What the Evidence Supports in Community Media Literacy Programming

Executive summary

Short, practical media literacy instruction has measurable effects on people's ability to distinguish mainstream from false news, but the effects are modest, they are not uniform across populations, and the strongest evidence concerns discernment rather than downstream sharing behaviour. Separately, attention-based interventions that prompt people to consider accuracy at the moment of sharing show effects on what people subsequently share. Programme designers should expect small improvements from a single session, should teach a procedure rather than a list of warning signs, and should not promise behavioural change that the evidence does not support.

Analytic question: What does the available published evidence support, and fail to support, about the design of short community media literacy interventions?
Overall confidence: Moderate confidence

Scope and analytic question

This review addresses one question: what does published evidence support about the design of short media literacy interventions delivered by community institutions?

It is written for people who will run a session, not for researchers. It covers four frequently cited studies, states what each establishes, and identifies where practitioners commonly extend the findings beyond what they support.

It does not attempt a systematic review. Four studies were selected because they are widely cited in programme design discussions, they address distinct mechanisms, and their limitations are instructive. A systematic review would be a substantially larger undertaking and is noted as a gap below.

Key judgments

Judgment 1. Short practical instruction improves discernment between mainstream and false headlines, with a modest effect size. Moderate confidence.

Guess and colleagues tested an intervention derived from a publicly circulated list of tips for spotting false news, using survey experiments in the United States and both online and in-person field experiments in India. They report increased discernment between mainstream and false news headlines in the United States sample and in the online India sample. The effect was not observed in the in-person rural India sample [1].

The practitioner-relevant reading is twofold. Brief instruction does something measurable. And the null result in one population is not a footnote; it indicates that transfer across contexts should be tested rather than assumed.

Judgment 2. Interventions that redirect attention toward accuracy at the moment of sharing affect subsequent sharing. Moderate confidence.

Pennycook and colleagues report a set of studies in which participants frequently indicated willingness to share headlines they would have rated as inaccurate when asked directly. A simple prompt drawing attention to accuracy, including a field experiment on social media, shifted the quality of subsequently shared content toward more reliable sources [2].

This supports a design choice: place the intervention at the decision point rather than only in advance. A session that ends with a printed card people keep near the point of sharing is closer to the tested mechanism than a session that ends with a lecture.

Judgment 3. The verification procedure worth teaching is lateral reading. Moderate to high confidence in the procedure’s superiority; lower confidence about how well it survives outside instructional settings.

Wineburg and McGrew compared professional fact checkers with historians and undergraduates evaluating unfamiliar websites. Fact checkers reached sounder conclusions faster by leaving the page and consulting independent sources rather than studying the page itself [3].

The strength of this finding for practitioners is that it identifies a procedure, not a trait. Procedures are teachable in a two-hour session in a way that dispositions are not. The limitation is that the comparison is of expert performance on a task, and durable adoption of the habit in everyday conditions is a separate question this study does not address.

Judgment 4. Programme messaging should not attribute the spread of false content primarily to automated accounts. Moderate confidence.

Vosoughi and colleagues analysed roughly 126,000 cascades shared by about 3 million people on Twitter between 2006 and 2017. False stories diffused further, faster, deeper, and more broadly than true ones across every category examined, with the largest differences in political news. The authors report that this difference persisted when automated accounts were removed from the analysis, indicating that human sharing accounted for the differential spread [4].

For a community session, this matters because it changes the framing from “there are bad actors out there” to “the mechanism runs through ordinary sharing decisions, including ours.” The second framing is both better supported and more actionable for participants.

What the evidence does not support

Three claims are common in programme materials and are not supported by the studies reviewed here.

That a single session produces durable behaviour change. The studies measure discernment in a task, or sharing over a short window. None of them establishes durable change in everyday information behaviour over months. Programme materials should not promise it.

That teaching people warning signs is sufficient. Warning-sign lists date quickly, particularly visual artefact lists for synthetic media. The better supported approach teaches a procedure that does not depend on the current state of production technology.

