Amy Kolsky thought she got a good deal. Before a stay in France this summer, this 53-year-old American managed to find a travel guide on Amazon − Paris Travel Guide 2023 −, very well rated and, what’s more, at a great price completely competitive, or 16.99 dollars (15.82 euros), compared to 25.49 dollars for the equivalent guide from Lonely Planet. Once in hand, however, it’s a disappointment, the latter being nothing more than a compilation of vague descriptions with no real added value: “It’s as if the author went online and just copied and pasted from Wikipedia.”says this traveler, whose testimony was collected by the New York Times in August.
This example illustrates a growing phenomenon, that of the intrusion of artificial intelligence (AI) in publishing, including in tourism publishing, as evidenced by this proliferation, according to the American daily, of plagiarized or falsified travel guides. Several clues attest to the fact that the work in question was written by artificial intelligence.
Starting with the author, a certain Mike Stevers, who, although presented as a “renowned travel writer” on Amazon, looks like it was created from scratch. Unlike Rick Steves – a surname which is confusing -, an American author, who does exist and whose notoriety was made on trips to Europe, thanks to his immersive style in local life.
Thin guardrail
Another element, the British daily Times screened entire pages using a detection tool called Originalyti.ai. Here again, the result is clear, the probability that these texts were written by AI being extremely high. This phenomenon would also concern, according to the British daily which conducted other tests of the type, categories of works and subjects such as cooking, gardening, business, crafts, medicine, religion and mathematics.
Contacted by The world, Amazon says the titles that were mentioned in these articles have now been removed from sale. “ We invest significant time and resources to ensure compliance with our policies, and we remove books that violate them.”, indicates the spokesperson for the multinational. However, the company does not wish to communicate on the number of works concerned, nor on the extent of the phenomenon, nor on the compensation which has been made following complaints from its customers.
The group acknowledges that it does not prohibit the publication of this type of book, but reserves the right “to reject or delete content generated by artificial intelligence when[’il] determine[e] that it creates a disappointing customer experience”, continues this same source. On September 7, it also indicated that it had changed its policy towards authors who self-publish on Amazon without going through a publisher. The latter will now be obliged to declare whether or not their online content is produced by artificial intelligence. Whether it concerns texts, images or translations in particular. A very thin safeguard, however, against scams of this kind.