Flagged but Fooled? Rethinking Misinformation Inoculation on Short-Form Video Apps

Abstract

Social media platforms are fighting misinformation by adding warning flags and suggesting related content. However, it is still unclear how users understand these flags and whether they influence users’ willingness to share content or believe false information—especially on video platforms such as TikTok. Do users notice these flags? If so, do the flags change how they view the content? Inoculation theory suggests that if individuals are forewarned about the potential for misinformation and are exposed to weakened forms of such misinformation, they may become less susceptible to misinformation. Does inoculating users with flagged content reduce false acceptance and sharing intentions later? To address these questions, we conducted a user study (N = 322) utilizing a TikTok-like interface, employing a 2 (misinformation warning: absent, present) × 3 (counterargument in pre-suggested false-flagged content: absent, cue, action) between-subject experimental design. While users noticed flags as intended, their perceptions regarding the frequency and harm of misinformation remain unchanged. Furthermoe, neither warning nor counterargument messages effectively reduced users’ acceptance of misinformation or their intention to share it, highlighting the boundary conditions of inoculation theory within immersive short video contexts. Interestingly, media literacy and TikTok dependency have emerged as significant predictors of false acceptance. This study discusses the practical implications for the design of ethical social media interfaces.

Keywords

fact-checking false flag inoculation theory misinformation TikTok

Introduction

The rumor that Samsung withdrew a $1 billion advertising deal for the 2024 Paris Olympics was widely circulated online. This claim originated from a TikTok user’s post marked with a “Satire” flag.¹ Despite the label indicating the content was satirical, the video went viral on TikTok and contributed to the spread of misinformation about the Paris Olympics. This raises important questions: Why do users trust information marked as satire? Does such a misinterpretation occur even when a video is officially flagged as misinformation by the platform? If so, is this due to the persuasive nature of the video modality, or do users simply disregard the labels, even when they notice them?

The New York Times has argued that TikTok is becoming a new home for misinformation.² TikTok has embedded tools that allow users to separate video and audio tracks from existing content, allowing them to create seemingly original videos. When a video goes viral, this process allows the same or similar content to essentially go viral multiple times.³ Although content may circulate repeatedly across platforms, misinformation labeling remains a fundamental mechanism for mitigating the inadvertent dissemination of misleading information. A misinformation flag is a type of label, tag, or warning notice placed on content—especially online posts, videos, or articles—that signals the information might be false, misleading, or unverified. It was first introduced by Facebook in 2016 during the U.S. presidential campaign, following evidence that highly viral false news articles were disproportionately shared on the platform.⁴ Initially, Facebook applied warning labels to content that had been debunked by third-party fact-checkers. The platform later shifted to a “Related Articles” format, which presented fact-checking information beneath questionable stories.⁵ TikTok has since adopted a similar approach. Specifically, TikTok addresses potentially false or misleading content—sometimes mistakenly referred to by users as “false flags”—through a combination of user reporting, independent fact-checking partnerships, and automated moderation systems.⁶

However, the effectiveness of this approach remains controversial. Many users struggle to comprehend misinformation flags⁵ and prior research has produced mixed findings: some studies suggest that such flags reduce users’ acceptance of false information,^7,8 whereas others argue that they have limited or no effect.^9,12,18 Moreover, the lack of a standardized format, placement, or wording for misinformation flags can lead to greater confusion.^13–15 To narrow this research gap, this study aims to explore (1) whether users notice misinformation flags in actual video social media interfaces, (2) whether the misinformation flag reduces users’ credibility of the content, and (3) if not, whether additional intervention helps users resist misinformation and decreases their intention to share it.

