Google Killed the Blue Link: AI Overviews Now Answer 43% of All Searches

Google killed the blue link. The data shows you probably didn't notice until now.

There's a number that should make every website owner, every blogger, and every business that depends on search traffic stop what they're doing. Google's AI Overviews — those AI-generated summaries that appear at the top of your search results — now show up in 43% of all searches. A year ago, that number was 15%.

In twelve months, Google quietly transformed from a search engine that sent you to other websites into an answer engine that keeps you on Google.

And here's the thing: you've probably been experiencing this shift without fully registering it. You search for something, you get an answer at the top of the page, and you leave. No clicking. No visiting the source. The answer was just... there.

That's not a bug. That's the plan.

The numbers are staggering

Market intelligence firm Similarweb published the data in late July 2026, and it paints a picture of a fundamental restructuring of how humans find information online.

AI Overviews went from 15% of searches to 43%. Google's AI Mode — the more conversational follow-up layer you can drop into after seeing an Overview — saw visits jump from 126 million in June 2025 to 279 million by May 2026. More than doubled in a year.

But the most revealing data point isn't about volume. It's about behavior. The average length of Google search queries has been steadily rising. People aren't typing "best laptop 2026" anymore. They're typing "what's the best laptop for video editing under $1500 with at least 32GB of RAM." They've stopped optimizing for a machine that matches keywords and started talking to a machine that understands language.

This is what the transition from search-as-directory to search-as-conversation looks like in the data. And it happened faster than almost anyone predicted.

The web's original sin, repeated

Here's the uncomfortable part. Every AI Overview that answers your question without you clicking through is a website that didn't get a visit. A publisher that didn't get ad revenue. A small business that didn't get a potential customer.

The old deal was simple: Google indexed your content, showed a snippet, and sent users your way. You got traffic. Google got the most useful search engine. Everyone won.

The new deal is: Google indexes your content, uses it to generate an answer, shows that answer to the user, and the user never leaves Google. Google gets the engagement. You get nothing.

This is actually the same fight that's been simmering since Google introduced Featured Snippets in 2014. The difference is scale and finality. A Featured Snippet still felt like a preview — a teaser that often drove clicks. AI Overviews feel like the answer. Why would you click through to a website when Google already told you what it says?

Similarweb flagged this trend last year, noting how devastating it was for news publishers specifically. The situation hasn't improved. Some categories got hit harder than others — recipe sites, product reviews, "how to" guides. These are content categories that existed almost entirely to capture search traffic, and they're watching their business model evaporate in real time.

The internet fights back

The broader industry is starting to push back, and the opening moves look like the start of a much bigger fight.

Cloudflare — the infrastructure company that sits in front of roughly 20% of the entire internet — launched tools that let website owners block AI crawlers from their sites entirely. Unless those AI companies pay for access through Cloudflare's marketplace. There's now a toll booth between AI companies and the content they need to generate answers.

Think about what that means. A company that powers a huge chunk of the internet's plumbing is essentially saying: if you want to scrape our customers' content to train your AI, you have to pay. The internet is unionizing against the AI companies. Whether it works is an open question, but the fact that Cloudflare is positioning itself as the broker tells you something about where this is going.

Content creators are realizing that AI search might be the most aggressive intermediary in the history of the internet — more extractive than Google's old snippet problem, more systematic than social media's algorithmic feed. At least with social media, you could post your content directly and build an audience. With AI search, your content gets absorbed, summarized, and served back without attribution that drives action.

The citation numbers tell the story. Citations in AI responses have risen more than fivefold over the past year, which sounds great until you see the baseline: just 6.8% of US ChatGPT desktop queries included citations as of May 2026. Five times almost nothing is still almost nothing.

There's a sliver of good news. After ChatGPT's May 7 search update, the proportion of visits actually landing on webpages jumped from 25% to nearly 60%. People do click when the links are prominent and relevant. But that's one platform's UX decision, not a structural shift. The larger trend remains unchanged.

The query revolution

Here's something that doesn't get enough attention in these discussions: the way we search is changing at a fundamental level, and that change has implications far beyond Google's ad revenue.

When you type "pizza near me," you're speaking search engine. When you type "I'm looking for a good pizza place that's open late and has outdoor seating near downtown," you're speaking to an AI. The shift from keyword-speak to natural-speak is happening in real time, and it's happening because AI Overviews and AI search actually understand the second query.

This matters because it changes what "being findable" means. Traditional SEO was built on keyword optimization — figuring out which terms people type, then structuring your content to match those terms. Keyword density, meta tags, backlink profiles, H1 tags. The entire multi-billion-dollar SEO industry existed to help websites speak search engine.

