Google AI Overviews Are Eating Publisher Traffic

Google AI Overviews have materially reduced organic click-through rates for a broad range of publishers since their wider rollout in 2024. When Google answers the question directly in the search results page, a meaningful share of users never click through to the source. That is not a hypothesis. It is playing out across traffic data for news sites, content publishers, and informational websites right now.

The scale of the effect varies by query type, industry, and how well a publisher’s content is structured. But the direction of travel is consistent: zero-click search is growing, and AI Overviews are accelerating it.

What Are Google AI Overviews and Why Do They Matter for Traffic?

AI Overviews are the AI-generated summary boxes that appear at the top of Google search results for a growing range of queries. They synthesise information from multiple sources and present a direct answer, with source citations shown as small cards beneath the summary. Google began rolling them out broadly in the United States in May 2024, with wider international expansion following through the second half of the year.

The commercial logic for Google is straightforward. Users get faster answers and stay within the Google ecosystem longer. The cost is borne by the publishers whose content trains and feeds the summaries but who receive less traffic in return.

I have watched the search landscape shift in waves over 20 years. Featured snippets reduced clicks for a lot of informational queries. Knowledge panels did the same for branded searches. Each time, the industry debated whether it was a real problem or just noise. AI Overviews feel categorically different in scale, because the technology can now synthesise and present complex answers, not just pull a single sentence from a page.

If you want a broader view of how AI is reshaping search and marketing strategy, the AI Marketing hub at The Marketing Juice covers the commercial implications across channels and business functions.

What Does the Traffic Data Actually Show?

The honest answer is that the data picture is messy, and anyone presenting clean universal numbers should be treated with scepticism. The effect varies significantly by query type, vertical, and content format.

What the data broadly shows is this. Queries where AI Overviews appear consistently see lower click-through rates than equivalent queries without them. The drop is most pronounced for informational queries, particularly those structured as questions. How-to content, definition queries, and listicle-style content that answers a single clear question are the most exposed categories.

Publishers in the health, finance, and general knowledge space have reported the steepest declines. These are categories where Google’s AI can produce a credible-looking answer from existing published material. The irony is that the publishers who invested most heavily in SEO-optimised informational content over the past decade built the training data that now competes with them.

Transactional queries, local searches, and content requiring genuine expertise or current reporting are less affected. If someone is searching for a specific product, a local service, or breaking news, AI Overviews are less likely to appear or less likely to satisfy the intent fully.

Tools like those covered in Semrush’s AI SEO guidance are starting to help publishers identify which of their queries are most exposed to AI Overview displacement, which is useful triage work even if it does not solve the underlying problem.

Is Being Cited in an AI Overview Worth Anything?

This is the question publishers are asking most urgently, and the answer is more complicated than Google’s framing suggests.

Google has positioned AI Overview citations as a benefit for publishers. The argument is that appearing as a cited source increases brand visibility and can drive qualified traffic from users who want to read more. There is some truth in this for certain query types. A user researching a complex financial decision or a medical question may well click through to a cited source for depth they cannot get from a summary.

But for the majority of informational queries, the citation cards are small, positioned below the fold of the AI summary, and clicked by a fraction of users. The user got what they needed from the summary. The citation is a courtesy, not a traffic driver.

I spent several years managing agency relationships with major publishers, and the pattern I saw repeatedly was that visibility metrics and business metrics diverged sharply when you looked closely. Impressions went up. Revenue did not follow. The same dynamic is playing out here. Being cited inside an AI Overview is a form of visibility. It is not the same as traffic, and traffic is not the same as revenue.

The Ahrefs perspective on AI and SEO is worth reviewing for a technical view of how citation patterns in AI Overviews work and what factors influence whether your content gets pulled.

Which Content Categories Are Most Exposed?

Not all content is equally at risk. Understanding the exposure profile of your content library is the first step toward a rational response.

The most exposed categories share a common characteristic: they answer a single, well-defined question that AI can summarise accurately without needing to read the full article. Think “what is a P&L statement”, “how long does SEO take to work”, or “best time to post on Instagram”. These are queries where the informational need can be met in three sentences. Google’s AI can produce those three sentences. The click never happens.

Less exposed categories include original reporting, where the content contains information AI does not have access to. Opinion and analysis, where the value is a specific person’s perspective rather than a factual answer. Technical depth, where the query requires nuance that a summary cannot adequately compress. And transactional intent, where the user needs to complete an action rather than receive information.

