shortages?” A business owner could ask which candidate is most likely to improve permit processing and reduce administrative delays.

The answers depend on the information available to the AI system, how current that information is, and whether the system can distinguish reliable evidence from political claims.This creates an incentive for campaigns to publish accessible, well-structured and evidence-based information about their candidates and policies. A detailed municipal programme explaining proposed interventions, budgets, implementation timelines and measurable targets could be more useful to both voters and AI systems than a slogan alone.

However, the same incentive could encourage campaigns to flood the internet with repetitive promotional material, publish misleading claims or attempt to make one candidate appear more credible than the evidence warrants.

South Africa's Electoral Commission has already warned about the risks posed by generative AI and hyper-local disinformation in the 2026 local government elections. Source: SAnews, 18 February 2026.

The danger is not limited to fabricated videos or fake campaign posters. It also includes subtler distortions: a chatbot repeating an unverified allegation, overlooking a candidate's relevant record, confusing an aspirational promise with an achieved result, or presenting an incomplete comparison as an objective assessment.

Can political campaigns optimise for AI without misleading voters?

There is a legitimate distinction between making accurate political information easier to find and attempting to manipulate the answers voters receive.A campaign that publishes its full manifesto, provides verifiable data about its record and explains how proposed municipal reforms would work is improving public access to information. It may also improve the likelihood that AI systems can accurately summarise its policies.

A campaign that manufactures independent-looking endorsements, purchases deceptive coverage or deliberately spreads unsubstantiated claims to influence chatbot responses crosses a different line.For South African political parties, responsible AI visibility should involve at least four practices.

First, publish authoritative primary information. Manifestos, policy documents, budget proposals and performance reports should be accessible from official party and candidate websites.

Second, make promises measurable. Instead of promising to “fix the city”, candidates should identify the services they intend to improve, the resources required, the proposed timelines and the indicators by which residents can assess progress.

Third, maintain transparency about evidence. Campaigns should distinguish verified achievements from political interpretations and future commitments.

Finally, monitor AI-generated answers without treating them as a popularity contest. If a chatbot misrepresents a candidate's programme, the appropriate response is to correct the public record with reliable information not to manufacture content designed to overwhelm competing perspectives.

The democratic question: who shapes what voters believe?

The emergence of AI-assisted political research raises a wider question about the relationship between political communication and democratic accountability.

If voters increasingly rely on AI assistants to compare candidates, these systems could become influential intermediaries between citizens and the people seeking public office. Their answers may shape which issues receive attention, which achievements are remembered and which candidates are perceived as credible.

Yet AI systems are not electoral authorities. They do not independently establish whether a municipality has improved, whether a campaign promise is affordable or whether a politician's claims are accurate. Their answers can reflect gaps, biases and inconsistencies in the information they process.

That places responsibilities on three groups: political parties must provide trustworthy information; AI developers must improve accuracy, transparency and source attribution; and voters must continue to question the evidence behind political claims.

Independent journalism and fact-checking remain essential. Resources such as Africa Check's 2026 local election coverage provide a useful way to compare parties' promises and assess whether claims can be substantiated.

Beyond the 2026 elections

The immediate question is whether AI visibility will become a deliberate feature of South African local election campaigns. The longer-term question is how political competition will change when candidates must communicate simultaneously with citizens, search engines, social media algorithms and AI assistants.

For Johannesburg, the contest between political figures such as Helen Zille and Herman Mashaba demonstrates why this matters. Voters need more than polished campaign narratives: they need credible evidence about municipal performance, realistic policy proposals and clear explanations of what different administrations can actually deliver.

The future of political campaigning may involve making candidates more visible to AI systems. But visibility must not become a substitute for accountability.

Ultimately, the most important question is not which candidate an AI assistant recommends. It is whether the evidence available to voters enables them to make an informed decision about who should govern their city.

Because in local government, the quality of a political answer matters far less than the quality of the services residents receive after the election.