Given that platforms have adopted a range of mitigation strategies already—such as de-ranking misleading content, applying fact-checking labels, and providing related informational links—this study turns to inoculation theory as an alternative explanatory framework to examine the effectiveness of additional interventions. Inoculation theory explains an individual’s resistance to persuasion through a biomedical analogy. Just as vaccines trigger the production of antibodies to protect against future infections, a weakened version of a persuasive attack can stimulate an individual’s cognitive–motivational processes, making them more resistant to persuasion.¹⁶ When applied to misinformation, inoculation theory suggests that if individuals are forewarned about the potential for misinformation and exposed to weakened forms of it, they may become more resistant to misinformation.¹⁸ Thus, we argue that warning users about the presence of misinformation on social media, along with the counterargument procedure—consisting of reasoning about why misinformation should not be trusted and exposing users to actual misinformation—will reduce false acceptance.²³

In highly immersive video-based social media such as TikTok, prompting the required action with a counterargument message could help users be attentive to the preexposed false-flagged content and generate their own rebuttals. Contextual warnings, which do not interrupt the user or require any action, are much less effective at changing behavior compared to interstitial warnings, which interrupt the user and necessitate interaction. Thus, we propose the research question and hypotheses (Table 1).

Table 1.

Proposed Hypotheses

	Hypotheses
H1.	A warning about frequent misinformation reduces (1) users’ false acceptance and (2) their intention to engage and share.	Not supported
H2.	A counter-argument message reduces (1) users’ false acceptance and (2) their intention to engage and share.	Not supported
H3.	A counter-argument procedure with action reduces users’ false acceptance and intention to engage or share more than a counter-argument with just a cue.	Not supported
H4.	A warning and counter-argument have an additive effect on false acceptance and sharing intention, so together they result in the lowest false acceptance.	Not supported

RQ.

For video-based social media users, controlling for media literacy and TikTok dependency, what is the relationship between a misinformation warning message (absence vs. presence), counter-argument procedure (absence vs. presence with cue vs. presence with action), and false acceptance, intention to engage, and intention to share the misinformation?

Method

A between-subjects experiment was used to test the conceptual model with two warning conditions and three counterargument conditions.

Participants

CloudResearch was used to recruit participants who resided in the United States, were 18 years of age or older, and had a TikTok account. This study was approved by the Institutional Review Board of the researcher. A total of 322 participants who passed the attention check were included in the final analysis; participants’ ages ranged from 21 to 66, with an average age of 39.5 (SD = 9.5). Of the participants, 55.2 percent were male, and 41.8 percent were female. The survey took ∼8.3 minutes to complete, and participants received $1.6 as compensation.

Stimulus materials

Using Figma, which is a collaborative interface design tool, we developed six conditions. The flow began with the TikTok login screen, followed by a warning message (presence vs. absence), four videos, a counterargument message (absence vs. cue vs. action), and a misinformation video (Figure 1). We also added a control condition without a false flag in the final video (Figure 2). The details of this flow are presented in Figure 1.

FIG. 1.

Stimuli flow. The series of photos shows screenshots of the actual short-form video stimuli presented to participants. They were first exposed to a warning message, followed by two videos, a counter-argument message cautioning them about false-flagged content, one general video, and finally a false-flagged video.

FIG. 2.

Misinformation flag. This shows the screenshot of short-form video stimuli with false flag and without false flag.

The warning message was operationalized as the presence or absence of the first susceptibility message. The counter-argument procedure was operationalized under three conditions: absence (control condition), presence with a cue, and presence with action (Figure 3). In the cue condition, a counter-argument pop-up appeared for five seconds before disappearing. In the action condition, the pop-up required participants to click “Ok” to proceed. In both cases, false information flags were shown afterward.

FIG. 3.

Cue vs. action in counter-argument message. This shows how counter argument message was manipulated. In the cue condition, a counter-argument pop-up appeared for five seconds before disappearing. In the action condition, the pop-up required participants to click “Ok” to proceed.

We considered two covariates in this study: media literacy (i.e., how much people understand digital media information) and TikTok dependency (i.e., how frequently participants use TikTok and how dependent they are on it).

Procedure

Participants were assigned to one of the seven conditions (six experimental condition + one control condition without false-flagged misinformation). Once they went through the TikTok interface, all relevant variables, including mediators and dependent variables, were measured (Table 2). Finally, the respondents were asked about their demographic information.

