Artificial intelligence is no longer just for writing texts, summarizing documents, or preparing presentations. It is also beginning to enter one of democracy's most delicate spaces: the voting decision.
According to The New York Times, several voters in the United States are using tools like Claude or ChatGPT to inform themselves before voting. They upload a photo of the ballot, ask them to compare candidates, summarize platforms, or explain which option best aligns with their political values.
The scene seems futuristic, but it's not so much. A voter who doesn't know the candidates in a local election can ask AI to explain who's who. Another can ask which candidate best matches their priorities. Another can ask for a strategic recommendation to prevent a specific candidate from winning.
AI, in theory, shouldn't say "vote for this one". But it can do something almost equally influential: organize information, prioritize options, explain risks, and help the user reach a conclusion.
From the electoral guide to the chatbot
For years, many voters have turned to media outlets, electoral guides, platform comparators, recommendations from parties, unions, associations, or politicized family members. The novelty is that they can now substitute part of that process with a conversation with a machine.
The appeal is evident. A ballot can be long, confusing, or full of unfamiliar names. Local elections often include positions, lists, measures, and candidacies about which many citizens have little information. AI offers a quick solution: it summarizes, compares, and translates political noise into seemingly clear answers.
Therein lies the power of the phenomenon. A chatbot doesn't need to engage in explicit propaganda to influence. It's enough for it to decide what information is relevant, what nuances it includes, which sources it prioritizes, and how it presents each candidate.
Why it could happen in Spain
Spain is not the United States, but the ground exists. In municipal elections, for example, many citizens vote with limited information about local candidacies, pacts, platforms, lists, or secondary candidates. In regional elections, the level of knowledge also varies greatly depending on the territory. Even in general elections, many voters may have doubts about strategic voting, blocs, constituencies, or the real impact of each ballot.
In this context, it would not be strange for a citizen to ask ChatGPT: “I am progressive, I live in Madrid and I want to vote strategically, what should I take into account?”. Or for another to pose: “I am liberal in economics and conservative in immigration, which party best suits me?”. Or for someone to upload electoral programs and ask for a quick comparison.
AI could become a kind of private political advisor, available in seconds and at no direct cost. It would not replace parties or the media, but it could mediate between them and the voter.
The risk: seemingly neutral answers
The big problem is that artificial intelligence responses usually sound orderly, reasonable, and neutral, even when they may contain errors or biases. This appearance of balance can make the user trust it more than they should.
AI models do not understand politics like an informed citizen. They process patterns, available sources, and user formulations. If someone asks from a specific ideological position, the response can adapt to that view and reinforce their previous preferences.
In other words, AI may not change a person's vote, but it can give them arguments to feel more confident about what they already thought. It can also present an option as "strategic" that depends on polls, electoral rules, or contexts that may not be updated or well interpreted.
A useful, but not innocent, tool
Using AI for information does not have to be negative. It can help understand programs, compare proposals, or translate complex political language. It can be useful for citizens who do not have time to read dozens of documents or follow all electoral news.
The problem arises when too much is delegated. If the voter asks for a quick answer and does not check the sources, AI can become an opaque electoral guide. It is not always known what information it has used, what it has left out, or why it has ordered the candidates in a certain way.
Furthermore, it can favor those who have a greater digital presence. A party with a lot of optimized content, well-structured programs, and easy-to-track messages may appear better represented than a small, local, or less resourced candidacy.
Parties will also learn to talk to AI
If voters start asking artificial intelligence, parties will try to appear better in their answers. The campaign will no longer be directed only at journalists, social networks, or voters, but also at the systems that summarize political information.
That can change the way electoral programs are published. More documents with bullets, more comparisons, more pages designed to be read by AI models, more simple messages, and more content designed for a chatbot to understand and reproduce.
The battle for political positioning could increasingly resemble a battle to appear well explained in an automatic response.
Spain and the useful vote: an especially sensitive area
In Spain, the use of AI could have a particular effect on the debate about the useful vote. The electoral system, with provincial constituencies and distribution through the d’Hondt method, makes the value of each vote highly dependent on the territory.
A user could ask which party has the best chance of obtaining a seat in their province or which vote can prevent a specific bloc from winning. That information, if well calculated, can help understand the system. But if misinterpreted, it can lead to errors.
In close elections, an apparently technical recommendation can have political consequences. Explaining scenarios is not the same as pushing the voter towards a specific option.
The democratic question
The phenomenon opens a fundamental question: do we want artificial intelligence to be a tool for electoral information or an intermediary in political decision-making?
The difference is important. It is one thing to ask an AI to summarize the programs of several parties. It is another to ask it which ballot paper is advisable to put in the ballot box.
The former can improve access to information. The latter can shift part of the citizen's political judgment towards a tool that does not vote, is not accountable to anyone, and does not always explain how it reached a conclusion.
AI does not vote, but it can influence the vote
Artificial intelligence does not decide an election. But it can begin to influence how voters understand candidates, compare programs, and resolve doubts before voting. In the United States, the phenomenon already appears in local elections and primaries. In Spain, it could easily arrive in any campaign marked by multiple candidacies, close blocs, and undecided voters.
The question is not whether citizens will use AI to inform themselves politically. They probably already are. The question is whether they will know how to distinguish between help in understanding information and a recommendation that could influence their vote without appearing to be a recommendation.
Because the next electoral advisor for many voters may not be a journalist, a family member, or a militant. It could be a chat window.