✅ THIS PART IS LEGIT · REAL RESEARCH · REAL LINKS · HONEST VERDICTS · BY DOPAMINEKART
A person typing on a laptop keyboard, real photograph.
✓ STATE OF PLAY · AI SEARCH

AI Search Engines, State of Play 2026: Who's Winning, and Who's Paying for It

AntCode
Image: Wikimedia Commons (CC0) — photo by Alejandro Escamilla, via Unsplash.
🧭

Quick take

  • Google runs two separate AI search products at once — AI Overviews, the inline summary bolted onto classic search, and AI Mode, a standalone, fully AI-generated results page — and together they now reach billions of monthly users.
  • ChatGPT Search, launched October 2024, inherited the scale of the world's most-used chatbot. OpenAI put weekly ChatGPT usage at roughly 900 million in February 2026, with later reporting suggesting it's approaching 1 billion.
  • Perplexity is far smaller by raw user count — tens of millions of monthly users, not billions — but it pioneered the "answer engine" format: cited sources, no ad-supported results page, and now a browser (Comet) built around it.
  • Underneath all of it: independent research from SparkToro and Pew Research both find that AI-generated answers are measurably pulling clicks away from the websites those answers are summarizing.

Typing a question into a box and getting back ten blue links used to be the entire product. That's no longer true for a majority of the people searching the web in 2026. Google now answers a huge share of queries directly inside the results page, or skips the results page altogether. ChatGPT, once a chatbot with no idea what happened after its training cutoff, now searches the live web on every relevant question. Perplexity built its whole identity around never showing you a results page in the first place. And Microsoft's Copilot quietly does the same thing inside Bing, Edge, and Windows, for anyone who never left the Microsoft ecosystem. None of these products work identically, and the honest state of play in 2026 is messier — and more consequential for the sites all of this is built on top of — than "AI is eating search" gets across.

Google is running two AI products at once, and they're not the same thing

It's easy to conflate "AI Overviews" and "AI Mode" because Google talks about both under the umbrella of AI-powered search, but they're structurally different products. AI Overviews is the inline module that appears above traditional blue links on a normal search results page — it uses large language models to generate a summary from web content, cites a handful of sources underneath, and leaves the rest of the results page intact below it. AI Mode is a different experience entirely: a dedicated tab where the results page itself is AI-generated, built for multi-turn, conversational follow-up questions rather than a single query-and-answer exchange.

Google began testing the concept as the "Search Generative Experience" at I/O 2023, rebranded and rolled it out broadly as AI Overviews in the US in May 2024, and expanded it to more than 100 countries by that October. AI Mode followed as a separate experiment starting in March 2025 before graduating to general availability. By Google's own count at I/O 2026, AI Overviews had crossed 2.5 billion monthly users across 190-plus countries and 100-plus languages, while AI Mode — younger and more contained — surpassed 1 billion monthly users and 75 million daily users within its first year, with query volume more than doubling every quarter since launch.

The feature hasn't been free of embarrassing misses. Early AI Overviews drew widespread ridicule for suggesting people eat rocks or put glue on pizza to keep cheese from sliding off, and later mistakes included incorrect claims about which countries' names start with certain letters and wrongly identifying unrelated athletes as siblings. Google's own help documentation is direct about the tradeoff, telling users the feature "can and will make mistakes" and encouraging people to rate answers and fact-check anything important.

ChatGPT didn't bolt search onto a chatbot — it built one specifically for search

OpenAI's path into search ran in the opposite direction from Google's. Rather than layering AI onto an existing results page, OpenAI built ChatGPT Search as a feature inside its chatbot: ask a question that needs current information, and a fine-tuned version of GPT-4o fetches live web results — sports scores, stock quotes, news — and weaves them into a conversational answer with linked sources and, where relevant, embedded photos or video. It launched publicly on October 31, 2024, evolving out of a "SearchGPT" prototype OpenAI had shown off that summer, initially limited to ChatGPT Plus and Team subscribers before OpenAI opened it to everyone by December 2024.

