Author: Nisrina Khotimah (Call for Commentary)Editor: Ayom Mratita Purbandani

Real versus AI-enhanced images depicting explosions at a U.S. military base in Iraq. [Source: X]

On 11th March, in one of the first communications of its kind in the region, Digital Dubai and the Al-Ameen service alerted residents about viral, AI-generated content depicting dramatised explosions and other attacks.1 While some comments under these images and videos reflected passive consumption, the growing number of comments identifying them as fabricated suggests a behavioural shift towards greater resistance to AI-generated misinformation.

The airstrikes and bombardments across the Arab states of the Persian Gulf, part of the broader geopolitical conflict involving the US, Israel and Iran, generated a surge in synthetic media on social media.2 During crises like these, when users are actively seeking real-time updates about their local situation, the saturation of fabricated content could spread confusion and panic. At the same time, repeated exposure to such content could also generate frustration and scepticism that motivates some users to warn others of the misinformation through comments and reposts.

Inoculation theory suggests that repeated exposure to weakened forms of misinformation, accompanied by corrective comments, could strengthen individuals’ resistance to future misinformation.3 As users become more exposed to synthetic media, and as they read more comments identifying them as fake, they could build familiarity with common AI signal markers such as visual inconsistencies and exaggerated destruction.

Drawing on inoculation theory, cognitive dissonance and evidence of active debunking from social media users during the airstrikes across the Gulf States, I argue that the proliferation of AI-generated misinformation could paradoxically help generate the conditions for its own resistance.

Too Much Misinformation Motivates Users to Flag and Warn

While specific, aggregate percentage surges for all AI-generated content across the region are difficult to isolate from general adoption rates, images and videos falsely depicting the impact of airstrikes flooded the online news environment in March 2026 with some examples garnering millions of views.4 Among the most widely circulated examples were videos falsely depicting the aircraft carrier USS Abraham Lincoln on fire and dramatising the impact of bombings on US air bases such as NSA Bahrain.56

Earlier examples of content falsely depicting bombardments in the Middle East, when generative AI tools were less widely adopted, attracted relatively limited public resistance against it.7 However, at the height of the 2026 airstrikes, when the global user figure increased by more than 1.4 billion,8 AI-generated misinformation was increasingly met with users actively contesting its authenticity. Debunking comments received high engagement, and corrective reposts circulated alongside the misinformation itself.

Comments identifying a viral video as AI-generated as top-liked comments (screenshot).

Cognitive dissonance theory offers one possible explanation for this corrective behaviour, which is distinct from the longer-term skill development that inoculation theory addresses.9 As users encounter false depictions of locations they are personally familiar with, they experience tension between the misinformation and their existing understanding of reality. The resulting psychological discomfort could motivate more media-literate users to engage in corrective behaviours, with publicly labelling content as fake serving both as a dissonance-reduction strategy and as a way to make their corrections more visible on social media.

This may also explain why comments identifying images and videos as AI-generated often received high engagement. Other users experiencing similar psychological discomfort may have been drawn to comments that validated their scepticism and contributed “likes” that amplified those corrective responses. As a result, what began as an individual attempt to resolve cognitive dissonance became socially reinforced through collective engagement, gradually normalising the act of publicly challenging AI-generated misinformation through comments and reposts.

Too Much Misinformation Trains Users to Recognise It

Where cognitive dissonance explains the impulse to flag misinformation, inoculation theory addresses what happens over sustained, repeated exposure. As synthetic media continues to saturate social media, and as corrective comments appear regularly, users could eventually become more trained to recognise AI signal markers.

This process is visible in the comments and reposts themselves. Beyond labelling content as AI-generated, users have also voluntarily demonstrated or explained why certain images and videos appear fabricated. Frequently cited AI signal markers during the airstrikes across the Gulf States included architecture inconsistencies, poorly rendered hands and discrepancies with official news reporting. In effect, these comments function as informal forms of “prebunking”, unintentionally equipping users with interpretive tools that could later be applied to future encounters with AI.3

An X user identifying a viral image as AI-generated (screenshot).

