In February 2026, Perplexity disclosed that it was processing over 150 million queries per day — up from roughly 15 million a day a year prior. ChatGPT’s search feature, launched in late 2024, now handles an estimated 200+ million daily search-like queries. Google still processes approximately 8.5 billion searches per day. But for the first time in a quarter century, the trendline matters more than the absolute numbers: Google’s share of the informational search market is declining, and the decline is accelerating.

This is not a story about Google dying. It is a story about search being unbundled — different query types migrating to different tools, fragmenting a market that one company has dominated since the early 2000s.

The Query-Type Breakdown

The critical insight most coverage misses: “search” is not one market. It is at least five distinct use cases, and AI is winning some while losing others badly.

The pattern: AI search wins on queries where users want understanding. Google wins on queries where users want to transact with the real world — buy something, go somewhere, find a business. This is not a coincidence. It reflects a fundamental architectural difference: AI search is built for synthesis, Google is built for indexing and connecting users to the commercial web.

The Market Share Picture

Raw market share numbers do not tell the whole story, because they treat all queries as equal. A navigational query (“facebook.com”) and a 20-minute research conversation count the same. But the shift is real and measurable:

That 8.2% figure for AI search is almost certainly an undercount. It does not capture queries made directly in ChatGPT, Claude, or Gemini that users think of as “chat” but that functionally replace what would have been a Google search. Internal data from Anthropic and OpenAI suggests the real volume of search-replacing queries across all AI tools may be double the tracked number.

More telling than the absolute share is the demographic split. Among users 18-34, AI search tools account for an estimated 15-20% of information-seeking behavior. Among users 55+, it is under 3%. This is the classic adoption curve of a technology that will grow as younger cohorts age up.

Where AI Search Actually Wins: Specific Examples

Abstract comparisons are less useful than concrete examples. Here are five queries where AI search is measurably better:

“Explain the difference between RSUs and ISOs, and help me figure out which is better for my situation as an early employee at a Series B startup.” Perplexity or ChatGPT will walk you through the tax implications, ask about your strike price and vesting schedule, and give you a personalized analysis. Google gives you 10 Investopedia articles that each explain the basics without addressing your specific situation.

“I’m getting CORS errors in my Next.js app when calling my Express API on a different port. I’ve tried setting the headers manually but preflight requests are still failing.” Any AI chat tool will diagnose the specific misconfiguration, ask follow-up questions about your setup, and give you the exact middleware configuration you need. Google gives you a Stack Overflow post from 2019 that may or may not match your setup.

“Compare the climate, cost of living, healthcare quality, and tax burden of Portugal vs. Spain for a remote worker earning USD from a US employer.” AI search synthesizes this across multiple dimensions in one response. Google requires four or five separate searches and a spreadsheet.

“What’s the current scientific consensus on creatine supplementation for cognitive performance in sleep-deprived adults?” Perplexity will pull from recent meta-analyses with citations. Google gives you a mix of PubMed abstracts you need a PhD to parse and supplement company blog posts trying to sell you something.

“My 2019 Honda CR-V is making a clicking sound when turning left at low speed. What’s the most likely cause and what should I expect to pay for repair?” AI search will diagnose the likely CV joint issue, explain the repair, and give a price range. Google gives you forum posts from 2014 about a different model.

Where Google Still Wins: And Why

Google’s enduring advantages are infrastructure advantages, not intelligence advantages:

Local search is Google’s deepest moat. Google Maps has 200 million+ business listings with real-time hours, reviews, photos, and directions. No AI search tool has anything comparable. When you search “Italian restaurant near me open now with outdoor seating,” Google is not just better — it is the only viable option.

Breaking news depends on real-time web indexing. Google’s crawlers index new pages within minutes. AI search tools that rely on web retrieval are faster than they were — Perplexity typically surfaces breaking news within 15-30 minutes — but Google is still faster for the first hour of a developing story.

E-commerce is Google’s revenue engine. Product search with prices, availability, merchant ratings, and price history is deeply integrated into Google’s infrastructure. AI search tools can compare products in the abstract, but they cannot tell you that the specific item you want is in stock at the store three miles from your house for $47.99.

