SEO lessons from publishers one year after Google’s AI Overviews

These are just some of the ways publishers are shifting their strategies, having learned what’s working — and what isn’t — since the launch of Google’s AI Overviews.

“The old rules of SEO and optimizing content don’t apply like they used to,” said an SEO manager at a lifestyle publisher, who agreed to speak under condition of anonymity. “It’s getting harder and harder to keep up with the [search engine results page] changes that Google keeps throwing at us… It’s a fundamental shift that’s happening.”

Publishers have struggled to measure the impact of AI Overviews on search traffic. But as the feature appears in more search queries — and with Google’s AI Mode looming in beta, they’re already looking for ways to anticipate and adapt to a changing search landscape.

But publishers aren’t all on the same page. Some are proactively pivoting their SEO strategies for a generative AI search future, while others are betting that their focus on building direct relationships with their audiences via subscriptions and newsletters will pay off in the long run and make them less dependent on referral traffic.

Here are some lessons publishers have learned about AI Overviews so far:

For years, SEO at publishers centered on identifying high-performing keywords and crafting content around them to rank their pages high in search engines. Now, some publishers are reorienting those efforts toward keywords that don’t trigger AI Overviews, in hopes of preserving visibility and click-through.

That’s because the data available so far is showing that publishers receive lower CTRs if their content is surfaced in the AI Overview, compared to when it isn’t.

Mail director of SEO and editorial e-commerce Carly Steven told delegates at the WAN-IFRA World News Media Congress in Krakow last week that even when the Daily Mail showed up as the top link in AI Overviews, there was a 43.9% lower clickthrough rate on desktop and 32.5% lower on mobile.

Using keyword data to determine what topics and categories of content are triggering AI Overviews, has become more of a go-to method for figuring out which keywords are impacted and how to adjust their SEO strategies accordingly.

And some publishers figure, there are few reasons to optimize to appear in AI Overviews if they receive fewer click-throughs. So they’re targeting fewer of those than before.

“Targeting those keywords [that trigger AIOs] aren’t necessary,” the previous SEO manager said. “It’s not worth our time [anymore] to try to rank in search when we can get so much more traffic by optimizing for Google Discover.”

For example, there’s little incentive to go after search queries like “How much juice is in a lemon?” when the Google search page will now answer that question in an AI Overviews summary. Other keyword-based queries include ones that contain “how long” or “why is.”

Certain types of search queries circumvent AI Overviews

Branded searches — such as “NewFronts 2025 Digiday,” for example — are bringing in higher click-throughs, according to publishers.

These keywords are less likely to trigger AI Overviews (about 4.8% do so) according to a report by marketing agency Amsive. And when they do trigger an AI Overview, they get an average 18.7% increase in clickthroughs (unlike the 20% CTR decline for non-branded keywords), likely due to individual users searching for specific events or brand names and brand loyalty to certain publications, greater user intent and brand familiarity, per the report.

That may be why smaller, independent publishers without the same brand strength as The New York Times, for example, have suffered traffic drops since AI Overviews started taking up a big chunk of Google’s search results page.

Smaller recipe sites and blogs in particular are getting hurt by AI Overviews, according to Marc McCollum, chief growth officer at Raptive, which oversees over 6,000 independent publishers. Users can get recipes summarized right in the AI Overviews section, providing less of an incentive for that user to click through to the website to get that recipe, he said.

AI Overviews are mainly triggered when a user types in more complex, longer informational queries, such as “What is a hurricane?” and “What’s the difference between a hurricane and a tornado?” Users can get answers to those questions in an AI Overviews summary without clicking through to an article. Science and health content experienced the largest share growth of AI Overviews (up about 20%) from January to March 2025, according to a recent Semrush study.

“We’re not seeing [click-throughs] from those [long-tail, informational] keywords like we used to, [from the ones] that trigger AI Overviews,” the SEO manager said. While they did not share data that supported this, they did say that the impact was notable enough that their team was no longer targeting those keywords.

Google AI Overviews still a measurement black hole 

Google still isn’t sharing click-through rates from AI Overviews, and isn’t separating that data from regular search referral traffic in its analytics dashboards.

Publishers have found their own ways to track the impact of AI Overviews, using third-party tools like Semrush and Moz. Ziff Davis compared CTRs on top search queries from a year ago to similar queries that now include AI Overviews, and found that AI Overviews is triggered in “a little over 20%,” Ziff Davis CEO Vivek Shah said in its May earnings call.

“About a third of the time, we’re getting cited [in AI Overviews], and presumably citations in AIOs is a good thing. Because it comes with links. I think that [is], something to watch,” Shah added.

AI Overviews appear for about a third of search results related to Dotdash Meredith’s content, CEO Neil Vogel said in a recent earnings call. Dotdash Meredith is working to reduce its reliance on Google referral traffic, which now accounts for about a third of the publisher’s traffic, he added.

However, these calculations aren’t foolproof.

“We’re basically relying on third-party tools, and that can be hit or miss,” said one news publishing exec who asked to speak anonymously. While their news outlet hadn’t seen much of a change in Google search traffic or clickthroughs since the rollout of AI Overviews, they noted it was difficult to measure apples to apples, given the fluctuations in news cycles.

Publishers concerned AI Mode is the real threat to referral traffic 

News publishers in general seem to be more insulated from Google’s generative AI search features so far. Hard news queries don’t seem to trigger AI Overviews. Real-time information categories — such as news and sports — were the least affected by the increase in AI Overviews in Google Search results, according to a recent analysis by Semrush and Datos.

But that may not be the case for long. There’s a growing sense of unease among publishers around how Google’s AI Mode will amplify and accelerate the slow erosion of CTRs publishers have seen to date with AI Overviews.

AI Mode is an experimental feature for search that resembles the search experience of ChatGPT Search or Perplexity, letting users ask follow-up questions without leaving the page, and is powered by Google’s Gemini 2.0 AI model.

Although the feature remains in beta for now, some publishers with access to Google’s Search Labs incubator are testing it. And Google recently started testing it on its homepage.

One news publishing exec said recent tests they conducted in Google’s AI Mode showed the search engine was able to ingest breaking news content and summarize that information “much quicker” than AI Overviews.

In one test, they input some queries into AI Mode about a breaking news story they hadn’t published yet, and the search engine couldn’t answer questions about that story. Ten minutes after the news org published its story, AI Mode “had everything available to it” about that topic.

The news publishing exec said AI Mode is the real threat to publishers’ referral traffic, once it’s rolled out widely.

“AIOs is just a trial balloon to collect data for AI Mode,” they said.

Google did not respond to a request for comment before publishing time.