Table 2.

List of Measures for Variables

Variables	Sources	Items	α, M, SD
Media literacy	Austin et al. (2018)²¹	I compare the information I see on social media with other information I have seen before I accept it as believable. I look for more information before I believe something I see on social media. It is important to think twice about what social media posts say.	α = 0.97 M = 5.97 SD = 0.90
TikTok dependency	Yang et al. (2024)¹⁹	TikTok is part of my everyday activity. I would be sorry if TikTok shut down. I regularly check my TikTok.	α = 0.95 M = 4.29 SD = 1.68
False acceptance	Garrett and Poulsen (2019)¹⁷	The events described in this post occurred. I believe this post proves that walking protein exists in the human body. I believe this post is genuine.	α = 0.94 M = 2.90 SD = 1.37
Engagement intention	Sigerson & Cheng (2018)²⁰	I would “like” this post if it showed up on my wall. I would like to engage with this content. I would like to comment on the content.	α = 0.96 M = 2.41 SD = 1.43
Sharing intention	Garrett and Poulsen (2019)	I would like to “share” this post if it showed up on my wall.	α = 0.96 M = 2.90 SD = 1.37
Recognition of warning manipulation		At the beginning of your interaction with TikTok, did you see the message warning you how often you see fake videos?
Recognition of counter-argument manipulation		In the middle of your interaction with TikTok, did you see a pop-up explaining what percentage of people refuse to engage with misinformation to protect their online experience?
Manipulation of warning		Based on your interaction with TikTok you just explored, please answer the question.How likely are you going to be exposed to misinformation or fake content in TikTok?
Manipulation of counter-argument		Based on your interaction with TikTok you just explored, please answer the question.How much likely do users try to avoid fake content to protect their online credibility?

Media literacy, TikTok dependency, false acceptance, engagement intention, and sharing intention were measured with a 7-point Likert scale (1 — Strongly disagree, 7 — Strongly agree). Manipulations of warning and counter-argument were measured with a 7-point Likert scale (1 — Extremely unlikely, 7 — Extremely likely). Recognition of manipulation was measured using a three-option, yes, no, and I do not remember.

Measurement items were adapted from existing scales and adjusted to reflect the specific context of this study.

TikTok dependency refers to the relationship in which social media users’ goals rely on the resources offered by the platform (Sun, Rubin, and Haridakis, 2008; Yang et al., 2024). We argue that, rather than mere usage frequency, the level of dependency may affect how users interpret flagged misinformation.

Results

We conducted two manipulation checks for the warning message: whether participants saw it and whether it affected their perceived exposure frequency to misinformation. Only the first manipulation check was successful x² (2, 276) = 105.27, p < 0.001. We evaluated the counterargument message manipulation using the same procedure. Similarly, participants identified the message x² (4, 276) = 148.19, p < 0.001, but the message did not change their perception.

To understand whether the misinformation flag itself has an effect without additional intervention, we compared participants’ credibility of the content depending on the existence of the flag. Involvement in the topic was controlled for all the tests. A one-way ANCOVA showed that the misinformation flag did not affect participants’ credibility of the content. To test the hypotheses, a MANCOVA was conducted. The results revealed that neither the warning message nor the counter-argument had a significant effect on users’ false acceptance^a, engagement intention^b, and sharing intention. Warning and counter-arguments also did not show an interaction effect (Table 3).

Table 3.

False Acceptance: Warning × Counter Argument

	Counter argument
	No	Cue	Action
No warning
M	3.18	2.78	2.78
SE	0.20	0.20	0.21
Warning
M	2.63	2.84	2.71
SE	0.20	0.19	0.21

F(2, 268) = 1.31, p = 0.27, partial η² = 0.10.