What makes ChatGPT Search different in scale, not just design, is that it rides on top of the single largest AI product audience in the world. OpenAI's official count put ChatGPT at roughly 900 million weekly active users as of February 2026, up from 400 million a year earlier — and by late July 2026, reporting from The Information put weekly actives close to the 1 billion mark, a threshold OpenAI hadn't officially confirmed at the time of writing. Whatever the exact figure, no other conversational AI product is remotely close, which means ChatGPT Search doesn't have to win users the way a standalone search engine would — it just has to be the thing people reach for when a chat needs a fact.

Most of these products don't compete for the same click. They compete for whether a click happens at all.

Perplexity: a fraction of the users, an outsized share of the conversation

Perplexity is the smallest of the major players by raw usage and, by most accounts, the one that set the template the others eventually followed. Founded in August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, Perplexity launched its core search engine that December with a simple pitch: instead of a ranked list of links, give people a synthesized answer built from real-time web search, with every claim tied to a visible, clickable citation. No traditional results page, no display ads — a direct answer-engine model, years before "answer engine" became an industry buzzword.

Growth was gradual and then not: 2 million unique visitors by February 2023, a billion-dollar-plus valuation by April 2024, and a jump to roughly 780 million monthly queries by May 2025. Reported monthly active users across Perplexity's products passed 45 million in 2026, with successive funding rounds pushing its valuation past the twenty-billion mark by late 2025 and higher still into 2026. The company has also pushed well past search itself: Comet, a Chromium-based AI browser, went free to all users in October 2025; a more autonomous "Perplexity Computer" agent followed in February 2026; and the company struck shopping-integration deals with Amazon and Nvidia. Perplexity offers free and paid subscription tiers alongside its consumer and enterprise products — the tier structure itself is a real feature of how the company operates, separate from any pricing detail attached to it.

The quieter fifth player: Microsoft's Copilot Search

Microsoft's entry gets talked about less, but it's been running the longest. What's now called Copilot Search started life as "Bing Chat" on February 7, 2023 — one of the very first mainstream AI-search hybrids, drawing a million people to its waitlist within 48 hours. Microsoft folded Bing Chat, Windows Copilot, and Microsoft 365 Copilot under one unified "Copilot" brand that September, and on April 4, 2025, launched Copilot Search specifically: a system Microsoft describes as an "Orchestrator" that iteratively generates its own search queries, runs them against the Bing index, and combines the results with GPT-model-generated synthesis. The underlying models have kept expanding — OpenAI's GPT line through GPT-5.5, plus Anthropic's Claude added for commercial customers starting January 2026.

By usage share, Copilot Search is a distant fifth: Statcounter's July 2026 data put Bing at roughly 4.47% of global search share and Copilot at about 3.07% of AI-assistant usage specifically — nowhere near Google, ChatGPT, or even Perplexity's reach. Its real advantage isn't audience size, it's placement: Copilot is the default AI layer inside Windows, Edge, and the Office suite that hundreds of millions of people already use for work every day, whether or not they'd ever choose it as a standalone search engine.

Five products, five different bets on what "search" means now

Lined up next to each other, the differences are less about who has the biggest model and more about what each company is optimizing for — inline convenience, conversational depth, citation-first trust, or ecosystem lock-in.

2.5B+ monthly users

Google AI Overviews ↗

The inline AI summary sitting above Google's normal results page. Fast to see, easy to ignore, and the single largest AI-generated answer surface on the web by user count.

1B+ monthly users

Google AI Mode ↗

A separate, fully AI-generated results tab built for follow-up questions rather than a one-shot query — Google's answer to what a conversational search page looks like.

~900M–1B weekly (app-wide)

ChatGPT Search ↗

Live web search folded directly into the world's most-used chatbot. Launched October 2024; free to all users by that December.

45M+ monthly users

Perplexity ↗

The original citation-first answer engine. Smallest audience here by a wide margin, but the template most of the others are now visibly copying — plus its own Comet browser.

