Theory of the Gimmick (2020) is Ngai's third major work, analyzing the aesthetic and economic structure of devices, techniques, and products that seem to work too hard and not hard enough simultaneously. The gimmick is a trick, a shortcut, a labor-saving device — and also an elaborate apparatus, an inflated solution, a form that draws attention to its own cleverness. Ngai traces the gimmick through patent medicines, infomercial products, marketing techniques, and artistic practices, demonstrating that the gimmick is not a defective commodity but a diagnostic form — one that reveals capitalism's tendency to generate solutions that inflate the problems they claim to solve. The book's application to AI is direct: large language models are gimmicks in the structural sense, enormous infrastructure producing adequate outputs, promising capability while delivering competence, working spectacularly hard to produce the mildly interesting.
The book builds on Ngai's earlier aesthetic theory while introducing a new dimension: economic form. The zany, cute, and interesting were aesthetic categories diagnosing affects. The gimmick is a form that produces those affects through its structure. The gimmick is zany (working frantically), cute (presenting complexity as simplicity), and interesting (perpetually novel without being transformative). The form synthesizes the earlier categories into a single diagnostic object — the capitalist artifact par excellence, promising too much and delivering just enough to sustain the oscillation between belief and skepticism.
Ngai's method is close reading of specific gimmicks: the Chia Pet, the Snuggie, the Flowbee, the ShamWow. Each is analyzed for its structural features: the promise (labor saved, problem solved), the disproportion (enormous apparatus for trivial result, or trivial apparatus for enormous claim), the affect produced (the oscillation between 'this is brilliant' and 'this is ridiculous'). The oscillation is the point. The gimmick destabilizes aesthetic judgment by producing output that evades stable evaluation. It is neither clearly good nor clearly bad. It is both and neither — and this equivocality is its diagnostic power, because it reveals a culture in which stable aesthetic judgment has become impossible.
Applied to AI, the gimmick framework cuts through the triumphalist/pessimist binary. Triumphalists say AI is transformative. Pessimists say it is fraud. Both are wrong in the same way: they try to settle the evaluation. The gimmick framework recognizes the evaluation cannot settle — and that the inability to settle is information. AI tools are genuinely impressive (infrastructure) and produce reliably adequate outputs (not excellence). Both assessments are correct. The oscillation between them is not confusion but accurate perception of a form that is structurally equivocal.
Malesic's 2025 application of the gimmick to ChatGPT is the clearest demonstration to date. The infrastructure is 'genuinely remarkable.' The outputs include 'brain-dead books, scam-filled ads, polished but boring essays.' The disproportion is the gimmick. The affect — simultaneous admiration for capability and disappointment in result — is the gimmick's signature. And Ngai's formulation: 'The gimmick lets us down only because it has also managed to pump us up.' The AI tool pumps up (reveals capability, expands possibility, collapses friction) and lets down (produces adequacy, displaces depth, trains tolerance for the interesting) in the same gesture — and the gesture is so smooth that the letdown passes as the cost of the pump-up, absorbed into the affect of progress.
The gimmick as economic form predates Ngai's theorization — patent medicines, miracle cures, too-good-to-be-true offers have existed as long as capitalism. Ngai's contribution was treating the gimmick not as a defective commodity (which fails to deliver) or a fraudulent one (which deceives intentionally) but as a legitimate form that delivers on its promise while betraying something else the user valued without knowing she valued it. The gimmick is not a lie. It is a partial truth that presents itself as total — and the partiality is where the analysis begins.
The gimmick oscillates. Simultaneous over- and undervaluation, working too hard and not hard enough, impressing and disappointing in the same gesture.
Labor-saving inflates labor. The promise is real — the shortcut works — but the saved labor is redirected into new demands the gimmick itself generates.
Evaluation cannot stabilize. The gimmick is neither clearly good nor clearly bad — it evades judgment by being both.
Disproportion is diagnostic. Enormous apparatus producing trivial results, or trivial apparatus making enormous claims — the gap revealing structural truth.
AI is the gimmick perfected. Billions of parameters producing competent adequacy — the most sophisticated labor-saving device that also inflates the demand for judgment.