The correlation analysis showed that TikTok dependency and media literacy were highly correlated with false acceptance, engagement intention, and sharing intention (Table 4). Linear regression analysis was employed to examine TikTok dependency and media literacy as predictors of false acceptance, engagement intention, and sharing intention. The results show that TikTok dependency and media literacy accounted for a significant portion of the variance in the outcome variables (Tables 5 and 6). False acceptance was positively associated with TikTok dependency and negatively associated with media literacy. A similar pattern was observed for both the engagement and sharing intentions.

Table 4.

Correlations of All Measured Variables

	1	2	3	4	5	6	7
1. Media Lit	1
2. TikTok DP	0.033	1
3. Prcv Und	0.228**	0.212**	1
4. Message Cre	−0.153**	0.139*	−0.324**	1
5. False Acpt	−0.119*	0.149**	−0.313**	0.848**	1
6. Engagement	−0.288**	0.307**	−0.214**	0.529**	0.503**	1
7. Sharing	−0.297**	0.254**	−0.221**	0.524**	0.501**	0.917**	1

Correlations significant at *p < 0.05 and **p < 0.01.

Media Lit, media literacy; TikTok DP, TikTok dependency; Prcv Und, perceived understanding; Message Cre, message credibility; False Acpt, false acceptance; Engagement, intention to engage; Sharing, intention to share.

Table 5.

Predictors of Engagement Intention

	β
TikTok dependency	0.32**
Media literacy	−0.29*

F (2, 271) = 28.77 R² = 0.18, p < 0.001.

*p < 0.5, **p < 0.01

Table 6.

Predictors of Sharing Intention

	β
TikTok dependency	0.26**
Media literacy	−0.32**

F (2, 271) = 23.66, R² = 0.15

**p < 0.01.

Discussion

Previous studies on preventing the spread of misinformation in social media have primarily focused on specific topics or on educating users,^23–25 often without considering how these strategies can be implemented within real social media interfaces.²⁶ Research applying inoculation theory also has typically been conducted in controlled settings²³ —for example, using games where users take the role of fake news generators to examine how such mechanisms affect their ability to combat misinformation.²⁶ While these studies offer valuable insights, very few have tested whether inoculation strategies are effective within actual social media environments. The current study sought to examine whether previous findings can be meaningfully implemented in a real social media platform.

The study results suggest that applying inoculation effects within real social media interfaces has important limitations. Participants noticed the message, but it did not meaningfully change their perception or subsequent false acceptance. Specifically, the experimental condition—characterized by a brief, single exposure to the stimuli—may have contributed to the null findings. Prior research indicates that warnings are often effective only after repeated exposure,²⁸ and meta-analyses have shown that inoculation treatments require time to build meaningful resistance to misinformation.²⁹ Given that implementing repeated exposures to achieve full inoculation in real social media interfaces poses considerable practical challenges, this study contributed to understanding a boundary condition of the theory in real-world social media contexts. Future research should explore how to address this boundary condition without undermining users’ social media experience.

In addition, the limited effectiveness of warnings and counterarguments may be explained by the immersive and fast-paced nature of short video formats,³⁰ which can easily divert users’ attention from the intended messages. Participants remembered seeing the intervention message, but it did not affect their perception. This might be related to the immersive nature of the series of videos to which they were exposed.³⁰ Furthermore, as the findings reveal that users’ media literacy and TikTok dependency significantly affected false acceptance and sharing intentions, platform designers should consider tailored strategies that account for users’ behavioral patterns and knowledge levels.

Conclusion

To reduce users’ false acceptance of misinformation, sustained efforts and interventions that provide actionable advice are essential. In our study, users noticed the misinformation flag even within engaging video stimuli; however, the message did not meaningfully affect their perceptions and, consequently, failed to reduce false acceptance. Identifying effective interventions that can meaningfully shift user perceptions and inoculate users in real social media interfaces remains a critical area for future research. In the meantime, recognizing that user perceptions may not be easily altered, enhancing media literacy over the long term may offer a more sustainable solution. Given that TikTok dependency is a significant predictor of false acceptance and engagement intention, platforms can develop tailored approaches for users with a high TikTok dependency. In addition, considering that content-sharing intention is negatively associated with users’ media literacy, we suggest that platforms can also consider targeting a specific audience that has low media literacy but high TikTok dependency with the right intervention.

