Launched at the end of 2010, Ngram Viewer can be used to detect trends in word usage in the millions of documents digitized by Google Books, covering a period from the sixteenth century up to the present day (eighteenth century for the French corpus). This article exploits the capabilities of this new application to examine the changing visibility of demographic vocabulary in written culture. It begins by looking at how data are selected and organized in Ngram Viewer, and shows that the counting of word sequences (or ngrams) without reference to context – a shortcoming pointed up by critics – is not an insurmountable problem. It then focuses on the main themes of demography, showing that the decline in demographic terminology since the 1990s is not an artefact. This decline is most visible for the demographic concepts linked to the marriage model, and for technical terms now confined to scientific journals (not covered by Ngram Viewer). An upward trend is observed, on the other hand, for terms linked to the social questions attracting a new generation of researchers, such as infecundity, perinatal mortality, sexual orientation, new transitions to adulthood, causes of death, health inequalities, gender relations, integration and discrimination, violence, systems of values. This suggests that demography must broaden its horizons if it wishes to maintain its former visibility and restore the link between science and society that has become so fragile today.