Advertisement
SKIP ADVERTISEMENT
You have a preview view of this article while we are checking your access. When we have confirmed access, the full article content will load.
Opinion
Supported by
SKIP ADVERTISEMENT
Guest Essay
Political Campaigns Have No Idea What’s About to Hit Them
April 28, 2026,
5:05 a.m. ET
Credit...
Chase Barnes
Listen
· 15:26 min
Share full article
By
Thomas B. Edsall
Mr. Edsall contributes weekly essays from Washington on politics and demographics.
See more of our coverage in your search results.
Encuentra más de nuestra cobertura en los resultados de búsqueda.
Add The New York Times on Google
Agrega The New York Times en Google
For better or worse, artificial intelligence is driving a major upheaval in American politics that will alter the substance and the character of campaigns.
A.I. has emerged as a powerful political tool with the potential either to improve the quality of decision-making on Election Day or to do the opposite and subvert the process of deliberation.
Perhaps surprisingly, a number of studies have shown that A.I. chatbots and large language models have stronger persuasive powers than humans.
In “
When Large Language Models Are More Persuasive Than Incentivized Humans, and Why
,” which was published last May, an international team of 40 researchers wrote:
In our first large-scale experiment, humans vs. L.L.M.s (Claude 3.5 Sonnet) interacted with other humans who were completing an online quiz for a reward, attempting to persuade them toward a given (either correct or incorrect) answer.
Claude was more persuasive than incentivized human persuaders both in truthful and deceptive contexts and it significantly increased accuracy if persuasion was truthful, but decreased it if persuasion was deceptive.
The authors suggested “that these effects may be due to L.L.M.s expressing higher conviction than humans.”
In a separate paper, “
Persuading Voters Using Human-Artificial Intelligence Dialogues
,” which was published in December, eight researchers studied the persuasive effectiveness of A.I. models through “conversations” with human participants about political candidates. They found “significant and positive treatment effects on candidate preference that are larger than typically observed from traditional video advertisements.”
One of the authors, Adam Berinsky, a political scientist at M.I.T., responded by email to my inquiries:
In preregistered experiments across the 2024 U.S. election, the 2025 Canadian election, the 2025 Polish election and a 2024 Massachusetts ballot measure on psychedelics, dialogues with frontier L.L.M.s meaningfully shifted voter preferences.
The effects ran larger than what we typically see from traditional video ads — roughly three to four points on candidate preference in the U.S., about 10 points in Canada and Poland, and 14 to 22 points on the Massachusetts measure.
Berinsky, however, pointedly put the study’s conclusions in a more cautious context:
Anxiety about A.I. in politics is often anxiety about persuasion itself, displaced onto a new technology. We went through versions of this with television, direct mail, cable news, social media and microtargeting. Each time, the predicted revolution arrived smaller than advertised, because voters are harder to move than observers assume.
Our results show A.I. can shift attitudes more than traditional ads in a controlled conversational setting, which is new. But the jump from “A.I. moves voters in a lab” to “A.I. decides the next election” still has to clear the delivery problem — and nobody has.
I asked two practitioners of the dark arts of politics for their assessment of the influence of A.I. They are not happy campers.
Thank you for your patience while we verify access. If you are in Reader mode please exit and
log into
your Times account, or
subscribe
for all of The Times.
Thank you for your patience while we verify access.
Already a subscriber?
Log in
.
Want all of The Times?
Subscribe
.
Related Content
Advertisement
SKIP ADVERTISEMENT
댓글
0