Inoculation theory compares this process to getting vaccinated.3 Much like how a vaccine exposes individuals to a weakened version of a virus in order to build resistance, repeated exposure to misinformation could strengthen audiences’ ability to resist future manipulation. A key condition of inoculation theory–the presence of refutation alongside the misinformation–may be met through corrective comments and reposts. Where this condition is met, each debunking could act as a distributed warning signal that allows users to build familiarity with the characteristics of synthetic media over time. Therefore, it is possible that the proliferation of AI-generated misinformation paradoxically contributes to improved media literacy and stronger public resistance against misinformation.

Debunking May Be Uneven

The picture is not straightforwardly optimistic. Research on the continued influence effect has shown that the corrections do not fully eliminate false beliefs, particularly when misinformation is emotionally charged or politically sensitive.12 In the context of the airstrikes across the Gulf states, fabricated depictions of attacks on familiar locations are likely to satisfy both conditions simultaneously, entailing that even widely circulated corrections may not entirely prevent misleading narratives from continuing to shape perceptions.

It is also important to note who is doing the debunking. Some of the comments and reports identifying synthetic media during this period demonstrated contextual familiarity. The forms of knowledge varied, ranging from aviation expertise and awareness of Google’s SynthID watermarking system, but they consistently reflected a level of discernment that not all users possess. This suggests that the visible corrective behaviour was likely disproportionately driven by users who already possessed higher levels of digital literacy, which may limit the extent to which this organic resistance can be interpreted as evidence of a general shift.

An X user identifying a viral image as AI-generated using contextual familiarity (screenshot).

Literacy Can Diffuse Through Visible Corrective Behaviour

Although debunking comments and reposts could predominantly come from more media-literate users, their increasing visibility still indicates the emergence of participatory resistance practices. In addition, research on incidental learning suggests that repeated exposure to corrective information could shape knowledge and attitudes without deliberate engagement.13 This may occur when users repeatedly encounter viral false depictions of airstrikes alongside debunking comments with high engagement, particularly given that reading comment sections has become a common aspect of online news engagement.14 Therefore, repeated exposure to such corrective exchanges may reinforce a consistent set of diagnostic cues, reinforcing pattern recognition even without active analysis. Even users who lack the technical expertise to independently verify footage may gradually internalise the visual and contextual markers that are characteristic of fabricated content.

The Comment Section as a New Site of News Verification

There has been increasing interest, in both corporate and academic discussions, on the role of comment sections as part of contemporary online news consumption.1516 As AI-generated misinformation becomes more widespread, comment sections are increasingly emerging as informal sites of news verification.

This is not the first conflict after the launch of generative AI, but it is among the clearest instances where users have frequently turned to comment sections or other commentaries to assess the credibility of viral content. On top of the fact that 31% of heavy social media users in the MENA region actively participate in the comment section this year,17 the scale of engagement that debunking comments and reposts received suggests that users were actively reading and interacting with them as part of how they processed the news that they were consuming.

These developments suggest that verification practices are becoming more democratic and socially distributed. Rather than primarily relying on public service announcements from platforms such as Digital Dubai and Al-Ameen, although this is not to say that such platforms are unnecessary for public digital literacy, audiences are collectively participating in the authentication of online news through social media engagement. Consequently, comment sections may evolve into a form of participatory gatekeeping in which public scepticism itself becomes how online news is evaluated, interpreted and contested by social media users.

