In this newsletter, I will estimate the impact of generative AI-driven search on news consumers and online media business, and consider strategic responses.
One-third of online news business at risk if consumers adopt Gen-AI search in the next few years
Twenty-nine percent of online traffic to news sites and 38% of new subscription conversions are at risk if tech companies scale generative AI-driven search, but that will likely take a couple of years, an INMA analysis finds.
News Corp’s CEO Robert Thomson warned against the threats. AI companies scrape journalism from the Web for training the AI without consent, nor remuneration. And they summarise original articles, reducing the need to visit news Web sites and disrupting the business model.
How big is the risk, exactly?
The new search experience: In response to the rapid adoption of ChatGPT, which garnered 100 million users within two months of its launch in November 2022, both Google and Microsoft released their own chatbots and started integrating generative AI into search engines.
Bing lets users choose whether to view a traditional list of links or chat with a bot. In its demo, Google displayed AI-generated direct answers above the links. AI answers may link to the original sources, but the companies haven’t shared any stats on click-throughs.
At an INMA study tour in New York, Microsoft’s Nikhil Kolar showed a chart confirming the intuition that only a fraction of Bing users chose chatbot over traditional search. The chart had no scale.
Kolar said that since the launch of a chatbot, the total volume of clicks on links has increased. He did not say though what proportion of the chat users click for more information vs. the search users.
Quantifying search: What Google will do matters most, as it amassed 93% share in the search market worldwide in May, per StatCounter. Microsoft’s Bing had 3% share.
In terms of the volume of visits, 7-month-old ChatGPT has already surpassed 14-year-old Bing, with estimated 1.8 billion visits in May vs. 1.2 billion, per SimilarWeb. At the same time, Google.com enjoyed 88 billion visits — or 30 times more than ChatGPT and Bing combined.
Morgan Stanley estimated that last year Google handled 3.3 trillion search queries worldwide.
Academic studies classify search queries by user intent. In a 2023 book Understanding Search Engines, Dirk Lewandowski summarised the findings:
Navigational queries, when the user intends to browse a specific Web site, account for 22%-42% of all queries.
Informational queries, when the user wishes to learn something, account for 11%-39%.
Transactional queries, when the user wants to buy a product or use a service, account for 22%.
All three categories of search queries may lead to news Web sites, but generative AI is likely to impact the informational and transactional queries. AI-generated direct answers might serve the users well enough, leading to fewer clicks on links.
Academic studies find users scarcely look at the search results other than the first ones, and many don’t scroll — particularly not on mobile. Direct answers provided on the search result pages are found to reduce user effort, improve satisfaction, and drive engagement with the search sites, but result in less clicks to sites.
The roll-out of AI-generated answers by Bing and Google is likely to be slowed down by the computing cost, although both companies face pressure to take on the challenge despite the expense. Alphabet’s chairman John Hennessy told Reuters that handling an AI-powered search query is costing 10 times more than a standard keyword search. He said though it posed: “a couple year problem at worst.”
So, publishers have a runway to respond to this disruption.
Sizing the business at risk: In 1Q 2023, on average, 29% of all pageviews on news sites were referred from search, per Chartbeat. It’s double the share of pageviews referred by social platforms (15%).
This reliance on search varies by region, with North European and American media sites depending less on search (16% and 20%, respectively), and Asian and African sites hinging on search (38% and 34%, respectively).
This leaves Asian, African, South European (33%), and Latin American (29%) news sites more exposed to any disruption of search referrals. Publishers in those regions relied more on digital advertising than reader revenue, per WAN-IFRA’s World Press Trends 2022-2023.
Generative AI-driven search might put digital reader revenue at risk too. Per Piano, in December 2022, a median brand saw 50% of paywall stops on sessions referred from search and 38% conversions to subscriptions.
Search is the second biggest source of paid stops and conversions after direct sessions and more important to publishers than traffic from social (14% paywall stops, but only 5% conversions).
Amid disruption by Gen-AI search, consider the consumer challenge and rebalance the business model
Don’t loathe news consumers for choosing AI-generated search summaries over articles on news sites. We all are struggling with information overload. How to respond strategically?