Navigation queries — which account for nearly 28% of all searches — are the bedrock of Google’s volume. People type “Gmail,” “YouTube,” and “Amazon” into Google billions of times per day. AI search adds zero value to these queries, and they are not going away.

Google’s AI Overviews: Cannibalizing Itself to Survive

Google’s response to the AI search threat has been AI Overviews — AI-generated summaries that appear above the traditional blue links. As of early 2026, AI Overviews appear on approximately 40% of informational queries in the US.

The strategy is rational: if users want AI-generated answers, give them AI-generated answers inside Google rather than losing them to Perplexity. But the data on the side effects is sobering.

A study by Authoritas in late 2025 found that queries with AI Overviews saw a 25-30% reduction in clicks to organic results. Ziptie, a search analytics firm, published data in March 2026 showing that the average click-through rate on the first organic result dropped from 28.5% to 19.7% on queries where AI Overviews appeared.

This is the classic innovator’s dilemma. Google is cannibalizing its own click-through rates — the foundation of its $175 billion/year ads business — to avoid losing users to competitors. The alternative (doing nothing) is worse. But the economic model of “AI answers at the top, ad-supported links below” is not yet proven to be as profitable as the old model.

The Publisher Problem: Who Pays for the Content AI Summarizes?

This is the most important long-term question in the AI search debate, and it does not have a good answer yet.

The web’s content ecosystem runs on a simple exchange: publishers create content, Google sends them traffic, publishers monetize that traffic with ads or subscriptions. AI search breaks this loop. When Perplexity synthesizes five articles into a single answer, those five publishers get a citation but not a click. Attribution without traffic is worth very little.

The numbers are stark. Data from Chartbeat shows that referral traffic from Google to news publishers declined 17% year-over-year in Q1 2026. Some of that is AI Overviews reducing clicks on Google itself. Some is users bypassing Google entirely for AI tools. Either way, the publishers who create the information that AI tools depend on are getting less traffic, less revenue, and less incentive to create quality content.

Perplexity has launched a publisher revenue-sharing program that pays participating publishers a share of advertising revenue when their content is cited. The program is small — reportedly generating less than $10 million in total publisher payments in its first year — but it is the first serious attempt at solving the economic problem.

The uncomfortable reality: if the content ecosystem degrades because publishers cannot monetize, the AI tools that depend on fresh, high-quality web content will also degrade. This is a tragedy-of-the-commons problem that the industry has not yet solved.

The AI Search Tools: A Quick Comparison

Perplexity remains the leader in dedicated AI search. Its citation system is the best in the market — every claim links to a specific source. The Pro tier ($20/month) adds more powerful models and deeper research capabilities. The free tier handles 5-10 research queries per day well.

ChatGPT Search benefits from OpenAI’s massive user base and tight integration with the ChatGPT interface. It is the default for users who are already in ChatGPT. Sources are cited but less granularly than Perplexity.

Google Gemini has the advantage of real-time access to Google’s index, which makes it the strongest for queries that need current information. The integration with Google Workspace (summarize this email thread, find that document) is a genuine differentiator.

You.com offers a research-focused AI search with a strong emphasis on developer and academic use cases. Its free tier is generous, and the “Research” mode provides multi-step analysis that competes with Perplexity Pro.

Claude (from Anthropic) is not a search engine, but many users treat it as one for analytical queries. It does not browse the web by default, but its deep reasoning on complex topics makes it the tool of choice when you need analysis, not just information retrieval.

What Happens Next

The most likely outcome is not “Google dies” or “AI search fizzles out.” It is permanent fragmentation. Informational and analytical queries will continue migrating to AI tools. Commercial, local, and navigational queries will remain with Google. A growing share of “search” will happen inside AI assistants (Siri, Alexa, Google Assistant) that users do not think of as search at all.

Google will remain the largest single player in search for years. But its monopoly over the act of looking something up on the internet — the behavior that has defined the web for 25 years — is over. The unbundling has begun, and it is not going to reverse.