Concerns over China's online influence campaigns overseas, while valid, are often overblown as there is little evidence to show campaigns are actually working
On April 27, 2021, Tristan Harris, a tech critic and former Google employee, warned at a U.S. Senate hearing that we are facing … Tweets: @citizenlab and @yoyoel Tweets: @citizenlab : According to @lotus_ruan and @gabriellelim, fears of Chinese disinformation are often exaggerated by overblown assessments of the effects of China's propaganda campaigns. In other words: evidence of activity is not the same as evidence of impact. https://www.justsecurity.org/ ... Yoel Roth / @yoyoel : “Evidence of activity is not the same as evidence of impact” is the single most important differentiator of disinformation from other security challenges — and one of the most frequently overlooked. https://twitter.com/... Expand More For Next Unexpand More For Next
Context & Ripple Effects
The argument lands mid-debate. At a U.S. Senate hearing months earlier, tech critic Tristan Harris warned of a looming Chinese influence threat, while Stanford research comparing engagement across state campaigns had already found China driving far less traction than Russia or Iran on the same platforms. This piece — amplified by Citizen Lab researchers Lotus Ruan and Gabrielle Lim and endorsed by Twitter's then-head of trust Yoel Roth — crystallizes the pushback into one line: evidence of activity is not evidence of impact.
The follow-on coverage validates the thesis rather than overturning it: [[a:974155|an investigation traced China's propaganda machine to private contractors churning out content for Facebook and Twitter]] regardless of results, coordinated videos from expat YouTubers surfaced as another low-yield channel, and by 2024 experts concluded the yearslong Spamouflage Dragon operation against the U.S. had largely failed — tripped up, in part, by missing cultural context.
First-order effects
- U.S. policymakers weighing testimony like Tristan Harris's Senate warning must now separate two metrics their briefings tend to merge: how much foreign propaganda is posted versus whether anyone outside bot networks engages with it.
- Platform trust teams, following the standard Yoel Roth called the field's key differentiator, face pressure to report measured audience uptake alongside takedown volumes when assessing Chinese operations.
Second-order effects
- China's outsourcing model keeps producing anyway: contractors documented filling Facebook and Twitter are paid for content volume, an incentive structure that guarantees visible 'activity' even where persuasion fails — which in turn inflates the threat counts that alarmed observers cite.
- Rivals' more effective campaigns set the comparison benchmark; any assessment of Chinese influence now gets graded against Russian and Iranian engagement rates rather than judged on raw posting scale alone.
Third-order effects
- If the pattern holds — activity abundant, impact unproven, cultural context the recurring failure point — generative tools that cheaply produce native-quality content become the variable most likely to close the gap, making the current activity-versus-impact distinction far harder to draw.
- Western counter-disinformation resources risk systematic misallocation: budgeting against the loudest measurable signal (volume of posts) rather than the demonstrated effect, until impact auditing becomes a standing requirement of influence-threat assessments.
The trend: Influence-operations assessment is shifting from detecting and counting foreign propaganda activity toward proving whether it actually moves audiences — a bar Chinese campaigns have so far failed to clear.