When I was running iProspect, we grew the agency from around 20 people to over 100 across several years. A significant part of that growth came from helping clients understand which of their marketing activities were genuinely driving commercial outcomes versus which were generating metrics that looked good but changed nothing. The same analytical discipline applies here. If your content exists primarily to rank for informational queries, you need to be honest about what that traffic was actually worth and whether it was ever as durable as it looked.

What Should Publishers Actually Do?

There are two broad categories of response: tactical adjustments to how you produce content, and structural changes to how your business relates to search traffic.

On the tactical side, the most defensible content formats are those that AI cannot easily replicate. Original data and research that does not exist anywhere else. Interviews and quotes from named individuals with genuine expertise. Detailed case studies with specific numbers. Opinions that are clearly attributed to a real person with a track record. These formats either require access AI does not have, or they derive their value from attribution in a way that a summary cannot replace.

Structuring content for AI Overview citation is a legitimate short-term tactic. Clear, well-structured answers to specific questions, with proper schema markup, do improve the chances of being cited. Moz’s work on AI content briefs covers the structural elements that make content more likely to be picked up by AI systems. But optimising for citation is not the same as solving the traffic problem. It is managing the decline, not reversing it.

The structural response is harder but more important. Publishers who built their audience on Google search traffic have a dependency that this shift is exposing. Building direct relationships, through email lists, subscriptions, community, or social channels that you own, reduces that dependency. It is not a fast fix. But every publisher who ignored this advice for the past decade is now in a more precarious position than those who did not.

I had a conversation with a client a few years ago who was resistant to investing in email because “Google sends us all the traffic we need”. That logic has an obvious fragility to it. Google’s priorities have never aligned perfectly with publisher interests, and the introduction of AI Overviews is simply the latest, most visible expression of that misalignment.

For teams thinking about how to adapt their SEO and content workflows in light of AI, Moz’s overview of AI tools for productivity and Ahrefs’ webinar on AI tools for SEO teams are both worth the time.

How Should You Measure the Real Impact on Your Business?

Measurement is where most publishers are currently making mistakes, and it is worth being direct about this.

The instinct is to look at overall organic traffic and compare year-on-year. This is too blunt an instrument. AI Overviews do not affect all queries equally. A blended traffic number will obscure which parts of your content library are under genuine pressure and which are not. You need to segment by query type, by content category, and by intent before you can draw any useful conclusions.

Google Search Console is the starting point. Look at impressions and clicks separately. If impressions are holding but clicks are falling, that is a strong signal that AI Overviews are appearing for those queries. A widening gap between impressions and clicks, on queries that previously converted well, is the diagnostic signal to look for.

One thing I have always believed, and it comes from seeing measurement done badly at scale across hundreds of clients, is that the most dangerous number in marketing is one that looks stable while the underlying business is changing. Aggregate traffic figures can mask significant structural shifts for months. By the time the number moves, the problem is already well established.

Beyond traffic, measure revenue per session and conversion rates for organic traffic segments. If the traffic you are losing to AI Overviews was never converting to anything commercial, its loss matters less than it appears. If it was a meaningful part of your acquisition funnel, you need to know that clearly and quickly.

The Broader Question About Google’s Relationship with Publishers

There is a harder conversation underneath all of this that the industry is having in fragments rather than directly.

Google’s AI Overviews are built on content that publishers created, often at significant cost. The system extracts value from that content and returns a fraction of it as citation visibility. Publishers have limited recourse. Opting out of Google entirely is not a realistic option for most. Negotiating collectively is possible in theory but has not produced meaningful results in practice. Legal and regulatory challenges are ongoing in various jurisdictions but slow-moving.

What this means practically is that publishers cannot wait for an external solution. The businesses that will handle this period most successfully are those that treat Google search as one channel among several, rather than as a foundation. That requires investment in direct audience relationships, in content that has value beyond search discovery, and in distribution that does not depend on a single platform’s algorithm.

None of this is new advice. The dependency on Google has been a known structural risk for publishers for over a decade. AI Overviews have simply made the cost of that dependency more visible and more immediate.

There is more on how AI is reshaping marketing strategy across channels in the AI Marketing section of The Marketing Juice, including practical coverage of where these tools create genuine commercial value and where they generate noise.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions