People now speak thousands of words less per day than fifteen years ago, and Australian oral historian Shirleene Robinson warns that AI chatbots could deepen the loss. In an opinion column for The Guardian, she argued that the main risk is not weaker writing but weaker listening. Entering different worlds through other voices, she wrote, remains integral to humanity. For companies deploying conversational AI, this frames a direct question about what automated dialogue replaces.
Robinson argument and speech decline data
Robinson built her case around a long decline in everyday talk that predates current chatbots. Psychologists at the University of Missouri-Kansas City and the University of Arizona found people spoke about 338 fewer words per day for each year between 2005 and 2019. Daily volume fell from about 16,000 words to around 12,700 over that period. The steepest drop occurred among people under 25. The figures cover mostly the United States and Europe and stop in 2019, before widespread chatbot use.
Her authority comes from leading the oral history collection at the National Library of Australia from 2018 to early 2024. That archive holds more than 58,000 hours of recordings, with material dating back to 1957. During that work she heard First Nations Australians describe deep history and members of the Stolen Generations give testimony. She also listened to a refugee who fled Vietnam by sea and a Darwin doctor who survived Cyclone Tracy. The range of voices underpins her claim that hearing people unlike ourselves builds empathy.
Robinson contrasts that human process with what chatbots now offer: faster replies and so-called empathetic listening. Her point is that simulated attention is not the same as sitting with an unfamiliar account and staying with discomfort or difference. The research window matters here because the measured decline happened without generative assistants in daily life. The arrival of always-available interlocutors therefore lands on an already shrinking base of spoken exchange. That sequence makes the present moment different from earlier debates about writing quality.
What weaker listening means for business
For firms that use AI in sales, support, and internal processes, the practical effect concerns discovery and judgment. Customer interviews, complaint handling, and employee feedback depend on noticing hesitation, context, and unstated motives. If staff delegate first contact to bots and read only summaries, they gain speed but see fewer raw narratives. Small companies feel this in lost nuance from a handful of key accounts, while large organizations feel it as thinner signal across thousands of conversations.
There are clear limits to what this warning proves on its own. The speech data do not establish that chatbots caused the decline, since the series ends in 2019 and centers on the US and Europe. Robinson presents an opinion grounded in archival practice, not a controlled study of chatbot users. Business leaders should therefore ask vendors how systems preserve original wording, audio, and disagreement rather than smoothing them into a polite answer. It also helps to check whether automation reduces total human-to-human talk time or merely removes routine exchanges.
The marker to watch is whether daily spoken interaction keeps falling in post-2019 measurements, especially among workers under 25. New studies using the same word-count method would show if chatbots accelerated a prior trend or stabilized it. Corporate evidence will come from firms that track human conversation hours alongside bot resolution rates. If both human talk and comprehension of clients decline, the cost Robinson describes will have moved from cultural concern to operational fact.