AI search doesn't care about your keyword density. It cares about whether your content actually answers the question. Which means the entire SEO industry is being forced to evolve from gaming keywords to — and I can't believe I'm saying this — writing genuinely useful content.

There's something darkly funny about that. We spent two decades engineering content for algorithms, stuffing pages with keywords and building link farms, and the end result of AI search is that the algorithm now rewards quality. The robots are making us be better humans.

Whether that actually plays out in practice is another question. Early signs suggest that AI search creates its own optimization games — structured data for AI parsers, content designed to be quotable in summaries, "LLM SEO" replacing traditional SEO. The names change. The gaming instinct stays the same. But at least the new game requires you to be useful, not just keyword-optimized.

A structural shift, not a feature update

Let me zoom out for a second, because it's easy to get lost in the numbers and miss the shape of what's happening.

The web had a specific architecture for twenty-five years. You published content. Search engines indexed it. Users searched, clicked, and arrived at your door. The search engine was a directory — incredibly powerful, yes, but fundamentally a middleman that directed traffic.

AI search breaks that architecture. The AI doesn't just index and route. It reads, synthesizes, and answers. It's not a directory anymore. It's a competitor to every website it summarizes. When Google answers "how to fix a leaky faucet" with an AI Overview, it's competing directly with This Old House, Family Handyman, and every plumbing blog that used to get that click.

This is why the Cloudflare paywall matters. This is why publishers are panicking. This is why Google is being careful about rollout speed despite the aggressive numbers. They know they're burning bridges with the content ecosystem that made them dominant, and they're doing it anyway because the alternative — letting ChatGPT or Perplexity eat their search business — is worse.

Google didn't become a $2 trillion company by being sentimental. They saw the AI search writing on the wall, and they moved. The 43% number isn't a mistake or an experiment. It's a commitment.

Why this matters for you

If you're reading this, you're probably someone who cares about AI and automation. Maybe you run a business. Maybe you build things. Maybe you just want to understand where this is all going.

Here's the honest takeaway: the web is becoming an AI-mediated experience. Not might become. Is becoming. When 43% of Google searches show AI-generated answers and AI Mode is doubling year over year, that's not a trend — that's a new baseline.

And this raises a question that I think is going to define the next few years: whose AI is mediating your experience of the internet?

Right now, for most people, the answer is Google's. Or OpenAI's. Or Anthropic's. You're using their AI, on their terms, with their priorities baked in. Their AI decides what to summarize. Their AI decides what to cite. Their AI decides what to surface and what to bury. When you search, you're not seeing the web. You're seeing their model's interpretation of the web.

There's an alternative, and it's getting more practical every day. Open-source AI models — Qwen, DeepSeek, GLM, Kimi, MiniMax — are now capable enough to handle real tasks. You can run them locally, on your own machine, with your own files and your own browser. Not asking a cloud AI to summarize the web for you, but having your own AI that browses, reads, and acts on your behalf.

This isn't theoretical. Desktop AI assistants exist today that can open websites, run terminal commands, read your files, and automate workflows — all powered by open-source models that you control. The frontier gap between proprietary and open models has narrowed dramatically. Kimi K3, a 2.8 trillion parameter open-weights model released in July, landed within striking distance of GPT-5.6 on major benchmarks. DeepSeek V4 matched frontier coding models at a fraction of the cost. The idea that you need a cloud API to get capable AI is increasingly outdated.

When the AI layer becomes how you experience the internet, owning that layer matters. Not renting it from a company whose business model depends on keeping you on their platform and showing you their answers.

The deeper question

Google's AI Overviews data is interesting not because it's surprising — we all felt this coming — but because of how fast it happened. Fifteen percent to forty-three percent in a year. AI Mode doubling. Search queries getting longer and more conversational.

The web isn't dying. But the web as a collection of websites you visit is being replaced by the web as a conversation you have with an AI. That conversation might happen on Google. It might happen on ChatGPT. Or — and this is the part I find most interesting — it might happen with an AI that lives on your own computer, using open-source models, browsing the web on your terms.

The next time you search for something and get a perfect answer without clicking a single link, take a second to think about what just happened. An AI read the internet for you, decided what mattered, and gave you the summary. That's incredibly convenient. It's also a profound shift in who controls your access to information.

The companies mediating that access aren't charities. They have goals that aren't your goals. Google wants you on Google. OpenAI wants you on ChatGPT. Both will shape your AI-mediated web experience to serve their interests. The question isn't whether AI will mediate your experience of the internet — that's already happening, at 43% and climbing. The question is whether you'll use their AI or your own.

---

If that idea resonates, there are tools built around exactly this philosophy. CopperRiver is a desktop AI assistant for Mac that uses open-source models to browse websites, run commands, read files, and automate tasks — all locally, all on your terms. Plans start at $9/month. No intermediary required.