Earlier this month, Perplexity, the search engine that uses LLMs to answer search queries, launched a shopping assistant. You enter the type of product you are looking for (e.g., Gift ideas for a male in his early 20s, who likes X and Y. My budget is Z). Then, you get recommendations, with links to where you can find additional information or actually purchase the article. We are told that the recommendations are not sponsored and, therefore, that we will get “objective” recommendations.
The launch of Perplexity’s shopping assistant reminded me of a paper published earlier this year, and which looked at shoppers’ willingness to accept AI recommendations when buying gifts. The paper is entitled “Gift giving in the age of AI: The role of social closeness in using AI gift recommendation tools”, and was co-authored by Yuxin Fu, David B. Dose and Radu Dimitriu.
The key finding from the paper is that shoppers are more willing to accept AI recommendations when buying gifts for distant friends than for close ones. That is because, when buying gifts for a close friend, shoppers believe that they are better than the AI at choosing a gift that a) reflects the depth of the relationship, and b) matches the tastes of the recipient.
However, when users could engage in dialogue with the AI, their willingness to accept the recommendation of a gift for a close friend increased. They still perceived the AI to do poorly in terms of finding a gift that reflected the depth of the relationship between gift-giver and recipient. However, they now perceived the AI to be able to understand the tastes of the recipient and to find a gift that matched those tastes. In addition, in the case of a gift for a distant friend, there was no significant difference in the willingness to accept a recommendation from a conversational AI and a one-shot AI.
The paper’s finding regarding the conversational AI seems particularly relevant for the new generation of chatbots powered by LLMs, such as Perplexity. After all, the appeal of AIs based on LLM technology lies, exactly, in their ability to understand natural language and to engage in a dialogue with humans. So, this all sounds very promising for sellers, who might now have an additional avenue to influence those of us trying to find a gift for their loved ones.
As I am actually looking for a present, I decided to give Perplexity a go. To be clear, this is NOT Perplexity’s shopping assistant, as that is available to Pro users in the US and, well, I am not one of them. Instead, my goal here is just to explore whether a conversational AI could help me find the right gift for a Gen-Z. Following my query, I got various suggestions organised by type, such as “Fun and Quirky Gifts”, “Practical and Useful Items” and so on. The recommendations included a short explanation, and the price in pounds.
That was a good start.
However, when I clicked on the links, I just ended up on webpages with curated gift guides. For example, one of the links took me to an article in Cosmopolitan magazine; another to the website thepresentfinder.co.uk.
I mean, in a way, the links were relevant. But they were no better than the recommendations I always got through a standard Google search. So, for the time being, my search for the perfect gift for this GenZ continues the old-fashioned way. But, who knows, maybe next year I will be able to outsource the task to an LLM.
Did you have any luck with AI recommendations for gift buying?
Ana Canhoto 