AI in elections and public speech: 22 records so far

Every measured gain and problem on the record for this topic, newest first.Too few to weigh yet. Updated November 5, 2026.

The count22 counted · 8 context

One square per record, darker for stronger evidence:Evidence: 1 high10 moderate11 emerging

22 records so far, too few to weigh yet. Is your study missing? Add it

The records

AI in elections and public speech

22 records so far

The records

AI in elections and public speech: 22 records so far · Truvace

The count

Gains: 2
Traces: 2
Problems: 18

A Trace is one record with both a gain and a problem, so it counts on both sides: 4 records show a gain and 20 show a problem. 8 context records are in the record list but not counted, because they report no measured gain or problem.

How strong the evidence is

Evidence	Gain side	Problem side
HighStrong study design, backed by at least two independent sources	1	0
ModerateA measured study or an official source	1	9
EmergingEarly, small, unreviewed or reported second-hand	2	11

By type of evidence: 8 news reports; 4 investigations by reporters; 4 surveys; 2 interview studies; 1 controlled comparison; 1 benchmark or lab test; 1 commentary or narrative review; 1 observational study. How evidence is scored

22 records so far, too few to weigh yet. A written summary appears once a topic has at least 20 counted records, with at least 5 on each side. Is your study missing? Add it

Strongest gains

Kazakh social media users who feel AI manipulates them report far lower trust in politics
In a survey of 1,795 Kazakh social media users, those who perceived algorithmic personalization sometimes reported a greater sense of control over their information choices.
Survey · Evidence: High · 77/100 · TRV-2026-0179
AI chatbot fact-checks cut trust in two false health claims more than web search did
In two experiments (280 and 232 people), those who checked a false claim about blood donation or hospice care with generative AI lost more trust in it than those who used a search engine, though effects were short-term.
Controlled or validated comparison · Evidence: Moderate · 68/100 · TRV-2026-2265
AI sped up voting admin in four Southeast Asian nations, but bias and data risks went unresolved
A comparison of Indonesia, Thailand, the Philippines and Myanmar found AI voter identification, registration and result monitoring streamlined election administration and improved voter-list accuracy.
Interviews or qualitative study · Evidence: Emerging · 45/100 · TRV-2026-0623

Strongest problems

Small German parties used more AI images than big ones, but got no extra engagement edge
A study of the 2025 German election found AI-generated posts drew more engagement than non-AI posts, but the boost was equal for major and minor parties, so minor parties gained no competitive advantage.
Observational study · Evidence: Moderate · 63/100 · TRV-2026-0109
AI image labels cut belief in false claims, but raised it for ones with human-made images
The same survey found people over-relied on labels: they believed false claims with human-made images more, and believed true claims with labeled AI images less.
Survey · Evidence: Moderate · 61/100 · TRV-2026-0198
How American political campaigns are using AI – and what they’re spending on the tools
Disclosure rules miss spending by consultants, media firms and independent committees, so the bulk of campaign AI use is likely hidden from public view.
Investigative reporting · Evidence: Moderate · 55/100 · TRV-2026-1531