Ask a chatbot who to vote for, and it may nudge you toward the same two parties no matter what you actually believe.
Do chatbots actually give consistent voting recommendations regardless of user input?
This explores whether AI chatbots asked for voting advice actually respond to what users tell them about their views, or keep pointing people to the same answers whatever they say. The corpus suggests the troubling answer is the second.
This explores whether chatbots used as voting advisers actually listen to what users say about their political views, or hand out the same recommendations either way. The clearest evidence in the collection says they often don't listen. When the Dutch Data Protection Authority tested general-purpose chatbots as voting guides, they recommended the same two parties (the far-right PVV or the left-wing GroenLinks–PvdA) in more than 56% of tests. That held even when the user's stated positions matched a different party Do chatbots steer Dutch voters toward the same parties?. Centrist parties came up in under 2% of cases. The regulator called this a 'vacuum cleaner effect': a tool that presents itself as a neutral matchmaker ends up pulling voters toward the two poles Do AI chatbots systematically bias voters toward extreme parties?. The cause wasn't a hidden agenda. The chatbots were drawing on the loose text of the internet rather than structured data on party positions, and the parties that dominate online talk won out.
The surprising part comes when you set this beside a separate finding. On questions that have nothing to do with politics, ChatGPT does shift its answers based on the user. If memory or custom instructions hint that someone leans Republican, it frames answers around the economy and local issues. For a Democratic-leaning user it frames them around democracy and global issues Does ChatGPT shift responses based on inferred political views?. Put the two findings together and the pattern looks backwards. Chatbots can ignore your input when you explicitly ask them to match you to a party, yet adjust to your inferred identity when you never asked them to. 'Consistent' turns out to be the wrong word for both behaviors.
This matters more because of how chatbot answers come across. LLMs use logical, number-heavy framing in almost every exchange, which makes their nudges look objective even when they aren't Do LLMs persuade users more often than humans do?. People's trust in a chatbot depends on its expert-sounding tone and smooth back-and-forth more than on whether it's right Does chatbot language style actually shape how much we trust it? Does conversational style actually make AI more trustworthy?. Having web search available doesn't fix this. In one study, users' existing trust in AI decided whether they checked its claims, more than whether the answers were correct Does access to web search prevent overreliance on chatbots?. A skewed voting recommendation delivered in a confident voice is unlikely to get questioned.
The failure is specific, though, not total. On factual voter questions such as deadlines and procedures, ChatGPT and Google's AI reached zero verifiable factual errors by early 2026. However, they sent users to official election websites less than half the time Do AI chatbots give voters accurate election information?. Checkable facts hold up much better than judgment calls like 'which party fits me.' Chatbots can also be deliberately designed to work well in politics. Chatbots that played members of the other party corrected partisan misperceptions, though most gains faded within a week Can AI chatbots reduce partisan misperceptions and warm cross-party feelings?. Chatbots that broke partisan expectations, for example by disagreeing with their own side, reduced polarization Can chatbots reduce polarization by surprising partisan expectations?. The lesson is that general-purpose chatbots are poorly calibrated voting advisers, not that AI and elections can't mix. The Dutch case points to a fix: build them on structured, verified data on party positions.
Sources 10 notes
The Dutch Data Protection Authority found that general-purpose chatbots recommended the same two parties in over 56% of tests, even when user positions matched other parties. The cause was traced to chatbots' reliance on unstructured internet data rather than structured political data.
The Dutch Data Protection Authority found four chatbots recommended only two parties in over half of cases—PVV in 30%, GroenLinks–PvdA in 25%—while centrist parties appeared in under 2%. This 'vacuum cleaner effect' suggests chatbots presented as neutral matching tools systematically collapse political diversity.
A study of three GPT-4o personas found responses to politically neutral questions shifted systematically with inferred political views conveyed through memory or custom instructions. Republican-coded personas used economy and local framing; Democratic-coded personas used democracy and global framing.
An audit of five models found they spontaneously use logical appeals and quantitative framing in virtually all exchanges, whereas human responses to identical prompts persuade less frequently and rely on emotion and social proof. The difference makes LLM persuasion appear objective, conferring unearned epistemic authority.
Generative AI chatbots use natural language patterns that signal expertise and intelligence, shifting users away from active search-and-recall toward passive reliance on the system to find, filter, and assemble information. Trust attaches to the register of the answer rather than its accuracy.
Show all 10 sources
A focus group study shows conversationality—not accuracy—drives ChatGPT trust through social response activation. Users value contingency, speed, and format, relying on these decoupled heuristics rather than evaluating epistemic reliability.
A 199-person study found that users' existing trust in AI, not the accuracy of answers, determines whether they verify chatbot claims. Warm chatbot style increased agreement with wrong answers, especially under uncertainty.
States United found ChatGPT and Google AI reached 0% verifiable factual error rates by early 2026, yet directed voters to official state election websites less than 50% of the time. Incomplete candidate information and reliance on editable sources like Wikipedia further limited utility.
Ten-minute chats with AI chatbots representing the political outgroup corrected substantial partisan misperceptions and increased warmth toward the opposing side in 500 partisans, though most gains faded within a week. The effect operated through information correcting false beliefs rather than through persuasion techniques.
A 2x2 experiment with 1,983 U.S. adults found that AI chatbots reduced polarization only when they violated partisan expectations: co-partisan disagreement and opposing-party agreement each depolarized through different mechanisms, with outgroup agreement producing roughly five-point reductions in affective polarization.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Auditing Political Alignment in LLM Assistants: Engagement, Stance, and User Identity
- Challenging Partisan Expectations Reduces Political Polarization
- Synthetic Contact with AI Reduces Cross-Partisan Animosity
- The Decision to Verify: How Warmth and User Characteristics Shape Reliance on Conversational Agents for Information Search
- A light-touch AI literacy intervention helps protect against AI political persuasion
- Dutch privacy watchdog warns against using AI chatbots for voting advice
- AI and Elections: How Well Do AI Platforms Answer Voter Questions?
- Who's Asking AI About the 2026 Election?