How persuasive are political cheapfakes disseminated via social media? The effects of out-of-context visual disinformation on message credibility and issue agreement
Open accessThe DOI matches its Crossref record; 38 of the 38 references it cites were found in a registry or an index; 4 of the 17 sources its claims cite were checked for support.
Abstract
Although extant research offered first insights into the effects of visual disinformation, we know markedly little about the effects of cheapfakes that rely on the deceptive re-contextualization of authentic visual materials disseminated via social media. Against this backdrop, we rely on online experiments (N = 874) and (N = 733) in which we exposed representative samples of participants to different types of (visual) disinformation in which the out-of-context presentation of visual material was central. Our findings indicate that cheapfakes based on video editing are not more credible than textual or image-only visual disinformation, while disinformation in general can lower support for a delegitimized politician. As a key contribution, we show that when cheapfakes are endorsed by other online users, they can mislead people and therewith succeed in delegitimizing targeted political actors, although richer modalities used in cheapfakes do not amplify deceptive content’s impact.
Overview
CoChat Summary
TL;DR
This survey experiment with Dutch participants found that out-of-context visual disinformation (cheapfakes) was not more credible or persuasive than text-only disinformation, though disinformation overall was less credible than authentic information. Social endorsement amplified effects mainly among people whose prior anti-immigration attitudes aligned with the deceptive message.
Key findings
- Disinformation was rated significantly less credible than authentic information (M = 3.65, SD = 1.16 vs. M = 4.88, SD = .93; t(872) = 10.69, p < .001), while issue agreement was similar across disinformation and authentic information (M = 4.25, SD = 1.84 vs. M = 4.31, SD = 1.89; t(872) = .37, p = .716). Candidate evaluations were slightly lower after disinformation (M = 4.45, SD = 1.11) than authentic information (M = 4.64, SD = 1.03; t(872) = 2.18, p = .030; regression B = −.26, SE = .11, p = .014).
- H1 was not supported: textual, image-plus-text, and cheapfake disinformation did not differ significantly in perceived credibility (all pairwise p = 1.00). However, the cheapfake was perceived as more directly reflecting reality (M = 4.56, SD = 2.10) than textual disinformation (M = 4.01, SD = 1.96, p = .034).
- Attitudinal congruence moderated effects: disinformation exposure interacted positively with prior issue attitudes on message credibility (H2a supported) and on positive evaluations of the depicted politician (H2c supported), but not on issue agreement (H2b not supported; prior issue attitudes strongly predicted agreement, B = 1.04, SE = .02, β = .89, p < .001).
- Endorsed disinformation had stronger effects for participants with congruent prior beliefs on credibility and issue agreement (H3a partially supported), but not on source evaluations; discrediting cues showed no significant interaction effects (H3b not supported). In a robustness check on MH370 re-contextualized footage, the cheapfake (M = 3.49, SD = 1.39) was equally credible as textual disinformation (M = 3.36, SD = 1.27), while real footage (M = 3.71, SD = 1.12) was slightly more credible than textual disinformation (ΔM = .35, SE = .11, p = .006).
Methods
- A survey-embedded online experiment used a 3 (disinformation type: textual vs. text-plus-image vs. cheapfake) × 3 (social endorsement: absent vs. endorsed vs. discredited) + 3 control conditions (authentic textual, text-plus-image, video speech) between-subjects design. Participants were 874 Dutch adults recruited via Kantar in the first week of December 2020 (from 1,267 responses; 69.0% completed); age M = 51.30, SD = 16.84, 49.0% female.
- The main topic was immigration, using an authentic speech by former Dutch politician Sybrand Buma altered to express radical right-wing anti-immigration claims. A second data collection served as a robustness check on re-contextualized footage about flight MH370 (authentic footage n = 250; textual disinformation n = 254; cheapfake n = 229).
- Analyses included independent-samples t-tests, one-way ANOVA with Bonferroni-corrected comparisons, and OLS regression models with interaction terms for attitudinal congruence and social endorsement. The study was not officially pre-registered.
