A research team has discovered that artificial intelligence can impersonate public figures in ways that feel more authentic to ordinary people than the real politicians themselves. Scientists created AI versions of statements from 112 public figures and tested them against actual statements from those same people. The result was consistent and troubling: people rated the AI impersonations as more coherent, more relevant, and more authentic than genuine political speech.
The study, reported by 404 Media, demonstrates that AI has crossed a threshold where it can mimic not just the content but the style and credibility signals of public figures. This is different from deepfake videos, which viewers can potentially spot as artificial through visual inconsistencies. Text-based impersonation is far harder to detect because readers have no visual cues and because political speech itself often sounds formal and rehearsed.
Why this matters becomes clear when you consider how information spreads. On social media, a false statement attributed to a politician will be more likely to convince people if it sounds coherent and authoritative. If that statement sounds more coherent than what the politician actually says, the false version becomes more credible. This creates a direct pathway for large-scale misinformation that could influence elections, shape policy debates, and erode public trust in institutions.
The capability to generate these impersonations requires no special access to the politicians themselves. Researchers used only publicly available statements and standard AI language models. A person with basic technical knowledge could create hundreds of convincing fake statements in minutes. This accessibility means the risk is not theoretical or distant. It is present now and growing.
What researchers flagged as particularly concerning is the mismatch between how easy AI makes deception and how difficult it is for people to detect it. Human brains evolved to spot lies through vocal tone, body language, and social context. Those detection mechanisms do not work against machine-written text. We rely instead on whether something sounds coherent and well-structured, exactly the strengths of modern AI.
The path forward remains unclear. Some platforms have begun labeling AI-generated content, but detection systems consistently lag behind generation capabilities. The technical arms race between better AI impersonation and better detection continues, with no guarantee that detection will win.
For people consuming political information online, the implication is straightforward: the coherence and polish of a statement are no longer reliable signs that it is genuine.
