Review of 327 studies shows customers eventually buy from chatbots as from people

The review finds that each of the three agent types has its own set of conditions that determine how closely it matches a human. Conditions identified in earlier work do not carry over from one type to another and can even push in opposite directions, so a finding about chatbots cannot be assumed to hold for robots or algorithms.

The authors use the results to say when automated agents can replace human ones in marketing roles to release capacity and alleviate labor shortages, and they set out a future research agenda.

The study asks when robots, chatbots and algorithms are equivalent to human agents in marketing roles. It combines 943 effect sizes from 327 studies, so the conclusions rest on a broad body of earlier experiments rather than one trial.

The main result is a gap between attitude and behaviour: customers may be skeptical of automated agents, but they value their performance and eventually choose or buy from them as if interacting with humans. Each agent type has distinct contingencies, and some earlier findings reverse from one type to another.

The authors extend the concept of automated social presence, describing automated agents as hybrid beings that give customers both a feeling of interacting with a humanlike entity and a feeling of interacting with a machine. They recommend where automated agents can replace human ones to release capacity and ease labor shortages.