References

1. R. Qaldi, “UAE authorities warn against fake AI content amid rising spread of deepfakes,” Khaleej Times, Mar. 11, 2026. [Online]. Available: https://www.khaleejtimes.com/uae/uae-warn-against-fake-ai-content-rising-spread-deepfakes ︎
2. S. A. Thompson and A. Cardia, “Cascade of A.I. fakes about war with Iran causes chaos online,” The New York Times, Mar. 13, 2026. [Online]. Available: https://www.nytimes.com/interactive/2026/03/14/business/media/iran-disinfo-artificial-intelligence.html ︎
3. S. Lewandowsky and S. van der Linden, “Countering misinformation and fake news through inoculation and prebunking,” European Review of Social Psychology, vol. 32, no. 2, pp. 348–384, 2021, https://doi.org/10.1080/10463283.2021.1876983 ︎
4. S. Yadav, “Deepfakes Across the Gulf: A Generation-Detection Gap,” ORF Middle East, 2026. [Online]. Available: https://orfme.org/expert-speak/deepfakes-across-the-gulf-a-generation-detection-gap/ ︎
5. T. Dube, “AI video of burning ship does not show USS Abraham Lincoln sinking after missile strike,” AFP Fact Check, Mar. 10, 2026. [Online]. Available: https://factcheck.afp.com/doc.afp.com.A2JQ2G3 ︎
6. M. Schenk, “Fact Check: Inconsistencies in before/after image showing damage to US base in Bahrain,” Lead Stories, Feb. 28, 2026. [Online]. Available: https://leadstories.com/hoax-alert/2026/02/fact-check-before-after-damage-picture-bahrain-dome-us-base.html ︎
7. A. Chopra, “What’s real anymore? AI warps truth of Middle East war,” AI Monitor, Apr. 1, 2026. [Online]. Available: https://www.al-monitor.com/originals/2026/04/whats-real-anymore-ai-warps-truth-middle-east-war ︎
8. S. Kemp, “Digital 2026 Mid-Year Global Update Report,” We Are Social, Apr. 22, 2026. [Online]. Available: https://wearesocial.com/id/blog/2026/04/digital-2026-mid-year-global-update-report/ ︎
9. A. Mehta, “Experience of discomfort due to information heterogeneity on Instagram: A study using cognitive dissonance theory,” SSRN Electronic Journal, 2024, http://dx.doi.org/10.2139/ssrn.4874497 ︎
10. S. Lewandowsky and S. van der Linden, “Countering misinformation and fake news through inoculation and prebunking,” European Review of Social Psychology, vol. 32, no. 2, pp. 348–384, 2021, https://doi.org/10.1080/10463283.2021.1876983 ︎
11. S. Lewandowsky and S. van der Linden, “Countering misinformation and fake news through inoculation and prebunking,” European Review of Social Psychology, vol. 32, no. 2, pp. 348–384, 2021, https://doi.org/10.1080/10463283.2021.1876983 ︎
12. C. Wittenberg and A. J. Berinsky, “Misinformation and its correction,” in Social Media and Democracy: The State of the Field, Prospects for Reform, N. Persily and J. A. Tucker, Eds. Cambridge, U.K.: Cambridge Univ. Press, 2020, pp. 163–198, https://doi.org/10.1017/9781108890960 ︎
13. J. K. Lee and E. Kim, “Incidental exposure to news: Predictors in the social media setting and effects on information gain online,” Computers in Human Behavior, vol. 75, pp. 1008–1015, 2017, doi: 10.1016/j.chb.2017.02.018 ︎
14. N. J. Stroud, E. van Duyn, and C. Peacock, “Survey of Commenters and Comment Readers,” Center for Media Engagement, Austin, TX, USA, Res. Rep., Mar. 2016. [Online]. Available: https://mediaengagement.org/research/survey-of-commenters-and-comment-readers/ ︎
15. T. Naab, D. Heinbach, M. Ziegele, and M. T. Grasberger, “Comments and credibility: How critical user comments decrease perceived news article credibility,” Journalism Studies, vol. 21, no. 4, pp. 1–19, 2020, https://doi.org/10.1080/1461670X.2020.1724181 ︎
16. C. Peacock, J. M. Scacco, and N. J. Stroud, “The deliberative influence of comment section structure,” Journalism, vol. 20, no. 4, pp. 503–523, 2019, https://doi.org/10.1177/1464884917711791 ︎
17. M. Burger, T. Greyling, S. Rossouw, F. Sarracino, and F. Wu, “Social media use and wellbeing in the Middle East and North Africa,” World Happiness Report, 2026. [Online]. Available: https://doi.org/10.18724/whr-tv9w-9h41 ︎