Limitations
- Stated: the text, image, and video stimuli differed on cues that are hard to isolate (e.g., speech, moving images, background), and the study focused on a single right-wing issue in a single country, limiting transferability.
- Stated: the design used a single disinformation treatment; future research should expose participants to multiple manipulated messages and vary content plausibility.
- Evident from the text: the achieved sample (874) fell short of the pre-study power estimate of around 900 completes and the target of 1,000, and the authors note that a larger sample would be preferred for testing moderators.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures








Verification
Claims the paper makes
3 Valid · 2 Not valid · 12 Not checked · 28 Own results and interpretations
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“In addition, because real (unaltered or authentic) materials are used and re-contextualized in a deceptive manner, they are difficult to detect and flag as disinformation (e.g., Fazio, 2020).”
What the sources say
- Could not verifycites Fazio, 2020 sourceExperimental — accuracy not yet measured
Our copy of Fazio (2020) is entirely about an explanation-task 'friction' intervention: the Abstract and Findings report that pausing to explain why a headline was true or false lowered intent to share false (but not true) political text headlines, and Methods (p. n/a) shows the stimuli were 24 text headlines. No passage addresses cheapfakes, authentic materials used out of context, or the detectability/flagging of such content (searched: 'cheapfakes / recontextualized real materials difficult to detect', 'manipulated images, out-of-context video', 'types of misinformation ... hard to identify', 'introduction ... detection challenges'). The closest remark — Limitations, 'the results may differ when it is less obvious which stories are likely true or likely suspect' — concerns headline plausibility, not the detectability of re-contextualized real material, so it neither supports nor contradicts the attributed claim. (LLM, conf=0.72)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Hence, although concerns on AI-powered disinformation are voiced loudly in the public domain, most examples of visual disinformation and information warfare today are based on the deceptive re-contextualization or low-effort editing of visual materials.”
What the sources say
- Partly supportedcites Brennen et al., 2021 sourceExperimental — accuracy not yet measured
The Findings (Serving as Evidence; page not provided) report that 35% of the sampled COVID-19 misinformation used misleading visual context and about 10% used manipulation, all involving simple editing. The Discussion (page not provided) reports no AI-based manipulations in the sample. These findings support the contrast with AI-powered content, but not the broader claim that most visual disinformation and information warfare today use those methods: the study concerns COVID-19 misinformation, and the two categories together account for less than half its sample. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The processing of videos that combine different modalities can result in cognitive overload, as many cognitive resources are devoted to encoding. This comes at the cost of systematically processing substantial arguments of the message (Sundar et al., 2021).”
What the sources say
- Supportedcites Sundar et al., 2021 source
The Literature review (p. ~2-3, passage 1) states it almost exactly: cognitive overload/depletion is exacerbated when processing video because information arrives through many modalities (audio, video, graphics) plus peripheral structural features, so 'more cognitive resources end up being allocated to encoding, at the cost of other memory subprocesses,' and a richer modality such as video 'may be encoded more quickly but not processed as systematically as a text-only message.' The Abstract (passage 2) restates the same framing as the paper's own theoretical account (MAIN Model and Limited Capacity Model: video overburdens finite processing capacity, pushing receivers toward superficial cues). The claim's hedge ('can result in') matches the source's hedged wording, so nothing is overstated. (LLM, conf=0.90)
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“there is a strong relationship between participants' prior issue attitudes related to crime and immigration and agreement with the arguments made in the congruent message ( B = 1.04, SE = .02, β = .89, p < .001)”
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“this effect was not found for source evaluations (Table 4, Model III).”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
15 more claims
Create a free account to see all 45 claims and to verify the 1 reference still open.
- Probably not a substantial claim. 18
- The quote only partly states this claim. 1
- It may state two separate claims as one. 11
Statistics recomputed
Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.
