Google processed an estimated 5.9 trillion searches in 2025, up 18% year-over-year according to eMarketer and StatCounter data. That is 16.4 billion searches per day, 189,815 per second. Google’s search advertising revenue hit $198 billion in 2024, up from $175 billion in 2023. Alphabet’s Q4 2025 earnings showed search revenue growing faster than any quarter since 2021.

These are not the numbers of a dying platform.

Yet the narrative persists: SEO is dead. Google is finished. AI chatbots have replaced search. The claims range from dramatic (“25% decline in organic traffic”) to apocalyptic (“50% decline by 2027”). They appear in marketing blogs, VC newsletters, conference keynotes, and LinkedIn posts with thousands of likes.

The claims are wrong. Not slightly wrong. Wrong by an order of magnitude. And the reasons they are wrong tell you more about the state of digital marketing analysis than about the state of search.

What the Largest Dataset Actually Shows

Graphite, a search optimization firm, partnered with Similarweb to analyze organic search traffic across 40,000 of the largest US websites. The dataset compared monthly visits from all search engines from February 2024 through November 2025. Similarweb’s estimates combine opt-in user panels, ISP partnerships, mobile carrier data, and direct site measurement, covering hundreds of millions of devices across desktop and mobile.

The result: organic search traffic declined 2.5% year-over-year. Not 25%. Not 50%. Two and a half percent.

This is consistent with Google’s own statement from August 2025 that total organic click volume from Google Search to websites had been “relatively stable year-over-year.” It is also consistent with the observable fact that Google’s search ad revenue grew 24% in the same period. Advertisers do not increase spend on a platform losing a quarter of its traffic.

The 2.5% decline is real. It is measurable. It is worth understanding. But it is a rounding error compared to the catastrophic numbers circulating in marketing media.

Where the Fake Numbers Come From

The inflated decline figures trace to three sources: surveys, small sample sizes, and anecdotal extrapolation. Each is methodologically flawed in specific, documentable ways.

Surveys. Multiple research firms published studies claiming 25%+ declines in search usage and proportional increases in LLM adoption. These studies used self-reported surveys asking questions like “Do you use ChatGPT instead of Google?” The problem: human memory is unreliable for quantitative recall. A 1987 study published in the Journal of Verbal Learning and Verbal Behavior showed that when asked to quantify past behaviors, people “resort to inferences that use partial information from memory to construct a numeric answer” rather than recalling actual counts.

Survey design introduces additional bias. The question “Do you use ChatGPT to search for products?” leads respondents to consider LLMs and asks a hypothetical. The question “Recall the last time you purchased a product. How did you find it?” asks about real behavior without suggesting an answer. The first form inflates LLM usage estimates. Most published surveys used the first form or something equivalent.

There is also selection bias. Surveys about AI adoption attract early adopters who use AI tools more than the general population. A survey about “changes in search behavior” self-selects for people who have changed their search behavior. The people who still use Google exactly as they did in 2023 do not click on surveys about search disruption.

Small samples. Several widely cited analyses examined 10 to 20 websites and extrapolated to the entire web. When one study found that HubSpot’s organic traffic declined significantly, the finding became a headline about the death of SEO. HubSpot’s traffic patterns are real but not representative. A sample of 19 sites, some of which actually increased in traffic, cannot support conclusions about 40,000 sites.

Anecdotal extrapolation. “I stopped using Google.” “My students all use ChatGPT.” “Nobody I know searches anymore.” These statements reflect three documented cognitive biases. The availability heuristic (people estimate frequency based on how easily examples come to mind). The false consensus effect (people overestimate how much others share their behavior). And egocentric bias (people overweight their own perspective). A VC who uses Perplexity for research is not a representative sample of the 5 billion daily Google users.

Google Search Volume Is Actually Increasing

The “SEO is dead” narrative requires Google search volume to be declining. It is not.

