Voice-activated shopping assistants have lasting benefits for online retailers, according to HKUST-led research. Zijun Shi and colleagues documented significant purchase increases on Tmall following the adoption of its voice AI, “Genie.” The mechanisms and limitations of this effect were also uncovered.

“The vocal search process can make it convenient for consumers to efficiently locate specific products while multitasking,” the researchers note. Nonetheless, “information presented by voice could be more difficult to process than information presented in text or visually.” Furthermore, for online shoppers with privacy concerns, “the perception of invasiveness associated with voice AI may also deter adoption.”

Despite these opposing possibilities, the data revealed a 16.6% rise in average weekly spending by voice-AI users. This climb appeared within four months of Tmall customers’ adoption of Genie and remained significant in the long term, despite some attenuation. “It is likely that the novelty effect is at work,” the authors propose: after a surge in purchases using the new device, “the product becomes more familiar, [and] consumers settle into a more modest pattern of increased spending.”

Consumption particularly increased among products that are low in substitutability, like books, art, and prescription drugs. Repeat purchases, and markets that buyers were already familiar with, also benefited more. This is explained by the limitations of voice assistants in comparing products or making detailed information digestible. “Consumers prefer to purchase products that do not require complicated selection processes through the voice channel,” the authors surmise.

Meanwhile, the spillover from voice AI into other sales channels mainly boosted purchases made using PCs—for buyers whose purchase intention is initially sparked by voice AI, large-screen PCs facilitate product comparison better than cellphones. “To realize the fuller potential of voice AI technology,” the authors recommend, “future product developers [should] be cognizant of the cost of information processing in their product design.”