The surprising reality: Why AI didn’t disrupt the 2024 election as feared

With all the pre-election buzz about AI derailing the 2024 elections, the actual impact seems surprisingly modest. Alexandra Popken, VP of Trust & Safety here at WebPurify, reflects on this, highlighting the gap between our worst fears and the current reality.

Andrew Chow with Time recently captured a sentiment I’ve been wrestling with, yet perhaps hesitant to voice: that, in truth, AI has had a surprisingly underwhelming impact on the 2024 elections. Chow argues, “watchdogs and technologists warned that artificial intelligence would sow chaos into the 2024 U.S. elections, spreading misinformation via deepfakes and personalized political advertising campaigns…yet fears of the election being derailed or defined by AI now appear to have been overblown.”

Count me among the early alarmists of 2024. I’d voiced concerns that AI-generated content would severely exacerbate election-related falsehoods, potentially even influencing outcomes. Maybe it’s residual anxiety from my time at Twitter, where few anticipated the depth of Russian interference on social media in the 2016 U.S. Presidential election. I’ve come to expect the worst when it comes to election cycles and digital misinformation.

That’s not to say we’re not seeing AI’s presence in the election. For example, OpenAI reported that an Iranian group attempted to use ChatGPT to incite division ahead of the election, although its reach was limited. NewsGuard also highlighted the rise of AI-generated fake news sites, at times specifically crafted to spread false election information. These are real concerns.

However, on the front lines of content moderation, we’re seeing far less harmful AI-generated content than anticipated. At WebPurify, we observed that only 1.3% of the misinformation flagged via our content moderation efforts this year – only some of it election-related – was identified as AI-generated, such as synthetic or manipulated media.

So, why has AI had a milder impact than expected? Here are a few hypotheses…

Users are increasingly skeptical of what they view online.

A September 2024 Pew Research Center poll showed that well over half of Americans are concerned about AI being used to generate and spread false information during the election cycle. Call it increased digital literacy or heightened awareness, but the public’s understanding of potential AI misuse has become a protective factor.

AI’s capabilities are still limited – at least visually.

While AI-generated text remains challenging to detect, AI images often still display noticeable inconsistencies. Fingers, facial symmetry, and unnatural lighting are telltale signs that allow the public to detect manipulation, keeping the worst of synthetic visuals at bay. For example, the fake photo former president Trump shared of Harris speaking at a communist event in Chicago was unmistakably AI to the naked eye.

Tech companies and platforms were better prepared this time.

Unlike the 2016 election, when tech platforms were largely unprepared, today’s industry leaders – such as Meta and TikTok – have the benefit of hindsight, AI doomers in their ear, and the fear of regulation on their back. By and large, these companies have invested heavily in measures to identify and flag potentially harmful AI-generated content before it gains traction. Although the use of content labels for AI has been somewhat inconsistent in practice, they’re taking proactive steps to give users authenticity signals which is a good thing.

Candidates spreading AI-generated misinformation are often fact-checked almost immediately.

With their extensive reach and influence over millions, candidates (namely Trump) have become major sources of AI-generated misinformation. However, the visibility of these figures means that misinformation shared on such a large stage is more likely to be fact-checked in near real time, helping to mitigate its spread. Take the fake Taylor Swift images endorsing Trump that he shared to Truth Social. These posts were quickly identified as fabricated by Swift’s fans and various media outlets, challenging both the believability and impact.

The conclusion

It’s possible we have yet to witness the full impact of AI on this year’s elections, and we shouldn’t become complacent – as the technology advances, its influence will almost certainly grow. However, it’s worth acknowledging that, for now, the reality isn’t as dire as we once feared.