What Google's Search Overhaul Means for Publishers

The Search Overhaul I Noticed at Google I/O

At Google I/O 2026 on May 19, Google announced a redesign of Search. It’s now easier to type long, complex questions, and you can search using photos, files, and videos as inputs. Google also previewed an information agent that keeps tabs on topics you care about. Plenty of pieces have already covered the whole of I/O, so I want to zoom in on the part of the announcement that caught my attention most: search itself.

As someone who writes for a living, what matters to me is where search results actually send readers. If people get their answer directly from Google, they may have less reason to visit the original source. Or they might click through anyway, looking for more detail. Either way, “search got better” and “traffic to content sites changed” are two different claims — worth examining together, but measured separately.

I’ve argued before that GEO/AEO deserves criticism. I was skeptical about how different these “optimize to get surfaced in AI answers” methodologies really are from old-fashioned SEO. I still think the question that matters isn’t exposure itself, but whether that exposure actually converts into visits, subscriptions, and purchases.

The Search Box, Follow-Up Questions, and the Information Agent

This announcement boils down to three key points.

First, it’s now easier to enter long questions and multiple kinds of material. The search box expands to fit the length of your query, and beyond text, you can now use images, files, videos, and Chrome tabs as input. Google said it’s beginning to roll out the new search box in every country and language where AI Mode is available.

Second, you can now continue with follow-up questions from an AI summary. If you ask another question within AI Overviews in your search results, it carries over into an AI Mode conversation while retaining the prior context. This doesn’t mean the two features have become identical — it means it’s now easier for users to move between them.

Third, Google previewed an information agent that finds changes matching your stated interests. For example, if you tell it the conditions you want in a home, it will monitor new listings and notify you when a match appears. Google announced it plans to make this available first to AI Pro and Ultra subscribers that summer.

Google said AI Mode’s monthly users have surpassed 1 billion, and that the number of queries has more than doubled every quarter since launch. These are usage figures the company disclosed itself. With longer queries, users can now pack in multiple requirements at once — for instance, specifying a travel destination’s distance, budget, and activities all together.

Even as search usage rises, whether clicks to external sites are rising along with it is a separate question that requires its own data.

The English draft looks accurate and complete — matches paragraph count, numbers, footnote, link, and image markers from the source with no distortions or omissions.

What the traffic data from news sites actually shows

Publishers running news sites and blogs have long met their readers through search. When a reader clicks through to an article, that’s an opportunity for an ad view or a subscription or purchase. That’s why traffic moving from search results to the original article has always been a key metric.

There’s research looking at how these clicks change when AI summaries appear in search.

Pew Research Center analyzed the March 2025 search histories of 900 US adults. On search pages with an AI summary, the share of visits where users clicked a regular search link was 8%; without an AI summary, it was 15%. The share clicking a source link inside the AI summary itself was a separate 1%. This is an observation based on search-page visits — not a controlled experiment with randomized conditions like search terms.

Chartbeat data cited in the Reuters Institute’s 2026 report found that Google search traffic to more than 2,500 news sites fell 33% in November 2025 compared to November 2024. In the US, the drop was 38%, and Google Discover¹ traffic fell 21% across the full sample. This reflects traffic changes at the sites surveyed, and the institute itself noted it’s unclear how much of this is attributable to AI Overviews.

Differences by site size were also reported. In Chartbeat’s comparison of search traffic by site size, small sites saw a 60% decline versus 22% for large sites. “Small” was defined as 1,000 to 10,000 average daily pageviews, “large” as 100,000 or more. This alone doesn’t prove that size is the cause of the decline, but looking only at the overall average risks missing sites that are struggling badly. Chartbeat data

The reasons behind declining traffic — and each outlet’s capacity to respond — vary by publisher. You have to weigh changes in search rankings, the topics a site covers, seasonal and news-cycle demand, and readers’ visiting habits together. It’s worth being cautious about attributing layoffs or closures to AI search alone.

What concerns me is the situation facing outlets heavily dependent on search traffic: they risk losing the chance to meet new readers altogether. This is especially difficult for sites with few direct visitors, since they have little cushion against shifts on external platforms.

The Reuters Institute surveyed 280 media leaders across 51 countries. Of these, the 268 who answered questions about search outlook expected search traffic to fall by an average of 43% over the next three years. To be clear, this is respondents’ forecast, not a decline that has actually happened yet.

Traffic arriving from AI chatbots also needs to be tracked. Even if it’s growing fast, if the starting base is small, it may not be enough to offset what’s being lost from search. What matters isn’t just the growth rate but the actual number of visits and the share they represent of total traffic.

Over the long run, the revenue of those producing original content matters too. To keep producing the articles and material that AI draws on, publishers need to be able to cover the cost of reporting and production. I think the convenience of search and the sustainability of content production need to be discussed together.

