News Corp-owned business intelligence search engine Factiva is adding generative AI summaries to its search results.
The news database, which is part of Dow Jones, has deployed Google’s Gemini technology as part of News Corp’s ongoing business partnership with the tech giant. News Corp signed a three-year partnership with Google in 2021 which included payments for content via Google News Showcase and technology sharing.
Factiva approached every one of its almost 4,000 sources for new generative AI permissions and received the go-ahead from a “significant subset” of them according to general manager Traci Mabrey.
Having generative AI summaries in the search results has added multiple new “royalty moments” to Factiva, she said, including when a publisher’s content is referenced within the summary.
Factiva is used by institutions and professionals in fields like business, finance, academia and government who buy subscriptions to access its searchable archive of news, analysis and research.
Factiva licenses content from publishers, including both paywalled and free websites, meaning they get royalties proportionate to how much their work is surfaced by users.
With the launch of the generative AI capability, responses to searches in Factiva will contain a multi-sentence summary of the question that was asked above citations for three direct sources with links back to their articles, three additional recommended searches, and then a more classic search results section with all the articles that were responsive to the question asked.
Factiva acted as ‘publisher first’ in outreach for new AI deals
Mabrey told Press Gazette Factiva has been guided by three main principles on its AI products: “being a publisher first, an arbiter for publishers and an innovator”.
The publisher has reached out and spoken to “every publication and every source” on its books asking for generative AI rights, meaning every publisher referenced in its search summaries has consented to be there.
This contrasts with the approach taken by Google itself, which summarises publishers’ work in its own AI summaries without giving them a “genuine” choice (if they want to stay in its main search results).
They include: The Associated Press, Swiss news agency AWP Finanznachrichten AG and The Washington Post as well as Factiva’s fellow News Corp properties in the UK and Australia.
Mabrey said publishers have been “very grateful and very pleased with the fact that we have come with a very transparent and very trustful approach, and I think that has been one of the rationales for us being able to get some very strong partners from around the world to come on board because we are being very candid about what we want to do with the publications, what we are going to be offering in royalty moments”.
She added that she did not believe hallucinations or inaccuracies would be a problem as they have been in some generative AI products because only trusted sources are being used.
“So we are not bringing in just unvalidated information and what we think that is able to do is that that’s able to really ensure that our cutting edge technology is built on the foundation of the trusted journalism, data and analysis that’s needed at the core to begin the process.
“What we want to ensure is that by identifying those sources and building a highly relevant and contextual semantic search, that is going to be able to preclude what we hope is going to be a significant amount, if not all, of the hallucinations, because it is a very clean dataset that is going to be delivering the generative AI results, and what we want to be able to do is continue that steady stream of high quality views and information as we go forward.”
Other AI-generated ways publishers can earn royalties at Dow Jones include through the Factiva Feed for GenAI and Dow Jones Newswires GenAI Feed, which provide enterprises and financial firms with “licensed, copyright-compliant full-text content to power custom generative AI applications” like chatbots, as well as the Risk Center product for financial crime and risk management screening, and the automated due diligence report product Dow Jones Integrity Check.
Google gave Factiva ‘privacy and security’
Factiva’s new generative AI addition has been engineered with Google‘s Gemini models on Google Cloud as an extension of the existing partnership with the tech giant.
The Google Cloud partnership initially rolled out semantic search on Factiva, meaning AI interpreted the meaning and intent of a search rather than simply matching keywords.
Of the latest addition, Mabrey said: “Why we were very pleased to partner with Google on with this is that they offered us the privacy and security that was paramount for us, given the fact that we have our publishing IP within this Google Cloud instance, we also have the publishing IP for the external publishers and media and market data entities that are part of the ecosystem.
“So having the privacy and having the security was critically important, but then also being able to have a semantic search paradigm that was built solely on relevancy and contextual search criteria was also paramount, and they were able to truly deliver that for us.”
Google Cloud’s North America president Michael Clark said in a statement: “Dow Jones’s use of Google’s Gemini models on Google Cloud to power gen AI search summaries marks a significant step towards leveraging the technology to enhance information access and analysis.
“By combining our models with the extensive Factiva content set, we’re showcasing how GenAI can be a powerful tool for growth for the media industry.”
Dow Jones owner News Corp has separately done a deal with Google’s biggest rival in the AI space, OpenAI, for the use of the publisher’s current and archived content in products like ChatGPT.
Mabrey said: “It’s a very different agreement than what we have with our IPs [information providers] for our generative AI search in Factiva, but it really does apply very similar guiding principles… We apply those same guiding principles to our own IP and to the partners that we have around the globe.”
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog