Kyanos

AI & Political Information

When a voter, donor, or journalist asks a chatbot about a candidate, the answer they get is now part of the campaign. These papers make the case for treating that answer as a channel worth measuring and managing, the same way campaigns already treat television, mail, and the doorstep.

The numbered papers are our own work: the framework for where AI answers fit in campaign strategy, what they cost when they go wrong, and what can and cannot be fixed. The research studies and commentary that follow are the outside evidence, the independent studies and reporting that show this is real, measurable, and already happening.

We publish them because the field is new and the stakes are concrete. A campaign that understands how AI describes its candidate can act on it. One that does not, cannot.

Papers

Six papers, building in sequence. They start with the scale of AI's reach into the electorate, connect AI presence to the mechanics of persuasion, apply it to the organizations that endorse candidates, then turn to the economics: where AI answers fit in a campaign budget, what that return looks like, and where remediation stops working. Each paper assumes the one before it.

📈
Kyanos · Paper I
How Many Voters Will Be Influenced by Chatbots in 2026 and 2028?
Campaign ManagersCandidatesDonorsConsultantsStrategy
A voter-influence projection for the 2026 and 2028 cycles, estimating the scale of AI chatbot contact with voters and how it grows as usage spreads across the electorate. The opening case for why AI answers belong in campaign planning at all.
🧠
Kyanos · Paper II
Legible to Both: AI Presence Optimization and the Psychology of Persuasion
Comms DirectorsCopywritersWebmastersMethodology
The qualities AI scoring rewards, entity clarity, position explicitness, factual density, coherent framing, are the same qualities persuasion research identifies as prerequisites for reaching human readers. A theoretical argument connecting AI presence work to cognitive load theory, the elaboration likelihood model, framing theory, and source credibility research.
🏢
Kyanos · Paper III
The Endorsement Economy
OrganizationsDonorsComms DirectorsStrategy
How progressive organizations can own the AI information environment. Scorecards, endorsements, and legislative ratings are exactly what AI systems are trained to trust, yet most organizations publish that data in formats AI cannot parse.
📄
Kyanos · Paper IV
The Last Mile Problem
Campaign ManagersCandidatesConsultantsROI
Where chatbot answers fit in campaign ROI. Introduces Effective Persuasion Cost (EPC), a five-dimension framework for channel effectiveness that accounts for AI-mediated verification, with a chain-of-loss model, value at risk by race tier, and monitoring budget breakdowns.
📄
Kyanos · Paper V
The Last Mile Problem: ROI
Campaign ManagersCandidatesDonorsROI
What AI says is now part of your persuasion budget. The business case for AI answer optimization as a voter contact channel alongside TV, mail, and canvassing, built on the EPC framework from Paper IV.
📄
Kyanos · Paper VI
The Limits of Remediation
Campaign ManagersComms DirectorsConsultantsStrategy
Why factual errors in AI platforms can be patched, why affective framing cannot, and what that means for campaigns and endorsing organizations. A lever map ranking remediation tractability from authoritative self-description to model retraining.
📋
Supporting Material
The Last Mile Problem: The Model
Campaign ManagersConsultantsROI
Technical addendum to Paper V. The underlying cost-per-voter model and assumptions for quantifying AI answer optimization value relative to other voter contact channels.
📊
Supporting Material · Interactive Tool
Last Mile ROI Calculator
Campaign ManagersCandidatesROI
Enter your race size, media market costs, and canvassing spend. Generates a customized estimate of where AI answer optimization fits in your persuasion budget.
Research Studies

The outside evidence. Independent studies, several from major journals and research institutions, measuring whether AI actually moves voters, how much the public trusts it, and where people now go for political information. We did not run these. We cite them because the answers are not ours to assert. External links point to the primary source.

📊
Research Study Best for Skeptics External ↗
The Levers of Political Persuasion with Conversational AI — Hackenburg et al., Science
Campaign ManagersCandidatesConsultantsEvidence
The largest persuasion study to date: 76,977 participants, 19 models, 707 political issues, 466,769 fact-checked claims. The techniques that made models more persuasive systematically made them less accurate, and the newest frontier models were among the least accurate. The strongest single proof that AI both moves opinion and does so partly with false claims.
🗳
Research Study Best for Skeptics External ↗
Persuading Voters using Human-AI Dialogues — Rand, Pennycook et al., Nature
Campaign ManagersCandidatesConsultantsEvidence
A Cornell-led causal experiment: roughly six minutes with a candidate-advocating chatbot shifted likely opposing voters about four times the average impact of traditional political ads, replicated with larger shifts in the 2025 Canadian and Polish elections. The clearest evidence that a single AI conversation measurably moves candidate preference.
📰
Research Study Best for Skeptics External ↗
Digital News Report 2025 — Reuters Institute, University of Oxford
Comms DirectorsConsultantsEvidence
The global benchmark for news behavior, surveying roughly 100,000 people across 48 markets. AI chatbots ranked last among the sources people use to verify whether something is true, behind trusted outlets and fact-checkers, even as weekly use kept climbing. The honest baseline on how little the public yet trusts AI for verification.
🏪
Research Study External ↗
How AI Is Shaping Elections — Yousefi, Tappin & Madsen, LSE
Comms DirectorsConsultantsWebmastersEvidence
A 2026 survey of 2,137 UK adults measuring where voters actually go for political information: roughly a quarter now get it from AI Overviews and a fifth from chatbots, more than double the 2024 figure, with no significant age gap. The demand-side evidence that voters already use AI for politics and that passive AI summaries reach more of them than deliberate chatbot visits.
🔎
Research Study Best for Skeptics External ↗
Google Users Click Less When an AI Summary Appears — Pew Research Center
Comms DirectorsWebmastersConsultantsEvidence
Pew's analysis of real browsing data found that when an AI summary appeared, users clicked a traditional result about half as often, while about 18% of all searches surfaced an AI Overview. Who, what, and why questions triggered a summary most of the time, the exact shape of voter queries. The proof that the last mile is the AI answer itself, read in place and rarely clicked past.
🏠
Research Study
AI's Political Architecture — Andy Hall
ConsultantsComms DirectorsEvidence
How AI systems are designed to handle political content, the structural decisions and constraints that determine what chatbots will and won't say about candidates and political organizations.
💰
Research Study
Political Markets and AI — Andy Hall
ConsultantsComms DirectorsEvidence
Andy Hall on the intersection of AI and political markets. Background reading for understanding the structural forces shaping AI's role in political information.
🧠
Research Study
Agents, Markets, and Political Knowledge — Andy Hall
ConsultantsComms DirectorsEvidence
Andy Hall on AI agents and political knowledge formation. More theoretical, useful context for understanding the mechanisms behind AI political content behavior.
🚀
Research Study
Building Political Superintelligence — Andy Hall
ConsultantsComms DirectorsEvidence
Longer-term implications of AI for political knowledge and democratic decision-making. More speculative, best for forward-looking conversations about where this goes.
Commentary

Reporting and practitioner perspective. Working journalists and political operators describing, in their own words, a problem they did not have a name for. Less rigorous than the studies above, closer to the ground.

🗳
Commentary Best for Skeptics
‘Who Should I Vote for?’ Voters Turn to A.I. Before Casting Their Ballots — Jennifer Medina, The New York Times
Campaign ManagersCandidatesConsultantsEvidence
A Times report from the 2026 primary season: named voters in Los Angeles, Atlanta, Corona, and Baltimore photographed or uploaded their ballots, asked Claude or ChatGPT who to vote for, and marked their ballots from the answers — one cut research from roughly 20 hours to one. Researchers quoted in the piece caution that answers tend to mirror the asker's own leanings and likely favor candidates who are more visible in local press and social media, and report that strategists have begun formatting online material the way chatbots prefer. The demand side of the NOTUS story: voters are not waiting for campaigns to catch up.
📰
Commentary Best for Skeptics
ChatGPT and Claude Will Be a Force in Elections. Nobody Knows What to Do About It. — Alex Roarty, NOTUS
Campaign ManagersCandidatesConsultantsStrategy
A Washington campaign reporter interviews more than a dozen strategists in both parties and finds the same gap everywhere: voters already ask chatbots who to vote for, and operatives have little idea what the bots say or how to shape it. The clearest mainstream-press confirmation of the problem, in a working reporter's words.
🏫
Commentary Best for Skeptics
AI Is A Shitty Political Advisor — Andy Hall
ConsultantsComms DirectorsEvidence
A Stanford political scientist documents how AI chatbots perform on political questions, inaccurate, inconsistent, and confident. The most pointed academic-credibility piece for audiences who need evidence before they take AI political content seriously.
📑
Commentary
You're Not Going to Be Replaced by AI. You're Going to Be Lapped. — Katie Harbath
Comms DirectorsOrganizationsConsultantsStrategy
A former Facebook global politics director on why political organizations need to engage with AI now. Accessible and non-technical, written for political practitioners, not technologists.
🌎
Commentary
The Future of Our Information Environment — Katie Harbath
Comms DirectorsOrganizationsConsultantsStrategy
A broader look at where AI fits in the political information ecosystem. Useful framing for communications staff thinking about the longer arc of AI and political media.