Google processed approximately 5 trillion searches in 2024 and an estimated 5.9 trillion in 2025. The 18% year-over-year increase reflects AI-assisted query expansion (users asking more follow-up questions within search sessions), deeper mobile penetration in Southeast Asia and Sub-Saharan Africa, and the integration of multimodal search inputs including image, voice, and video prompts via Google Lens.

Google Lens alone handles 12 to 20 billion visual searches per month, with over 3 billion active users. Voice search accounts for 20% of Google app queries. “Near me” searches grew 900% in two years. These are not categories that existed a decade ago. Search is expanding into modalities that the “SEO is dead” narrative does not account for because the narrative defines search as “typing words into a box.”

Traffic to Google itself also increased. Similarweb data shows traffic to Google grew 0.8% in 2025, and visitors increased 1.4% comparing Q4 2025 to Q4 2024. Traffic to search engines as a whole grew 0.4% in 2025. These numbers are small, but they are positive. Search engine usage is flat to slightly up, not collapsing.

What About ChatGPT? Isn’t Everyone Switching?

ChatGPT has 800 million weekly active users and processes 2.5 billion daily prompts. These are real numbers. They are also not substitutes for Google searches.

The key distinction: ChatGPT created new demand for AI interactions. It did not simply redirect existing search queries. People use ChatGPT for tasks Google was never good at: drafting emails, brainstorming ideas, coding assistance, explaining concepts in conversation, role-playing, creating content. Some of those tasks overlap with informational search queries, but the overlap is partial, not total.

The data confirms this. If 800 million people had switched from Google to ChatGPT for search, Google’s 5 billion daily users would show a 16% decline in activity. Instead, Google’s daily search volume increased. The most parsimonious explanation: ChatGPT expanded the total pool of information-seeking behavior rather than cannibalizing Google’s share of a fixed pool.

Google’s global search market share did decline, from 92.9% in 2023 to 89.6% in mid-2025. That 3.3 percentage point drop over two years is the largest in a decade. It is real. But 89.6% of 5.9 trillion searches is still 5.3 trillion searches. Google lost share and gained volume simultaneously. This is possible because the total market grew.

The Zero-Click Question

Approximately 60% of Google searches now end without a click to any external website. This statistic is real and frequently cited as evidence that Google is “stealing” traffic from publishers.

The math deserves scrutiny. If Google handles 5.9 trillion searches per year and 40% result in clicks, that is 2.36 trillion clicks to external websites per year. In 2015, Google handled roughly 2.5 trillion annual searches with a higher click-through rate (estimated 65-70%), producing approximately 1.6 to 1.75 trillion clicks. More searches with a lower click rate still produce more absolute clicks than fewer searches with a higher click rate.

Many zero-click searches are navigational (the user types “facebook” and clicks the first result, which counts as a “search” but was never going to generate exploratory clicks). Others are answered by the Knowledge Panel (“What time is it in Tokyo?”). These queries were never generating meaningful organic traffic.

The honest concern is about informational queries where AI Overviews provide a sufficient answer without requiring a click. That is a real dynamic. But it is narrower than “60% of searches produce zero clicks” implies.

AI Overviews: The Real Impact

Google’s AI Overviews do reduce click-through rates to organic results when they appear. The Graphite/Similarweb analysis found a 35% reduction in organic CTR on queries where AI Overviews are present.

Two critical caveats. First, AI Overviews appear on roughly 30% of queries, not all queries. Other estimates put the trigger rate at 13% to 30% depending on query type (informational queries trigger them most, transactional queries least). Second, in 80% of cases where AI Overviews now appear, Google previously showed Featured Snippets, which also reduced CTR. The net new impact of AI Overviews, beyond what Featured Snippets already caused, is smaller than the headline number suggests.

The category-level data supports this. News, health, cooking, and entertainment sites saw organic traffic declines exceeding 10%. These are precisely the categories where AI Overviews provide the most complete answers (recipes, symptom descriptions, entertainment schedules). Meanwhile, shopping, clothing, and marketplace sites saw traffic increases. Transactional intent, where users want to buy something, is largely unaffected because AI Overviews cannot complete a purchase.

