The motivational gain control effect: mechanisms of predictive load and brand signal-to-noise ratio on danmaku interaction with AI advertising

Purpose This study develops the motivational gain control model, examining how dual cognitive demands of AI-generated advertising – internal decoding burden from structural transitions and external attributional ambiguity from executional convergence – jointly influence real-time audience interaction. Predictive load captures expectation-updating demand across adjacent stages; brand signal-to-noise ratio (BSNR) captures attributional clarity relative to competitors, constituting two dimensions of cognitive decoding valence whose evaluative meaning depends on product involvement. Design/methodology/approach Using panel data from 248 AI-generated advertisements and their danmaku interactions – synchronous viewer comments overlaid on video – on a major Chinese video platform, each divided into five periods, a panel vector error correction model estimates short-run and long-run effects and product involvement's moderating role. Findings For low-involvement products, content-level predictive load suppresses danmaku interaction while BSNR enhances it across both horizons. For high-involvement products, the positive effect of BSNR is significantly attenuated, whereas the negative effect of predictive load is only marginally attenuated, mainly in the long run. The BSNR attenuation provides strong evidence against a pure resource-depletion account; the predictive load attenuation, though marginally significant, is directionally consistent. Together, these patterns support motivation-dependent valence reassessment. Cross-modal congruence exhibits non-monotonic temporal patterns distinct from content-level features. Originality/value This study provides evidence against the fixed valence assumption shared across processing fluency, competitive interference and elaboration likelihood frameworks, proposing motivation-dependent valence reassessment as an alternative mechanism. It jointly models internal narrative dynamics and external competitive similarity through predictive load and BSNR and employs danmaku interaction as a real-time behavioral trace of evaluative response in AI advertising research. Peer review The peer-review history for this article is available at: Link to the website.