Search Ads and Publisher Ads Are Different Businesses

We need to distinguish between ads attached to Google search results and ads that appear on external sites. Search ad revenue isn’t structured so that all of it gets split with content publishers. Ads on external sites run through a separate business — AdSense, Ad Manager, and the like.

If a user never lands on the original site, that site never gets the chance to show an ad. Google, on the other hand, can show ads right within the search screen itself. Because of this difference, Google’s search revenue growth and publishers’ earnings don’t necessarily move in the same direction.

Alphabet’s Q1 2026 Google Search & other revenue was about $60.4 billion, up 19% year-over-year. Google Network revenue was about $6.971 billion, down roughly 4%. These figures come from the official earnings release. Network revenue is revenue Google recognizes — it isn’t the same as the amount actually paid out to publishers.

These results show that search ads and external network ads grew at different rates. But that doesn’t mean we can treat the gap as money that shifted from publishers to Google, or attribute the entire difference to AI search.

What to Check in Korea’s Content Business

In Korea, you have to watch both Google and Naver together. Depending on your industry and audience, either search engine could be the more important source of traffic.

Before looking at overall market share, check the search visits actually landing on your own site. It helps to separate out where the shift happened — Google or Naver — and which posts or products saw traffic drop, since that shapes how you respond.

Naver also runs AI briefings in its search results. Whether users stop at the summary or click through to a blog or shop makes a real difference to content operators. You can’t just apply Google’s global numbers to a Korean site.

Businesses that meet customers through search — hospitals, academies, online shops — need to ask the same question: when visit counts shift, do inquiries, bookings, and purchases shift with them?

I think it’s worth building up touchpoints beyond search in parallel. An opt-in email newsletter and direct visits can be ways to reconnect with your audience. YouTube and social channels are useful too, but keep in mind that they’re also subject to their own platforms’ policies.

Search traffic will keep mattering. But you get a clearer read on how your business is changing when you measure it alongside returning readers and subscription/purchase conversions.

Oswarld’s Lens

In building GTM strategy, I’ve repeatedly seen how a platform’s operating changes reshape the terms of its partner businesses.

This time, too, I don’t take Google’s rising usage numbers alone as good news for content publishers. Alongside checking whether it’s convenient for readers, I want to verify whether traffic or revenue actually flows back to the people who created the original content.

The quality of the answers themselves needs separate scrutiny. In a Tow Center study using 1,600 questions designed to make AI tools locate the source of news excerpts in 2025, eight AI search tools gave wrong answers a combined 60%+ of the time. In a study by the EBU and BBC evaluating more than 3,000 news answers from four AI tools that same year, 45% had at least one serious problem — including issues with sourcing and context. These are two separate studies measuring different tasks, not the current error rate for Google’s entire new 2026 search feature.

Even when I see an AI-summarized answer first, I should still be able to click through to the source and check the evidence and context. Alongside search becoming more convenient, I care just as much about whether I still get the chance to compare other sources and judge for myself.

Closing

As more searches deliver instant answers, publishers need to track how their traffic patterns are actually changing. It’s better to start by looking at what’s happening to your own content and readers, rather than fixating on the industry-wide decline rate.

Search visibility, site visits, and subscription or purchase conversions are three separate metrics. Tracking whether people who read the original piece come back again gives you a more concrete basis for deciding which touchpoint needs work.

I’m also examining whether this newsletter gives readers who first arrive through search a reason to come back. I think that’s a question I’ll need to keep answering, right alongside the question of search rankings.

References & Further Reading

Primary sources

• Klaudia Jaźwińska, Aisvarya Chandrasekar, “AI Search Has a Citation Problem”, Columbia Journalism Review (Tow Center for Digital Journalism), March 2025. : This is the core study that tested citation accuracy across 8 AI search tools with 1,600 queries.
• EBU & BBC, “News Integrity in AI Assistants”, October 2025. : A large-scale study on AI news accuracy involving 22 public broadcasters across 18 countries.
• Reuters Institute/Chartbeat, “Journalism, Media, and Technology Trends and Predictions 2026”, January 2026. : This report covers changes in search referral traffic across more than 2,500 sites, comparing November 2024 to November 2025, along with media leaders’ outlook.
• Pew Research Center, AI Overviews CTR study, July 2025. : An observational study comparing link clicks with and without AI summaries, based on the March 2025 search histories of 900 American adults.

Background

• Google official blog, “Search at I/O 2026”, May 2026. : The official announcement for this year’s Google I/O search updates.
• AdExchanger, “The AI Search Reckoning Is Dismantling Open Web Traffic”, January 2026. : A comprehensive look at how AI search is reshaping the publisher ecosystem.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Google Discover: A feed that automatically recommends content based on user interests, shown either in the Google app or on Android home screens. Because articles can appear without any search action, it’s a crucial traffic channel for publishers. ↩