TikTok exposes its users to more misinformation than any other major social platform, according to new research, with around one in four posts flagged as containing “problematic” content, a figure well above rivals Facebook, Instagram, and even the notorious troll pit X.

A report from non-profit organisation, Science Feedback, which works with scientists to fact‑check online content about health, climate, and other scientific topics, found that 25% of posts analysed on TikTok contained misleading or false information, up from around 20% last year.

When including abusive content, like hate speech, and borderline content, which toes the line of disinformation without outright falsity, the researchers found 43% of all TikTok posts were rife with mis-and-disinformation, a far larger share even than X, with 32%.

Three of the world’s most popular social platforms, TikTok, X, and YouTube, now show more misleading content than credible content, compared with only X in the first round of analysis.

This issue has been magnified by a growing pool of AI slop on every platform, with the study revealing that AI-generated mis/disinformation posts have already accumulated around 34 million views, but that only 16.5% of such content carries any visible label.

Across almost every platform, the study found a pattern of low-credibility, low effort accounts, those that regularly share false, misleading, or unverified information, receiving the most engagement relative to audience size.

On X, for example, an account posting false or misleading information can expect around ten times the engagement per post as a credible source with a similar following. Not only that, these low-credibility accounts on X are growing their audiences roughly 3.5 times faster than high-credibility ones – a disparity that researchers termed the “misinformation premium”.  

With such high levels of engagement, there are real-world implications. Around 43% of all false and misleading content analysed was connected to health, with the most prevalent misleading narratives being “major diseases can be treated with diet, natural or folk remedies”, and that “conventional medicine is unnecessary or harmful”.

Such posts ranged from the use of lemon juice to “detoxify” the body, to much darker claims that essential oils can be used to cure cancer and the oft-quoted, but totally unproven assertion that vaccines lead to neurological conditions in children.

Instagram was found to be the worst offender here, with 80% of misleading or false posts on the platform relating to health issues, followed by TikTok and Facebook, at around half each.

The war in Ukraine is another favoured topic for misinfo agitators, accounting for almost a quarter of problematic content, with regular posts that Ukraine is a “corrupt puppet state” controlled by foreign powers eager to weaken Eastern Europe, and that Zelensky is personally transferring $50 million a month to a bank in Saudi Arabia.     

While it’s easy for those with the barest knowledge of current events to shrug off such outlandish claims, the researchers warned that major social platforms are built in a way that consistently allows misleading content to spread, with agitators having plenty of incentive to continue spewing poison.

On YouTube, for instance, the study found 81% of low-credibility channels appear to be monetised, compared to 90% of high-credibility channels, evidence that YouTube itself is financially sustaining the spread of misleading content.

Recommended reading

• The UK Public Are Struggling to Spot Fake News Online, Says Ofcom
• Half of Enterprises Will Invest in Disinformation Security By 2027
• Ofcom: Four in Ten UK Adults Encounter Misinformation

The same is true of Facebook, despite a wider gap of 22% vs 51%, with researchers arguing that the fact that a fifth of low-credibility accounts appear to benefit from the site’s ad structure shows enforcement is sorely lacking. 

There is plenty that could be done in response. For a start, large social platforms like TikTok could redesign their recommendation and amplification algorithms to lessen the misinformation premium, and begin a process of labelling every piece of AI-generated slop identified.

They could also address the monetisation of misleading posts to disincentivise the most frequent misinfo peddlers, or even begin publishing standardised datasets for further study.

At the moment, that seems unlikely at best, given such a move would require the global cooperation of governments, lawmakers, regulators, and the platforms themselves to make it happen – a tall order in the current political environment.