In 2022, Oxford’s Reuters Institute surveyed 93,000 news consumers in 46 markets — 38% said they avoided the news, often or sometimes.
The proportion has increased sharply across countries. For example, it doubled in Brazil and the U.K. over the last five years, with many respondents saying they are worn out by the amount of news.
Across the world, avoidance was up 31% since 2019. But is it any news?
Information overload: Since the invention of the printing press in the 15th century, people welcomed more access to knowledge but complained about drowning in it.
“The Internet accelerated the information overload,” said Professor Dominik Batorski, a sociologist from Warsaw University, when we met over Ukrainian dumplings this spring.
I was intrigued by his perspective on the history of the Internet’s adoption as the history of solutions to the ever increasing overload, which he published with the Polish Academy of Sciences.
In the 1990s, when the World Wide Web exploded, people navigated Web sites via directories, such as Yahoo!
Soon the directories became too large to be useful. Search engines, such as Google, offered a hand.
In the 2000s, people were overwhelmed by the number of search results, so they figured they could ask friends for help. Facebook, anybody?
In the 2010s, everyone had hundreds of friends on social media, so they sought a respite among the close friends — on WhatsApp and other messengers.
In the meantime, content aggregators popped up, curating content algorithmically and tailoring to our interests. Predictive AI helped us survive in this ocean of blogs, newsletters, books, songs, movies.
Here we are in 2023. Professor Batorski sees the primary use case of ChatGPT, Bard and other AI bots as synthesising tools rather than content generation tools: “Nobody wants more content!”
He explains: “The easier it is to produce content, the more is available. The more content there is, the less attention any piece gets. Consumers’ demand for solutions inspires and drives the supply.”
Consumer-driven disruption: In the 2019 book Unlocking the Customer Value Chain, Harvard professor Thales Teixeira and I analysed patterns of decoupling-driven disruption across industries.
We looked at how Airbnb took over business from hotels, how Amazon competed with retailers, or how Uber stirred taxis. The big idea of the book is that it’s customers who really disrupt businesses and not technology alone.
We saw broadly two ways the incumbents responded to the challenge:
Some tried to recouple the activities in the value chain that consumers decoupled with new technologies, e.g., with product tweaks, new distribution channels, protective measures by lobbying, lawsuits, patents, contracts.
Others adapted to disruption by rebalancing their business models, e.g., capturing value at different stages of the value chain or charging a different party.
If the search companies scale AI-generated answers, and consumers accept them, news publishers will face a similar dilemma — how to respond strategically. Let us consider both paths.
Recoupling path: Firstly, how could media firms reassemble decoupled activities: news discovery and access?
Consent, registration, and paywalls: Protect content from scraping or summarising by challenging these practices legally. Unlock content for logged-in users only.
Direct and alternative channels: Refocus on driving direct traffic. Invest in apps. Deliver content directly to inboxes with e-mail newsletters or to phones with notifications. Reassess social media platforms.
Differentiation: Invest in your brand awareness so people can find you when at need. Offer more in-depth or opinionated journalism. Engage users in conversations around content.
Non-text content: Expand audio, video, and interactive content, e.g., quizzes and games, which are harder for AI to synthesise.
Rebalancing path: Secondly, how could media firms adapt their business model to the looming disruption?
Reader revenue: Double down on subscriptions, shifting the revenue mix from advertising to the model less dependent on reach and casual users referred by search.
User experience: Leverage AI yourself to better serve consumers by personalising recommendations, summarising, and explaining information.
Collaborations: Consider alliances with other publishers to create a combined offering that’s more attractive to readers and harder for AI to decouple.
Big Tech payments: Demand fair payments from AI companies for the value your journalism creates but they monetise.
News publishers might pursue both avenues, although we found the rebalancing path usually more sustainable in the long term.
Interested in applying AI to news marketing and sales? Read my INMA report from January, AI Guide and ChatGPT Promptbook for News Marketers.
About this newsletter
Today’s newsletter is written by Grzegorz “Greg” Piechota, INMA’s researcher-in-residence and lead for the Readers First Initiative. In his letters, Greg shares original research, analysis, and best practices in growing reader revenue.