- Not confirmedt(872) = 0.37reported p = .716 · recomputed p = .711
“t (872) = .37, p = .716”
- Consistentt(872) = 10.69reported p < .001 · recomputed p < .001
“t (872) = 10.69, p < .001”
- Consistentt(872) = 2.18reported p = .03 · recomputed p = .030
“t (872) = 2.18, p = .030”
- ConsistentF(3, 870) = 38.21reported p < .001 · recomputed p < .001
“F (3, 870) = 38.21, p < .001”
- ConsistentF(2, 730) = 4.97reported p = .007 · recomputed p = .007
“F (2, 730) = 4.97, p = .007”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
§ Credibility, issue agreement and candidate evaluations in response to disinformation · p. 11
“The results indicate that there is a signifi cant (and substantial) difference in perceived credibility for participants exposed to dis information ( M = 3.65, SD = 1.16) compared to real information ( M = 4.88, SD = .93); t (872) = 10.69, p < 0.001, 95% CI [.76, 1.10].”
§ Credibility, issue agreement and candidate evaluations in response to disinformation · p. 11
“95% CI [.76, 1.10]”
Citations
Cites
Statistical power analyses using G∗Power 3.1: tests for correlation and regression analyses
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009) · Behavior Research Methods · doi:10.3758/brm.41.4.1149 · Cited by 37,615 · cited 1×
“Soft quotas were enforced to ensure a distribution of age ( M = 51.30, SD = 16.84), gender (49.0% female) and For the power analysis that we conducted to inform sample size before commencing fieldwork, we relied on the G*Power 3.1 tool (Faul et al., 2009).”
InfluentialThe case for motivated reasoning.
Kunda, Z. (1990) · Psychological Bulletin · doi:10.1037/0033-2909.108.3.480 · Cited by 9,022 · cited 1×
“…to be motivated by a confirmation bias (e.g., Hameleers et al., 2024): People are more likely to uncritically accept information when it reassures their worldviews and political attitudes, and critically scrutinize information when it does not confirm their existing worldviews (e.g., Kunda, 1990).”
The limited capacity model of mediated message processing
Lang, A. (2000) · Journal of Communication · doi:10.1111/j.1460-2466.2000.tb02833.x · Cited by 2,153 · cited 1×
“In addition, considering that videos require more processing resources than textual information due to the richer mode of presentation used, visual disinformation is more likely to be processed heuristically (Lang, 2000).”
Reuters Institute Digital News Report 2015
Newman, N., Levy, D. A. L., & Nielsen, R. K. (2015) · SSRN Electronic Journal · doi:10.2139/ssrn.2619576 · Cited by 1,890
The disinformation order: Disruptive communication and the decline of democratic institutions
Bennett, W. L., & Livingston, S. (2018) · European Journal of Communication · doi:10.1177/0267323118760317 · Cited by 1,552 · cited 2×
“Considering that immigration is a salient issue in political disinformation (e.g., Bennett & Livingston, 2018), it is essential to direct our focus to this politicized issue.”
Social bots distort the 2016 U.S. Presidential election online discussion
Bessi, A., & Ferrara, E. (2016) · First Monday · doi:10.5210/fm.v21i11.7090 · Cited by 993 · cited 1×
“This finding confirms literature stressing the influence of astroturfing and the reliance on trolls and bots in real-life, such as the 2016 U.S. elections (Bessi & Ferrara, 2016).”
Vaccari, C., & Chadwick, A. (2020) · Social Media + Society · doi:10.1177/2056305120903408 · Cited by 978 · cited 2×
“However, our findings are in line with research suggesting that video-based disinfor-mation messages, such as deepfakes, are not more persuasive or deceptive than other modes of disinformation (e.g., Barari et al., 2021; Dobber et al., 2021; Ternovski et al., 2022; Vaccari & Chadwick, 2020).”
Political communication in a high-choice media environment: a challenge for democracy?
