When an account online says it is an ordinary voter, what would tell you whether that identity is genuine? A convincing profile is not proof either way. A recent report from Graphika describes Spamouflage-linked activity on X that included accounts presenting themselves as American voters and posting divisive political content ahead of the U.S. midterms. The findings raise a useful question about political influence online: how can researchers identify coordinated or deceptive activity without claiming more than the evidence supports?
Graphika’s October 30 report, “The #Americans – Season 2,” describes three sets of Spamouflage-linked activity on X. One involved accounts posing as U.S. voters. Associated Press coverage adds examples from the findings, including four accounts with a history of Spamouflage activity that amplified MAGA-styled personas. AP also reports on another account that presented itself as a Philippine culture blog while posting pro-China and divisive U.S.-focused content. Graphika’s report and AP’s coverage describe the activity as linked to Spamouflage.
That phrase matters. “Linked to” describes an attribution made by researchers; it does not, by itself, establish who directed each account or who was ultimately responsible. The reporting documents alleged deceptive personas and political posts, but it does not prove that a Chinese-government order drove this specific activity. AP says China’s embassy denied that China interferes in U.S. elections. The available account should therefore be read as a report about researchers’ findings, not as a settled account of command or motive.
A useful way to picture the problem is a borrowed coat: an account can wear the appearance of a local identity, but the coat alone does not tell you who put it on, how it got there, or whether anyone believed it. Researchers look at the surrounding activity to make an attribution; readers should keep the limits of that attribution in view.
What a persona can—and cannot—tell us
A political persona is the identity an account presents to other users. In this case, Graphika says some Spamouflage-linked accounts presented themselves as U.S. voters. AP’s account of the findings describes MAGA-styled personas and a purported culture blog with a different mix of pro-China and U.S.-focused posts. The contrast is a reminder that deceptive presentation need not take one uniform form.
But identifying a persona is not the same as measuring its effect. The evidence provided does not establish that these accounts persuaded voters, shifted political debate at scale, or affected an election. Nor does it establish that the activity used artificial intelligence. Political content can be divisive without being effective; finding a post or account is not evidence that it changed anyone’s mind.
Follower counts need similar care. AP reports that Graphika could not determine whether an account with more than 200,000 followers had been hijacked, sold, or otherwise changed hands. That uncertainty makes the number a poor shortcut for estimating how many people saw or were influenced by the political posts. A large audience attached to an account does not establish the reach of particular content, especially when the account’s history or ownership is unclear.
Influence is not only persuasion
The most useful context in the AP report is an expert assessment about what this kind of activity may accomplish. Darren Linvill told AP that Spamouflage has been more effective at harassing members of the Chinese diaspora and suppressing information than at changing mainstream U.S. political conversation. That is an assessment of the broader activity, not proof that the accounts in this report achieved those outcomes.
It also widens the question readers might ask. Political influence is often discussed as a contest to persuade a large national audience. But harassment or suppression of information can matter even without a measurable swing in mainstream opinion. The evidence here does not establish the effects of these particular posts on any community. Linvill’s comment instead offers a reason not to judge influence operations only by whether they visibly alter a national political debate.
A careful reading can hold both points at once: the reported personas and posts are worth understanding, and their impact remains unproven. The same restraint applies to attribution. Graphika’s report and AP’s coverage identify the activity as Spamouflage-linked, but AP did not independently replicate Graphika’s account attribution. That does not make the findings meaningless; it clarifies what kind of evidence readers have in front of them.
A practical way to read the next report
When you see a claim about a suspected influence operation, separate three questions. First, what did researchers observe? Here, the reports describe accounts presenting particular identities and posting political or pro-China content on X. Second, what is being attributed? The activity is described as Spamouflage-linked, but the evidence supplied does not establish who directed each account or prove a government order for this activity. Third, what effect is demonstrated? The available reporting does not show persuasion, a large-scale shift in debate, or an election outcome.
That separation is useful because dramatic claims often bundle observation, attribution, and impact together. A post may be real, the persona may be deceptive, and the activity may be linked to a known operation—while the operator’s identity and the audience effect remain uncertain. Each is a different claim, and each needs its own evidence.
The Graphika report is therefore a record of alleged deceptive political personas ahead of the midterms, not a measurement of votes changed. The concrete takeaway is neither to dismiss such activity nor to assume it succeeded: notice what researchers say they found, look for the limits they acknowledge, and treat claims about who directed it or what it accomplished as separate questions.
Sources
- The #Americans – Season 2 (2026-10-30)
- AP coverage of Graphika’s Spamouflage findings (2026-10-30)