That the same intervention works across populations. The divergent result between online and in-person samples in the Guess study is direct evidence against assuming transfer [1]. Programmes serving populations unlike the study samples should treat effectiveness as an open question and collect their own simple pre and post measures.

Context

The environment these programmes operate in has shifted. The Reuters Institute’s 2026 survey across 48 markets, with a total sample of 97,520 respondents, reports 54 percent using social media and video networks for news, trust in news at 37 percent, the lowest recorded since measurement began in 2015, and 42 percent saying they sometimes or often avoid news [5].

Two implications follow for community programming. First, the intermediary between events and audiences is increasingly a ranking system, so digital literacy content about how feeds select is not optional background. Second, low trust and high avoidance mean that sessions framed as defending established media against falsehood are likely to meet resistance. A session framed around a method the participant controls is a better fit for the observed conditions.

Methods and limitations

This review consulted five sources, four of them peer-reviewed studies and one an annual survey report. Sources were selected for their prominence in programme design discussion, not by systematic search. No meta-analysis was performed and no effect sizes were pooled.

Three limitations should be weighed by any reader using this document.

Selection. Four studies chosen for prominence is a small and non-random sample of a substantial literature. Publication bias and citation bias both plausibly favour positive findings.

Generalization. Three of the five sources draw on samples from a small number of countries. Applying the findings to a specific community requires judgment about how similar that community is.

Currency. The platform environment described in the 2018 cascade study differs from the current one. The mechanism identified, human-driven differential spread, is more likely to persist than the specific magnitudes.

Confidence in the overall judgment is stated as moderate, reflecting consistent direction across independent studies alongside a small source base and known generalization limits.

Implications for a two-hour community session

The following design follows from the judgments above. It is a recommendation, not a tested curriculum.

1. Open with the mechanism, not with examples of falsehood. Explain that spread runs through ordinary sharing, which is supported by the cascade evidence [4], and avoids a framing participants may resist.
2. Teach one procedure, taught to fluency. Lateral reading, practised on three real sources during the session, rather than five techniques covered briefly [3].
3. Place a prompt at the decision point. Close by giving participants a physical or saved checklist intended for use at the moment of sharing [2].
4. Use current, local material. Fabricated examples are safe but less transferable. Where real material is used, verify it fully before the session.
5. Measure something. Six headline classifications before and after the session take five minutes and give the organization its own evidence, which no external study can supply for its population [1].
6. State what the session cannot do. Participants should leave knowing that one session is a start.

Gaps worth filling

A practitioner-facing systematic review of short community interventions, with effect sizes and population characteristics, does not appear to be readily available in a form community organizations can use. Producing or commissioning one would be a substantial contribution and is beyond the scope of this document.

Longitudinal evidence on retention beyond a few weeks is thin across the sources reviewed here. Organizations running repeat sessions with the same participants are well placed to contribute observational data on this point.

References

Numbered citations in the text above correspond to the entries below.

1. Guess, A. M., Lerner, M., Lyons, B., Montgomery, J. M., Nyhan, B., Reifler, J., & Sircar, N. (2020). A digital media literacy intervention increases discernment between mainstream and false news in the United States and India. Proceedings of the National Academy of Sciences, 117(27), 15536-15545. https://doi.org/10.1073/pnas.1920498117
2. Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A. A., Eckles, D., & Rand, D. G. (2021). Shifting attention to accuracy can reduce misinformation online. Nature, 592, 590-595. https://doi.org/10.1038/s41586-021-03344-2
3. Wineburg, S., & McGrew, S. (2019). Lateral reading and the nature of expertise: Reading less and learning more when evaluating digital information. Teachers College Record, 121(11), 1-40. https://journals.sagepub.com/doi/10.1177/016146811912101102
4. Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false news online. Science, 359(6380), 1146-1151. https://doi.org/10.1126/science.aap9559
5. Reuters Institute for the Study of Journalism. (2026). Digital News Report 2026: Overview and key findings. University of Oxford. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/dnr-executive-summary

Next step

Use the community discussion guide

Topics covered: Community PracticeMedia LiteracyInformation Resilience