~3% of AI-assistant usage

Microsoft Copilot Search ↗

Bing's search index fused with GPT-generated synthesis, built into Windows, Edge, and Office. Small standalone reach, enormous built-in distribution.

What this is actually costing the websites underneath it

This is the part of the story that's easy to wave away as vibes-based complaining from publishers, except the numbers keep landing in the same place from multiple independent sources that have no reason to agree with each other. SparkToro's audit of real search behavior found that 68.01% of Google searches in the first four months of 2026 ended with no click at all — up from 60.45% just two years earlier, and roughly 45% a decade ago. The same report cites Ahrefs data showing that click-through rate drops by nearly 60% specifically on the subset of searches where an AI Overview appears.

Pew Research Center ran its own analysis of roughly 68,000 real Google searches and found an 8% average click-through rate on results pages that included an AI Overview, versus 15% on pages that didn't — a 46.7% relative reduction. Similarweb separately tracked the zero-click share climbing from 56% to 69% between May 2024 and May 2025. Not every category is losing out equally: branded searches — someone looking for a specific company or product by name — actually saw an 18% increase in click-through rate when an AI Overview appeared, since the summary often just confirms the brand and sends the click straight to it. It's informational, "explain this to me" queries — historically a huge share of publisher traffic — where the decline is concentrated.

ProductHow it answers
Google AI OverviewsInline summary above normal search results
Google AI ModeFully AI-generated results page, conversational
ChatGPT SearchChatbot answer with live web results woven in
PerplexityCitation-first synthesized answer, no results page
Microsoft Copilot SearchBing index + GPT synthesis via an "Orchestrator"

Common questions

Is Google AI Mode the same thing as AI Overviews?

No. AI Overviews is a summary module that sits above a normal results page; AI Mode is a separate tab where the entire results page is AI-generated and built for follow-up questions.

Why do fewer Google searches lead to a click now?

Independent research (SparkToro, Pew Research, Similarweb) consistently finds that when an AI-generated summary appears, people are less likely to click through to a source website — the summary itself often answers the question.

Is ChatGPT actually a search engine?

Not in the traditional sense — there's no ranked results page — but its Search feature does pull live web results into its answers, which functionally competes for the same "I need current information" queries a search engine handles.

How does Perplexity work if it has no ads or results page?

It offers free and paid subscription tiers, plus enterprise and shopping-integration products, rather than running the ad-supported results-page model traditional search engines use.

Does Microsoft's Copilot Search matter if its usage share is small?

Its standalone usage share is small compared to Google or ChatGPT, but it's the default AI layer inside Windows, Edge, and Office — products with enormous built-in reach regardless of whether people would pick it as their primary search engine.

The takeaway

There isn't one AI search winner in 2026 — there are several products solving slightly different problems, at wildly different scales, with one shared side effect. Google's AI Overviews and AI Mode reach more people than anything else here by a wide margin, simply because they're stapled to Google Search itself. ChatGPT Search borrows its scale from the world's most-used chatbot rather than winning search users directly. Perplexity is the smallest of the group but arguably set the whole category's design language. And Copilot Search proves that raw usage share isn't the only kind of leverage — being the default inside Windows and Office counts for something too. What all five have in common is the same underlying shift: more of the open web's traffic is getting absorbed into an answer instead of a click, and the research keeps confirming that isn't a fringe effect — it's now the majority behavior of Google's own users.

Disclosure. DopamineKart has no business relationship with Google, OpenAI, Perplexity, or Microsoft. Nothing in this piece is sponsored, and no affiliate links appear above. Statistics and product details are drawn from company announcements (Google, OpenAI, Microsoft, Perplexity), independent research (SparkToro, Pew Research Center, Similarweb, Ahrefs, Statcounter), and reporting from TechCrunch, Search Engine Land, Search Engine Journal, CNBC, and Wikipedia (cited for founding and product-history background only). Cover photograph is Wikimedia Commons, credited above.

Corrections: email us