VOLUME 54
Correcting False Health Claims and Navigating AI-Generated Information
Highlights
Recent research on correcting false or misleading claims finds that corrections can improve the accuracy of people’s beliefs, but their impact on engagement may depend on reaching people before a claim has already spread widely.
And research on AI-generated search results suggests that people may be less likely to visit the original sources behind AI-generated answers. As these tools become more common, health organizations may need to consider how changes in information-seeking behavior affect their ability to reach audiences and build trust.
What We’re Watching
Social Media Fact-Checks Can Reduce Belief in False Claims, But Timing Matters For Limiting Reposts
As the public navigates information environments with large volumes of conflicting information from different sources, health communicators face the challenge of understanding how and when to address false claims. A growing body of research suggests that corrections after someone has been exposed to a claim on social media can reduce belief in misinformation. At the same time, the timing of corrections may matter for limiting reposts that give false claims further spread.
In recent years, social media platforms have scaled back or changed approaches designed to limit the spread of misleading content through professional fact-checking programs and instead invested in crowd-sourced fact-checking. Though professional fact-checkers may be perceived as more credible because of their training and expertise, corrections can still be effective in reducing belief in misinformation even when they come from a source that is perceived as less credible. In fact, crowd-sourced fact-checks can be as effective as expert fact-checks in reducing confidence in misinformation. X is one platform that uses a crowd-sourced fact-checking feature, known as Community Notes, to correct information deemed incorrect by its users. Recent investigations into Community Notes indicate that when added, they reduced the subsequent shares of misleading posts by 61.2%. But by the time a Community Note was published, the misleading post had often already been spread. The average time between a post being published and a note being added was about 63 hours, while posts typically had already received half of their first 36-hour reposts within about six hours after publication. As a result, Community Notes may be less effective at reducing the overall number of reposts than they are at limiting further reposts once they are added, suggesting that corrections may be most effective at limiting the spread of false claims when they appear early.
Why This Matters
While corrections can improve accuracy, their impact on reposts may depend on reaching audiences before a claim has already spread widely. Although different approaches can reduce belief in false claims to a similar degree, differences in their speed may matter alongside the accuracy of the content itself.
AI & Emerging Technology
As AI Expands, People May Be Getting Health Information Without Reaching the Source
AI-generated search results and other AI health tools are changing the way people encounter health information online, potentially making it less likely that users visit the organizations and websites that originally produced the information.
An analysis published last year by Pew Research Center tracked the web browsing activity of a nationally representative panel of 900 U.S. adults over one month in 2025. Google users rarely clicked on a link cited in an AI Overview in Google search results, doing so in just 1% of cases. When AI Overviews were present in search results, Google users were also less likely to click links from the search results that followed, doing so 8% of the time compared to 15% for pages without AI Overviews.
A March KFF poll found that about two-thirds of adults reported seeking physical or mental health information and advice from an internet search engine (68%) in the past year. With many search engines providing AI-generated summaries of search results, the Pew Research Center analysis suggests that people may be receiving information from websites without directly encountering those sources. For health organizations and other trusted messengers, evolving user behavior could change the relationship between providing information and building recognition or trust.
As AI features expand and user behaviors continue to evolve, some organizations are using techniques to increase the likelihood that their content will appear in AI-generated answers. Health care and pharmaceutical companies were early adopters of a practice called Generative Engine Optimization (GEO), designed to help sources appear more often in AI systems. One study showed these techniques increased a source’s visibility in AI responses by up to 40%, although effects varied across domains.
More From KFF
- KFF Issue Brief: Kindergarten Routine Vaccination Rates Continue to Decline
- KFF Issue Brief: State and Local Policies on School Vaccine Requirements: Overview and Current Status
- KFF Quick Take: New Guidance on Public Charge Determinations Does Not Specify Key Concepts, Likely Resulting in Immigrant Families Facing Ongoing Confusion and Fear
About The Health Information and Trust Initiative: the Health Information and Trust Initiative is a KFF program aimed at tracking health misinformation in the U.S., analyzing its impact on the American people, and mobilizing media to address the problem. Our goal is to be of service to everyone working on health misinformation, strengthen efforts to counter misinformation, and build trust.
The Monitor is a report from KFF’s Health Information and Trust initiative that focuses on recent developments in health information. It’s free and published twice a month.
Sign up to receive KFF Monitor
email updates
Support for the Health Information and Trust initiative is provided by the Robert Wood Johnson Foundation (RWJF). The views expressed do not necessarily reflect the views of RWJF and KFF maintains full editorial control over all of its policy analysis, polling, and journalism activities. The data shared in the Monitor is sourced through media monitoring research conducted by KFF.