Van Aelst, P., Strömbäck, J., Aalberg, T., Esser, F., de Vreese, C., Matthes, J., Hopmann, D., Salgado, S., Hubé, N., Stępińska, A., et al. (2017) · Annals of the International Communication Association · doi:10.1080/23808985.2017.1288551 · Cited by 974 · cited 1×
“In a communication setting characterized by the abundance of counter-factual narratives, conspiracies, and confusion or fragmentation regarding factual beliefs (e.g., Van Aelst et al., 2017), it can be argued that true and false information are difficult to distinguish.”
Media Manipulation and Disinformation Online
Lewis, R. A., & Marwick, A. E. (2017) · Cited by 924
Populist Political Communication in Europe
(2016) · Routledge eBooks · doi:10.4324/9781315623016 · Cited by 804
Research works printed but not linked: 1
- Barari, S., Lucas, C., & Munger, K. (2021). Political deepfake videos misinform the public, but no more than other fake media. OSF Preprints, 13, 1–16.
Cited by
Gendered disinformation as violence: A new analytical agenda
Gehrke, M., & Amit-Danhi, E. R. (2025) · Harvard Kennedy School Misinformation Review · doi:10.37016/mr-2020-177 · Cited by 14
Yarchi, M., & Blatt, E. (2025) · Information Communication & Society · doi:10.1080/1369118x.2025.2592772 · Cited by 6
“Research shows that social endorsement of manipulated visuals can significantly influence their persuasive power, regardless of how sophisticated the manipulation technique is (Hameleers et al., 2024).”
Parra, N. K. S., & Pinela, Á. V. S. (2025) · Arandu-UTIC. · doi:10.69639/arandu.v12i3.1443 · Cited by 3
Everyday encounters with misinformation online: examining sources, topics and modes
Hourigan, A., Thomson, T. J., Notley, T., Park, S., & Dezuanni, M. (2026) · Information Communication & Society · doi:10.1080/1369118x.2026.2636132 · Cited by 2
“In this study, we investigate user-identified misinformation and untrustworthy content, acknowledging that such subjective perceptions mean content may include more than objectively false claims (van der Meer & Hameleers, 2025).”
InfluentialTopraksoy, A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6623443
Metadata
- Authors
- Hameleers, M., van der Meer, T., & Vliegenthart, R.
- Affiliations
- University of Amsterdam (NL)
- Wageningen University & Research (NL)
- Corresponding author
- Hameleers, Michael
- Venue
- Information, Communication & Society
- Publisher
- Informa UK Limited
- Published
- 2024-08-14
- Volume
- 28
- Issue
- 1
- Pages
- 61-78
- ISSN
- 1369-118X (ISSN-L)
- 1468-4462
- OpenAlex
- W4401585291
- Cited by
- 22
- References
- 36
- Citations by year
- 2026: 15 · 2025: 7
- FWCI
- 9.98
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- public-domain
- creativecommons.org/licenses/by/4.0/ (vor)
- doi.org/10.1080/policies-text-data-mining (tdm)
- Locations
- Information Communication & Society (publishedVersion, public-domain)
- UvA-DARE (University of Amsterdam) (publishedVersion, cc-by)
- Socio-Environmental Systems Modeling (submittedVersion, public-domain)
- Topics
- Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
- Social Sciences › Communication › Social Media and Politics
- Social Sciences › Sociology and Political Science › Media Influence and Politics
- Keywords
- message credibility, video editing, persuasiveness, social media
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: How persuasive are political cheapfakes disseminated via social media? The effects of out-of-context visual disinformation on message credibility and issue agreement. https://app.cochat.ai/papers/10.1080/1369118x.2024.2388079
- The paper
- Hameleers, M., van der Meer, T., & Vliegenthart, R. (2025). How persuasive are political cheapfakes disseminated via social media? The effects of out-of-context visual disinformation on message credibility and issue agreement. Information, Communication & Society. https://doi.org/10.1080/1369118x.2024.2388079
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.