Footnotes

Acknowledgment

The author would like to thank Andrew Pacheco for his work on the stimuli and Dr. Aiping Xiong for her feedback on the direction of the article.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

The current study did not receive any funding.

Author’s Contribution

Y.J. contributed to the conceptualization, implementation of the research, analysis of the results, and writing of the article.

a

Warning: F (1, 270) = 1.49, p = 0.22, partial η_p² = 0.005 Counter argument: F (2, 269) = 0.16, p = 0.86, partial η_p² = 0.001

b

Warning: F (1, 270) = 0.02, p = 0.9, partial η_p² = 0.000 Counter argument: F (2, 269) = 1.13, p = 0.32, partial η_p² = 0.008

References

1. Asiedu

. Did an Olympic sponsor pull support after the opening ceremony? WRAL NEWS; 2025. https://www.wral.com/story/did-an-olympic-sponsor-pull-support-after-the-opening-ceremony/21561760/ [Last accessed: August 7, 2025].

2. Hsu

. Worries grow that TikTok Is New Home for Manipulated Video and Photos. The New York Times; 2022. Available from: https://www.nytimes.com/2022/11/04/technology/tiktok-deepfakes-disinformation.html [Last accessed: August 7, 2025].

3. Zadrozny

. Misinformation by a thousand cuts: Varied rigged election claims circulate. NBC News; 2020. Available from: https://www.nbcnews.com/tech/tech-news/misinformation-thousand-cuts-varied-rigged-election-claims-circulate-n1247476 [Last accessed: August 7, 2025].

4. Jamieson

, Solon

. Facebook to begin flagging fake news in response to mounting criticism. The Guardian; 2016. Available from: https://www.theguardian.com/technology/2016/dec/15/facebook-flag-fake-news-fact-check [Last accessed: December 14, 2025].

5. Clayton

, Spencer

, Busam

, et al. Real solutions for fake news? Measuring the effectiveness of general warnings and fact-check tags in reducing belief in false stories on social media. Polit Behav, 2020; 42(4):1073–1095; doi: 10.1007/s11109-019-09533-0

6. Hern

. TikTok to introduce warnings on content to help tackle misinformation. The Guardian; 2017. Available from: https://www.theguardian.com/technology/2021/feb/04/tiktok-to-introduce-warnings-on-content-to-help-tackle-misinformation#:∼:text=If%20users%20try%20to%20share,spread%20of%20potentially%20misleading%20information [Last accessed: August 7, 2025].

7. Martel

, Rand

. Misinformation warning labels are widely effective: A review of warning effects and their moderating features. Curr Opin Psychol, 2023; 54:101710; doi: 10.1016/j.copsyc.2023.101710

8. Barman

, Colan

. Does explanation matter? An exploratory study on the effects of Covid–19 misinformation warning flags on social media. In: 2023 10th International Conference on Behavioural and Social Computing (BESC). 2023, pp. 1–7. Available from: https://ieeexplore.ieee.org/abstract/document/10386371

9. Junger

, Montoya

, Overink

. Priming and warnings are not effective to prevent social engineering attacks. Comput Human Behav, 2017; 66:75–87; doi: 10.1016/j.chb.2016.09.012

10.

10. Carragher

, Williams

, Carley

. Misinformation resilient search rankings with webgraph-Based Interventions. ACM Trans Intell Syst Technol, 2025; 16(1):1–27; doi: 10.1145/3670410

11.

11. Mena

. Cleaning up social media: The effect of warning labels on likelihood of sharing false news on Facebook. Policy & Internet, 2020; 12(2):165–183; doi: 10.1002/poi3.214

12.

12. Zhang

, Wu

, Kang

, et al. Effects of security warnings and instant gratification cues on attitudes toward mobile websites. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. Association for Computing Machinery: New York, NY, USA; 2014, pp. 111–114; doi:10.1145/2556288.2557347

13.

