FUD & Political Labeling: More Research about FUD and Labeling Effectiveness
Based upon the CivilTalk’s daily FUD and Labeling Analysis from September 12–18, 2026
Prepared for Rick Fiorito by Claude • September 18, 2026
Bottom Line
The conclusion from the research literature hasn't changed with a new batch of statements: fear-and-doubt rhetoric is a high-variance, low-persuasion, base-consolidating tool. It rarely converts opposing or undecided voters, it carries a real chance of backfiring on the speaker, and its clearest effect is on the speaker's own side — energizing supporters while, in the case of election- and institution-legitimacy claims, measurably eroding their trust in the system. This round's reports add a new wrinkle: AI regulation has emerged as a third major FUD lane (alongside election integrity and character attacks), used by both parties to build urgency without citing evidence — which the research below treats the same way it treats any other unsubstantiated fear appeal.
This Batch, in Brief
The daily reports provided this round span statements dated September 4–17, 2026 (compiled across eight overlapping report-runs made between September 12–18). Coverage clusters around three storylines:
The tail end of the GOP Midterm Convention (Sept 9–11, Dallas) and the 9/11 25th-anniversary ceremonies — Vance, Trump, Mike Johnson, Abbott, Paxton, and Hegseth all produced flagged statements in this window.
A fast-growing AI-regulation fight — Trump, Cox, Kennedy, Vance, Mace, Daines, Pritzker, and Jeffries all made unsubstantiated fear-or-doubt claims about AI in the space of about a week, roughly evenly split by party.
Scattered late-week stories — Newsom's federal travel-funding probe, the $5,000 tariff-dividend proposal, the Iran war, and a North Carolina Senate rally all produced one or two flagged statements each.
The question below is the same one asked of last round's reports: what does the research literature say happens when politicians use these tools — not whether any individual claim is true.
What the Research Shows
These are the same seven findings from the prior brief — they don't change from one week of statements to the next, so they're restated here in brief rather than re-derived.
Attack rhetoric is close to a coin flip, not a winning formula. Lau, Sigelman & Rovner's meta-analysis of 100+ studies (Journal of Politics, 2007) found “simply no evidence that attack [messaging] is a smart move” — roughly as many studies found backlash against the attacker as found benefit. A 2019 French Politics review found the backlash effect on the sponsor's own image is, on average, slightly stronger than the damage to the target.
Fear appeals do work — but the political evidence base is thin. A large meta-analysis (Tannenbaum et al., Psychological Bulletin, 2015) found fear appeals reliably move behavior, but almost all of that evidence comes from health communication, not politics, and the appeals that work best pair fear with a specific, actionable response — something generic political catastrophizing rarely offers.
It doesn't suppress turnout — it mostly rallies the base. The old “demobilization” hypothesis hasn't held up in recent meta-analyses. What has replaced it is “negative partisanship” (Abramowitz & Webster, 2018): fear of the other side now drives turnout more than enthusiasm for one's own — but that effect is concentrated in the base, not the persuadable middle.
It spreads fast online — that's attention, not persuasion. Brady et al. (PNAS, 2017) found each moral-emotional word in a political message increases sharing by about 20%, and attacking the outgroup drives far more engagement than praising the ingroup — but that's reach within sympathetic networks, not documented persuasion of anyone else.
It measurably erodes trust — mainly among the speaker's own side. A 2021 PNAS panel study (Clayton, Davis, Nyhan, Porter, Ryan & Wood) found election-legitimacy rhetoric didn't move broad support for violence, but it did erode trust in elections specifically among the speaker's own supporters. This round's parallel: AI “existential threat” and “Trojan horse” framing functions the same way — building alarm inside one's own coalition rather than persuading the other side.
Unsupported factual and character claims are the most exposed. Wood & Porter's replicated study (2019) found factual corrections mostly work and rarely backfire. That cuts against this round's most checkable claims — a “dementia” diagnosis asserted with no evidence, a “bribery” characterization of a policy proposal, an unnamed “existential” AI threat — since these are the statements most vulnerable to a credible correction actually landing.
The two parties are running structurally different playbooks. Grossmann & Hopkins' Asymmetric Politics (2016) documents Republican rhetoric skewing ideological/civilizational (“communists in Congress,” “party of hatred”) and Democratic rhetoric skewing toward specific-actor wrongdoing (“briber-in-chief,” “corruption and grift”). This round tracks that pattern again, with AI fear-framing as the one lane both parties now use in nearly identical form.
The Undecided Voter: A Different Audience
This finding also carries over unchanged: undecided and persuadable voters are a smaller, more skeptical, more substance-seeking audience than either party's messaging usually assumes.
It isn't built to move them. Kalla & Broockman's meta-analysis of 49 field experiments (APSR, 2018) found conventional campaign contact — the vehicle FUD travels through — has an average persuasive effect on vote choice indistinguishable from zero.
Persuadable voters skew away from what FUD rewards. Nai & Maier (2021) found tolerance for negative/uncivil campaigning tracks personality traits (low conflict-aversion, high Machiavellianism) that persuadable, moderate voters are less likely to have.
Where research finds durable persuasion, it looks nothing like FUD. Broockman & Kalla's “deep canvassing” studies (Science, 2016) found structured, non-judgmental, two-way conversation produced attitude shifts larger than 14 years of background social change, still measurable months later.
The numbers haven't moved much in a week. Headline “undecided/other” on the generic congressional ballot remains small (~5%, Emerson College, July 2026). Independents are at a record ~45% of self-identification (Gallup, 2026), and roughly 28% of independents remain undecided on their House vote this close to the election. A broader definition that counts “leaning but could still change” alongside “don't know” puts true persuadability at roughly 12–15% of the electorate — concentrated among moderates (~25% still fluid) far more than strong partisans (~2%).
Applying This to the Daily Reports
Mapping this round's specific tactics to the research above:
Ideological/civilizational branding (“communists in Congress,” “party of hatred,” “crazy people”) — consistent with the long-documented GOP-skews-ideological pattern (Grossmann & Hopkins); no consistent electoral payoff per the negative-campaigning literature, and a real chance of backlash on the speaker.
Unsupported character/medical claims (Pritzker's “dementia” claim, Beshear's “briber-in-chief,” Wyden/Schumer/Padilla's “rig”/“bribe” framing) — the most exposed claims in the set; Wood & Porter's research finds corrections of specific, checkable claims like these mostly land and don't backfire.
Vague existential/threat framing (Cox's “existential event,” Kennedy's AI “terminator,” Pritzker's “existential threat to humanity,” Vance's “Trojan horse,” Mace's “regulate into the ground”) — a new cluster this round, all evidence-free predictions about AI; this is the fear-appeal genre the Tannenbaum meta-analysis says works best paired with a concrete action, which none of these offer.
Sweeping, unfalsifiable superlatives (Garcia's “corruption and grift,” Gruters' “looming disaster,” Trump's “blood-soaked killing fields” contradicted by the same coverage's own crime data) — the flavor most rewarded by social platforms per Brady et al., but reach within sympathetic audiences isn't documented persuasion.
Election/institution-legitimacy framing (Wyden/Merkley on mail voting, Newsom's “the tyrant” investigation framing) — the one tactic with a documented, narrow effect: it erodes the speaker's own supporters' trust in institutions rather than persuading anyone else, per the 2021 PNAS panel study.
Caveats
This is a synthesis of peer-reviewed, widely replicated literature on negative campaigning, fear appeals, and elite rhetoric — not a study of these specific statements.
Most of the underlying research comes from paid advertising and lab experiments, a different delivery mechanism than convention speeches, Truth Social posts, and cable-news soundbites; the incentives are the same, but effect sizes may not transfer exactly.
Effects are heterogeneous — what backfires with persuadable independents can simultaneously mobilize the base, so “does it work” depends on which audience is actually being targeted.
The daily reports themselves flag many calls as genuinely borderline (noted individually in the source reports); this brief treats those the same as the clearer cases for pattern-matching purposes, which will overstate confidence on the softer calls.
Sources