There have been plenty of anecdotal stories from publishers over the past year or so who have reported that the introduction of AI summaries into Google’s search results has seriously dented traffic. Now, new research released by Pew Research Center gives a broader overview of how AI overviews are impacting click-through rates. A study run over one month earlier this year found that users were roughly half as likely to click on a search link whenever Google generated an AI summary, compared with searches where there was no AI summary present.
Pew Research Center’s study, which monitored search activity from 900 US adults in March, found that 15 percent of Google searches ended in the user clicking on a search result in cases where there was no AI overview. When there was an AI overview, this figure dropped to 8 percent. And while Google’s AI overviews come with their own set of links, leading to sources which helped inform the overview, click rates on these were very low. Users clicked on a link in the AI summary on just one percent of searches where an AI summary was generated.
The data demonstrates one of the major challenges facing publishers: AI tools are increasingly using publishers’ own content to answer user queries, starving the publishers themselves of traffic to monetise and denting revenues. The Pew Research data doesn’t capture the full impact of AI chatbots and search engines by any means. Chatbots which don’t have a traditional search element are also drawing traffic away from publishers, and while click-through rates from the likes of ChatGPT haven’t been publicised, anecdotal evidence suggests they are similarly low.
Questions and answers
Not all search traffic is equally impacted by the introduction of AI overviews. While 18 percent of Google searches seen by Pew Research generated AI overviews, some types of AI searches were much more likely to include an AI summary than others.
Searches which form questions, for example, will lead to an AI summary more often than not. Sixty percent of search queries which began with question words (for example ‘who’, ‘how’, ‘what’, ‘when’, or ‘why’) resulted in an AI summary. Searches that use full sentences are also more impacted, as 36 percent of searches which included both a noun and a verb were found to generate an AI overview.
These findings could be useful for publishers in terms of understanding which types of content are more likely to be affected by AI. But this might require a strategic shift, since tactics which worked in the pre-AI world are now being effectively punished by the introduction of AI overviews. Previously, for example, publishers might create content specifically to target questions which they knew users would be likely to search. ‘What time does the England game kick-off tonight?’-style articles are an extreme example of this. Now however, with AI overviews directly answering questions, this tactic is punished rather than rewarded (though it’s worth noting that Google’s ‘helpful content’ algorithm update had already started cracking down on spammy content designed solely with SEO in mind).
And there’s only so much that publishers can work around this new search landscape without simply losing a significant chunk of their output. It’s not just articles which are directly designed to answer user search queries which risk being starved of search traffic. Reviews, for example, could be hit hard, if searches like ‘Is the new Fantastic Four film any good?’ generate AI overviews picking key quotes from a range of articles.
Tragedy of the commons
For its part, Google has (in some cases at least) taken more of a cooperative approach to publisher relationships than some of the other tech giants. While Meta essentially pivoted away from news in response to new laws in various markets which gave publishers power to seek compensation for use of their content, Google launched Google News Showcase as a means of remunerating news companies.
And ultimately, Google’s AI tools feed off of publishers’ output, so it’s in Google’s interest (and the interests of other AI giants) to make sure publishers’ businesses are financially sustainable. Bloomberg reported yesterday that Google is in talks with multiple news groups around licensing their content for its AI products.
But this issue, pointed out by Cloudflare CEO Matthew Prince in his widely circulated comments on AI’s impact on publishers, is all these AI giants are operating in a competitive environment, where they can’t afford to be the only ones paying for publisher data which others are simply taking without payment.
Follow VideoWeek on LinkedIn.
