Study: AI-Powered Social Media Can Subtly Manipulate Opinion at Scale

Is AI quietly manipulating democracy through biased content or are humans still the greatest threat to public opinion?

JUL 23·Updated JUL 24

Study: AI-Powered Social Media Can Subtly Manipulate Opinion at Scale

story

Above: A 15-year-old teenage boy looks at an iPhone smartphone screen in his bedroom on July 1, 2026, in Somerset, England. Image credit: Anna Barclay/Getty Images

The Spin

Establishment-critical narrative

AI writing tools are quietly reshaping public opinion without anyone's consent. LLMs consistently nudge social media posts toward certain political positions even when told not to. These biases compound across millions of interactions, gradually shifting how entire communities think. Unless we intervene, the public sphere may soon be replaced by a cheap imitation of civic discourse.

Yale News Taylor & Francis Online OII

Pro-establishment narrative

The panic over AI manipulating public opinion ignores a basic truth: humans are already far better at spreading bias and misinformation than any algorithm. AI bias is correctable in ways human bias simply isn't, and mass persuasion through AI-generated content faces the same stubborn bottlenecks that have always limited political messaging. The real threat to democracy is human behavior, not the tools we use.

Reutersinstitute ITIF The Bulletin