1 Year In, Google AI Overviews Are Not a Chronic Condition for Publishers, Yet

When Google first unveiled AI Overviews last March, the technology seemed poised to fundamentally reshape the landscape of search—then it started recommending glue as a pizza topping. But one year later, AIOs are no longer a joke. The product, which scrapes websites and uses the information to answer user queries at the top of the search page, threatens to send fewer people to publishers’ articles. In doing so, it could undermine one of the most fundamental ways in which media companies monetize their content. Below are four of the most important takeaways from the first year of the technology, pulled from interviews with search experts, digital media executives, and data measurement firms. AIOs are becoming more common When Google rolled out AIOs to the public in May 2024, they initially appeared in around 10% of all queries, according to Brightedge data. After some initial, high-profile hallucinations, Google began toggling back how frequently its AIOs appeared, adjusting their prevalence based on category and experimenting with them further. By July, AIOs had begun to appear only in around 7% of all queries. As Google improved the technology—it tested AIOs heavily during holiday shopping periods and introduced the snippets to entirely new categories—AIOs began appearing with more frequency. By February, AIOs surfaced in 16% of all queries, the highest prevalence since being unveiled. “As Google gets more confident, it’s ramping up AI Overviews,” said Jim Yu, chief executive of the search data firm Brightedge. AIO presence depends heavily on query category Despite the increased prevalence of AIOs on an average basis, the presence or absence of the snippets depends largely on the category of the query. Some categories, such as healthcare, business-to-business technology, and education, have a high likelihood of surfacing an AIO. Others, such as commerce, entertainment, and travel, have a far smaller likelihood of prompting the snippets to appear. Google is adjusting how often AIOs appear While AIOs fluctuate broadly based on category, their prevalence within those categories has also shifted over time. This shows that Google is testing and learning from the product in real-time, fine-tuning where it presents AIOs based on where they are most useful, according to search expert Kevin Indig. For instance, Google continues to deploy AIOs frequently in informational queries, but it has throttled their prevalence in transactional categories, where users are revealing a higher intent to make a purchase. In entertainment and commerce—the latter where queries suggest users are about to buy—Google rolled out AIOs at a high rate initially before later scaling them back. “Google does not want to disrupt its transactional business because that is where it runs its most lucrative auctions,” Indig said. The effect on clickthrough rates is minimal but negative Publishers’ primary fear with AIOs is that they will answer user queries so effectively that the amount of traffic they receive from Google will go down substantially, which would harm their business. So far, the data shows that AIOs do hurt clickthrough rate, but that it is minimal and heavily dependent on category. In fact, some publishers have claimed that the technology has hardly had an impact on their businesses at all. Executives from the digital media giants Ziff Davis and Dotdash Meredith have both shared in earnings reports that the effects of AIOs on their traffic have been minimal. And Google chief executive Sundar Pichai said in his most recent earnings call that AIOs actually lead to a net increase in the number of search queries. “People use search more with AI Overviews, and usage growth increases over time as people learn they can ask new types of questions,” Pichai said. But data from Brightedge and analysis conducted by Indig both find the opposite. According to Yu, there is a roughly 50% drop in the clickthrough rate when AIOs are present. Indig found the same, elaborating that if a publisher’s link makes it into the AIO, the decline in clickthrough rate is less dramatic—but still lower than in queries where no AIO is present. As is often the case, only Google knows the exact effect that AIOs are having on the ecosystem, and it has been reluctant to share the relevant data. Google Search Advocate Daniel Waisberg said at a recent panel that the company plans to debut a way for publishers to track AIO traffic through Search Console, but only after the product has stabilized. To quell some of the panic about AIOs, one media executive suggested that it would benefit Google to be transparent about just how benign the technology has proven to be so far. “Two things are true,” the executive said. “That AIOs are not as present as you might think, and that even when they are present, they are not as painful as you might think. It would seem to be in Google’s best interest to share that.”

Search Engine Optimization

1 Year In, Google AI Overviews Are Not a Chronic Condition for Publishers, Yet

The product threatened to upend search, but its impact has been more muted

Mark Stenberg

Paul Hiebert

About

Subscriptions

Events

Publications