The Size Distribution Matters

The Graphite analysis revealed a pattern that the “SEO is dead” narrative misses entirely. The top 10 largest sites by traffic actually increased organic search traffic by 1.6% year-over-year. Sites ranked in the top 50 were roughly flat. Sites ranked 100 to 10,000 experienced the largest declines. Smaller sites (ranked 10,000 to 40,000) were mixed, with some increasing.

This is not a uniform decline across the web. It is a redistribution. Google’s algorithm changes and AI features are concentrating traffic among the largest, most authoritative domains while squeezing the middle tier. Whether you call this an SEO problem depends on where you sit. For the top 10 sites, SEO is delivering more traffic than last year. For a mid-tier content site that relied on informational queries now answered by AI Overviews, the decline is real and painful.

This distinction matters for strategy. The response to “SEO is dead” is different from the response to “mid-tier informational content sites are losing traffic to AI Overviews.” The first is false. The second is true, specific, and actionable.

Follow the Money

Google’s search advertising revenue tells the clearest story. Advertisers spent $198 billion on Google Search ads in 2024, a 24% increase over 2023. The projected figure for 2026 is $198.4 billion according to eMarketer’s January 2026 forecast, with continued growth driven by AI-powered ad formats and Performance Max campaigns expanding to 14 new international markets.

Advertisers are sophisticated buyers. They measure return on ad spend with precision. If organic traffic were declining 25%, search ad performance would deteriorate (fewer users means fewer impressions means fewer clicks means worse ROI). Instead, the average Google Ads click-through rate across all industries rose to 3.54% in 2026, up from 3.17% in 2025. Legal services hit 6.98% CTR. E-commerce hit 5.12%.

When $198 billion in annual spend is increasing and CTR is improving, the underlying platform is not dying. The money is the most reliable signal because it is the hardest to fake.

What Is Actually Changing

SEO is not dead. But it is changing in specific, measurable ways that require specific responses.

Informational queries that can be fully answered in a paragraph are losing clicks to AI Overviews. Sites that relied on thin informational content (“What is X?” articles with 300 words of padding) are losing traffic. This is not the death of SEO. It is the death of low-value informational content that added nothing beyond what a search result snippet already provided.

Transactional, navigational, and complex informational queries (those requiring multiple perspectives, recent data, or domain expertise) are unaffected or growing. Technical content that explains mechanisms, original research and analysis, product reviews with first-hand testing, and content that requires trust (medical, legal, financial) are areas where organic search still delivers and AI Overviews cannot fully substitute.

Google’s market share is declining slowly (3.3 points over two years) as Bing’s AI integration and ChatGPT capture some search-adjacent queries. This is the first meaningful competitive pressure Google has faced in a decade. It is not an existential threat at 89.6% market share, but it is a trend worth watching.

The distribution of organic traffic is shifting toward larger, more authoritative domains. Mid-tier sites need to either build genuine authority (E-E-A-T), specialize in niches where depth beats breadth, or accept that some informational traffic is gone permanently. Original reporting on breaking events and deep technical analysis are examples of content that AI Overviews cannot replicate because they require primary sources, timing, and editorial judgment.

The Bottom Line

The “SEO is dead” narrative is built on surveys that misuse methodology, sample sizes too small to support conclusions, cognitive biases that make individual experiences feel like universal trends, and a conflation of “search is changing” with “search is dying.”

The data says: 5.9 trillion annual searches. 2.5% organic traffic decline. $198 billion in ad revenue. 89.6% market share. 0.8% increase in traffic to Google.

Search is not dying. It is the largest, most measurable, most commercially valuable information distribution channel on the internet. It is changing in ways that reward depth, authority, and originality while punishing thin content and commodity information. For publishers who build content worth ranking, SEO in 2026 is not dead. It is the competition that got harder.