AI entered the presidential race this week, but not in the way many might have been expecting. In a post on Truth Social, Donald Trump falsely claimed that Kamala Harris had “CHEATED” and “A.I.’d” an image showing a large crowd of people cheering for her at a campaign stop in Michigan. The charge was quickly and easily disproved by news organizations (thousands of supporters were in fact photographed there from multiple angles); this was certainly not the “deepfake” crisis that experts have warned about for years, in which the existence of high-fidelity synthetic media leaves the public without the ability to distinguish between reality and fabrication. Nonetheless, Trump’s claim instantly boomeranged around the internet, amplified not only by his supporters but by pro-Harris accounts (to ridicule and condemn it) as well as technical experts (to fact-check and debunk it). Some commentators also seized on the occasion to speculate as to Trump’s mental well-being, a persistent theme of the summer campaign season. Was the post one more piece of evidence that the former president is losing his grip?
Trump thrives on the unique dynamics of social media—tapping into both the algorithms that shape the information landscape and what it means for individual users to interact online. Loaded words and terms (which can also function as hashtags) are everything; they’re sometimes called dog whistles, but linguists also refer to them as signifiers. This is a term that refers to a word’s actual form—its appearance on page or screen, its sound to the ear, its feel on the tongue—as opposed to its semantic meaning. What “A.I.” signifies in Trump’s post is not just a technology but Trump’s superiority, his dominance and mastery of all eventualities: He gets it. He’s on it. Nothing gets past him.
Putting signifiers into play is not a new tactic, of course. Perhaps the best example to date is the word Benghazi, unfailingly uttered by a certain segment of the right-wing commentariat as an almost reflexive response to mere mention of Hillary Clinton. As a signifier, Benghazi stems from the 2012 attack on a pair of American government compounds in that Libyan city. Four Americans, including our ambassador, were killed. Then–Secretary of State Clinton was accused by her opponents of slow-walking the appropriate military countermeasures, costing lives. Numerous congressional hearings ensued, none of them proving negligence on Clinton’s part but all of them consuming bandwidth and implanting the word in the minds of the electorate.
In this respect, even the oddly painstaking punctuation in Trump’s “A.I.’d” post may not be beside the point. It functions much like the multisyllabic foreignness of Benghazi. The fussy periods abbreviating the acronym A.I., the placement of the apostrophe: all communicate precision and specificity of knowledge, a command of what’s going on. Trump’s got them cold. He knows exactly what this is all about.
To be clear: I am not claiming that Trump was conscious of any of this as he posted. This isn’t another Trump-as-multidimensional-chess-master argument. Whatever tactical savvy is behind the post is the product of the reflexive way Trump uses media—his instincts for how to spike the narrative and shift the discourse—as well as his reckless disregard for the truth, and his consistent treatment of nearly all language as mere filler, or mere bluster, malleable and millable for his own ends.
But Trump’s most effective signifiers have never been entirely arbitrary. In the case of “A.I.,” the signifier feeds on many of his supporters’ inherent distrust of the media, as well as legitimate fears of the menace of deepfakes and a paranoid belief that Democrats and the so-called deep state must surely have such technologies at their disposal (and are willing to use them). The signifier also feeds on their desire to believe that Harris herself is some kind of synthetic candidate, manufactured to spec and illegitimately inserted into the electoral process.
Can anything be done to counteract this behavior? Fact-checking may be necessary, but it is never going to be sufficient. It’s an entirely reactive move, one that succeeds only by granting its subject, however spurious, undeserved consideration. The better move might be to play a more tactical and targeted version of Trump’s game. This is where J. D. Vance’s alleged (and disproved) illicit relations with a couch come in.
Some media outlets have tsked-tsked the meme, which is popular on the left. What’s the difference between this and Trump’s endless canards? The couch meme may be unwholesome and unflattering, but it does not attempt to distort the truth of an actual event. It posits a nonevent, and the fact that the original tweet included phony page references to Vance’s own memoir also made it effortless to fact-check; an untruth with its own refutation built right in.
Not everything that may be factually untrue is equally liable as disinformation, and not every untruth operates in the same way. The Vance-couch meme does not demonstrate that Democrats are poisoning the information landscape in equal measure with Trump, creating even more work for all the hapless guardians of accuracy. It uses parody and humor to provide an outlet for people’s distaste for an individual who seems to take an inordinate amount of interest in the bedroom activities of others. Mockery and ridicule operate in a different register from outright fabrication. They are effective signifying tactics not because they are falsehoods but because they can achieve a unique kind of accuracy.
Both “A.I.” and Vance’s couch are signifiers, fungible tokens in the collective language game that is the internet. Democrats shouldn’t have to apologize, certainly not until the internet is a far less hospitable place for right-wing lies, memes, and disinformation campaigns that are far more harmful in the aggregate. By recognizing language games for what they are, it’s possible to be a more responsible player—while still throwing the occasional elbow.