Future is betting on AI to boost recirculation \u2013 and make that stickier audience more appealing to advertisers.<\/p>\n
The publisher\u2019s new proprietary AI-powered content categorization engine, called Advisor, acts like a \u201cbrain\u201d trained on Future\u2019s internal data, according to Jamie Samuel, head of commercial products at Future. Built into its audience data platform Aperture, Advisor lets Future quickly test an \u201cunlimited\u201d number of approaches to keep users engaged on its sites, ranging from chatbots to content recommendation widgets, he said.<\/p>\n
Advisor uses machine learning and OpenAI\u2019s large language models to analyze articles in real-time across Future\u2019s over 50 publications, including Marie Claire, Who What Wear, Tom\u2019s Guide and The Week.\u00a0<\/p>\n
Previously, Future\u2019s content categorization system analyzed pages as a whole. Now, the Advisor engine parses individual paragraphs, images and even price points mentioned within each article, Samuel said. Future\u2019s data science, engineering and product teams can then query Advisor\u2019s database to uncover audience insights data and build products tailored to user journeys, Samuel added.\u00a0<\/p>\n
Future took a measured approach to rolling out its AI-powered content recommendation engine, quietly introducing it in beta over the last six months across an increased number of sites.<\/p>\n
But the first test for Advisor was in 2023. Future wanted to trial the user insights coming from Advisor\u2019s content categorization, and see if they could be the foundation of an AI product that wouldn\u2019t hallucinate. The first use case was a chatbot for Tom\u2019s Guide called Hammerbot, which went live in beta in 2023<\/a>. Another chatbot \u2013 a shopping assistant chatbot for Who What Wear \u2013 debuted last October.<\/p>\n Advisor has now evolved from a limited experiment into a core pillar of Future\u2019s recirculation strategy. The AI-powered tool also replaced the search functionality on all of Future\u2019s sites in its CMS last September.<\/p>\n Advisor has led to a three-fold increase in onsite click-through rates, according to Samuel. This has also led to a \u201cnotable\u201d increase in revenue per session, Samuel added. He declined to share how much of an increase, but said users staying on Future\u2019s sites for longer and reading more content gives the publisher more opportunities to serve them ads or purchase products.\u00a0<\/p>\n Advisor analyzes each page to recommend relevant articles, products or deals \u2013 like surfacing top games for a new console or wall mounts for a TV. And it can change its recommendations around current events, such as Amazon Pride Day sales events.\u00a0<\/p>\n The engine powers four main widgets on Future\u2019s sites, Samuel said. The publisher will continue to use third-party content recommendation widgets on some of its sites from its deal with Taboola<\/a>. But the development of its own proprietary version signals a move toward in-house solutions, and the change means Future has gained more \u201ccontrol and precision\u201d over its onsite content recommendations, Samuel said.<\/p>\n \u201cEvery single page we\u2019ve got is an additional opportunity to serve someone [readers] something [they\u2019re] interested in. And [Advisor] is definitely doing it through a kind of experimentation framework. It\u2019s not doing it on its own,\u201d Samuel said. \u201cAs we run those experiments, we learn what works\u2026 We\u2019re just trying to complete the [user] journey, and as they complete that journey, we see the revenue per session go up.\u201d<\/p>\n