13. Glenday

. Facebook introduces ‘disputed’ label for fake news stories. The Drum; 2017. Available from: https://www.thedrum.com/news/2017/03/06/facebook-introduces-disputed-label-fake-news-stories [Last accessed: August 7, 2025].

14.

14. Smith

. Designing against misinformation. Medium; 2017. Available from: https://medium.com/designatmeta/designing-against-misinformation-e5846b3aa1e2 [Last accessed: August 7, 2025].

15.

15. Pennycook

, Rand

. Who falls for fake news? The roles of bullshit receptivity, overclaiming, familiarity, and analytic thinking. J Pers, 2020; 88(2):185–200; doi: 10.1111/jopy.12476

16.

16. Compton

. Inoculation theory. In: The SAGE Handbook of Persuasion: Developments in Theory and Practice, vol. 2. SAGE Publications, Inc; 2013, pp. 220–237.

17.

Garrett, R. K., Poulsen, S. Flagging Facebook Falsehoods: Self-Identified Humor Warnings Outperform Fact Checker and Peer Warnings. Journal of Computer-Mediated Communication 2019;24(5):240–258; doi:10.1093/jcmc/zmz012

18.

18. Lewandowsky

, van der Linden

. Countering misinformation and fake news through inoculation and prebunking. Eur Rev Soc Psychol, 2021; 32(2):348–384; doi: 10.1080/10463283.2021.1876983

19.

Yang

, Adnan

, Alivi

et al. Unveiling the Influence of TikTok Dependency on University Students’ Post-COVID-19 Health Protective Behavior. SMC, 2024; 12(1):390; doi: 10.11114/smc.v12i1.6625

20.

20. Sigerson

, Cheng

. Scales for measuring user engagement with social network sites: A systematic review of psychometric properties. Computers in Human Behavior, 2018; 83:87–105; doi: 10.1016/j.chb.2018.01.023

21.

Austin

, Austin

, French

et al. The Effects of a Nutrition Media Literacy Intervention on Parents' and Youths' Communication about Food. J Health Commun, 2018; 23(2):190–199; doi: 10.1080/10810730.2018.1423649

22.

22. Kaiser

, Wei

, Lucherini

, et al. Adapting security warnings to counter online disinformation. 2021. pp. 1163–80. https://www.usenix.org/conference/usenixsecurity21/presentation/kaiser [Last accessed: April 1, 2025].

23.

23. Suarez-Lledo

, Alvarez-Galvez

. Prevalence of health misinformation on social media: Systematic review. J Med Internet Res, 2021; 23(1):e17187; doi: 10.2196/17187

24.

24. Sylvia Chou

, Gaysynsky

, Cappella

. Where we go from here: Health misinformation on social media. Am J Public Health, 2020; 110(S3):S273–S275; doi: 10.2105/AJPH.2020.305905

25.

25. Trethewey

. Strategies to combat medical misinformation on social media. Postgrad Med J, 2020; 96(1131):4–6; doi: 10.1136/postgradmedj-2019-137201

26.

26. van der Linden

, Roozenbeek

, Compton

. Inoculating against fake news about COVID-19. Front Psychol, 2020; 11:566790; doi: 10.3389/fpsyg.2020.566790

27.

27. Banas

, Rains

. A meta-analysis of research on inoculation theory. Commun Monogr, 2010; 77(3):281–311; doi: 10.1080/03637751003758193

28.

28. Pennycook

, Cannon

, Rand

. Prior exposure increases perceived accuracy of fake news. J Exp Psychol Gen, 2018; 147(12):1865–1880; doi: 10.1037/xge0000465

29.

29. Ross

, Jung

, Heisel

, et al. Fake news on social media: The (in)effectiveness of warning messages. In: International Conference on Information Systems (ICIS). San Francisco, CA; 2018.

30.

30. Yan

, He

, Li

. Why time flies? The role of immersion in short video usage behavior. Front Psychol, 2023; 14:1127210; doi: 10.3389/fpsyg.2023.1127210