How to navigate AI manipulation this election season

An attacker doesn’t have to change a single vote to damage confidence in an election. Sometimes, creating confusion is enough. We can protect the systems that run elections and still leave voters exposed to manipulation.

For example, in 2024 , Iranian cyber actors stole private campaign material and attempted to distribute it to influence the election. U.S. intelligence agencies also attributed a fabricated video showing ballots being destroyed in Pennsylvania to Russian actors. These different tactics shared the goal of exploiting division and undermining confidence.

Artificial intelligence makes that challenge more complicated. It can make deception cheaper, faster and easier to personalize, but the risk goes beyond creating more convincing fakes. It can also make it easier for dishonest actors to cast doubt on real evidence by claiming it was generated by AI.

Social engineering exploits how people think, whom they trust and what prompts them to act. During an election, attackers have plenty to work with, including strong emotions, familiar public figures and a constant stream of consequential information.

Consider a hypothetical call that sounds like a candidate asking for an urgent donation. Or a text claiming your polling location has changed. Or a video appearing to show an election official admitting misconduct. Each exploits a familiar identity or institution, and AI can help attackers produce convincing variations at scale.

The same approach can target campaign staff. An attacker impersonating a colleague could use a cloned voice to request an authentication code. One successful deception could expose private communications that are then leaked, selectively edited or used to support a misleading narrative.

An attack might begin with a social media post, continue with a fake website and conclude with a text. However, seeing a claim repeatedly does not make it independently verified.

The deeper problem emerges when awareness of deception turns into disbelief of everything. Legal scholars Bobby Chesney and Danielle Citron call this the “liar’s dividend.” As people become more aware of deepfakes, dishonest actors gain a more plausible excuse to dismiss authentic recordings, legitimate reporting or other evidence by claiming it was generated or manipulated by AI.

That said, knowing that anything can be faked tells us something about what the technology is capable of; it doesn’t tell us whether a particular piece of information is falsified. The challenge for voters is to learn how to question suspicious content without assuming everything they encounter is fake. They need practical habits that make manipulation more difficult.

First, pause when a message pushes you to act. Urgency, outrage and fear deserve extra scrutiny. Before donating, sharing or changing your voting plans, verify the request through a separate channel.

Second, look for independent confirmation. Several accounts repeating the same clip may all be relying on one unverified source. Look for credible reporting that establishes the source of the information.

Third, voters must go directly to the relevant authority. Verify registration, polling locations and deadlines with your state or local election office. For campaign communications, navigate to the campaign’s known website rather than following an unsolicited link. An official campaign statement establishes what the campaign says, while independent reporting helps assess the claim.

Finally, be comfortable saying, “I don’t know yet.” You don’t owe every viral post an immediate verdict. Waiting for evidence is a responsible decision and ensures you apply that standard even when a claim supports your political preferences.

But the responsibility extends beyond voters. These habits can help voters navigate uncertainty, but they cannot carry the full burden.

Campaigns should secure their accounts, keep original communications accessible and respond quickly to impersonation attempts. Platforms need effective reporting and escalation processes when deceptive election content appears. Nonprofit organizations such as Defending Digital Campaigns , which my organization supports as a technology partner, exist today to provide eligible campaigns with cybersecurity resources that can support this work.

AI also has an important role to play on the defensive side. The same technology that allows attackers to create convincing impersonations at greater speed and scale can help defenders identify malicious infrastructure, connect coordinated activity across channels and disrupt campaigns before they reach more people.

But technology alone will not solve the trust problem. As synthetic content becomes easier to produce, we need to make authentic information easier to verify. That means giving voters reliable places to confirm what campaigns and election officials have actually said, responding quickly when impersonation occurs.

The danger is not simply that AI will make people believe something false. It is that repeated exposure to manipulation will make them question whether anything can be trusted at all. Election security must account for both threats. Protecting the vote means protecting the systems that count it, but it also means preserving voters’ ability to know what is real.

Bobby Ford is chief strategy officer at Doppel .