The Internet Is Splitting Into Two Worlds. What Does It Mean for Search?

The Internet is structurally splitting into two worlds – a volume world (dominated by video) and a transactional one (API calls and bots). Traditional search engines are losing their role as intermediaries, because 58.5% of Google queries end without a click, while conversational AI assistants like ChatGPT do not replace search but expand the overall market with new types of queries. The article analyzes the impact on different types of search – navigational and local queries remain Google's domain, while informational and creative queries are shifting to AI.

In 2024, Google processed 21.64% more search queries than the year before. At the same time, 58.5% of those queries ended without a single click on any website. Global internet traffic exceeded 33 exabytes per day — a thousandfold the monthly volume of 2010. Eight technology brands generate two-thirds of all data flow. And yet ChatGPT, despite 900 million weekly users, accounts for barely under two percent of the search market.

These seemingly contradictory statistics tell a single story: the internet is undergoing structural decomposition. It is breaking apart into two separate worlds — the volume world (video, where eight giants have absorbed most of the bandwidth) and the transactional world (API calls, automated bots, and artificial intelligences that make up the majority of requests to web servers). The classic web search engine, historically the intermediary between user and content, is losing that role from both sides: users either get answers directly on the results page, or they head off to specialized platforms. Conversational assistants, meanwhile, are not replacing the search engine — they are expanding the overall information market with a new category of queries that no one had asked before.

This report synthesizes data from the Sandvine Global Internet Phenomena Report, Cloudflare Radar, SparkToro/Datos, Imperva, Seer Interactive, the Pew Research Center, and Alphabet's financial statements. The aim is to answer the question: what should the classic search engine, and what should the conversational assistant, take away from current trends?

Understanding today's internet requires distinguishing two fundamentally different metrics.

By volume of data transferred, video dominates, making up — according to Sandvine — roughly 65% of all global traffic. YouTube takes 16% of download traffic on fixed networks (1.9 GB per user per day), Netflix 12% (1.4 GB), Facebook 9%, TikTok 7%, and Amazon Prime 6%. The eight largest brands — Alphabet, Meta, Netflix, Microsoft, Apple, Amazon, ByteDance, and Disney — generate 65% of traffic on fixed networks and 68% on mobile.

By number of requests, the internet looks entirely different. API calls make up 54% of all requests according to Cloudflare (2021), 71% according to Imperva (2023), and as much as 83% of web traffic according to Akamai. In 2024, automated traffic overtook human traffic for the first time — 51% of all web requests come from bots (37% malicious, 14% legitimate crawlers). Bots from large-model companies are massively indexing the web: Bytespider (23% of crawler traffic), ClaudeBot (8%), GPTBot (5.6%) — all according to Cloudflare Radar measurements for 2024.

This dualism is key to search. Google, Facebook, and AWS generate billions of tiny requests (searches, ad auctions, API calls), each transferring a minimal amount of data, but together forming the backbone of the transactional internet. Netflix generates relatively few requests but an enormous volume of data. The search engine operates primarily in the world of requests, not bandwidth.

One of the most resilient internet myths claims that 30% of the internet is pornography. That figure comes from a 2012 article on ExtremeTech that used crude estimates extrapolated from Google DoubleClick data — it was not a measurement of network traffic.

What the actual data say: researchers at Boston University analyzed the million most-visited websites and found that only 4% are pornographic. Pornographic queries make up 10–15% of web searches (in the 1990s it may have been 40%, but that said more about the narrow user demographic of the time than about data volume). In Sandvine's reports, no pornographic site has ever appeared in the top twenty by volume. XVideos, the largest pornographic site in the 2024 data, shows a mere 10 MB per user per day — compared with 1.9 GB for YouTube. The reason is simple: short clips (averaging 10 minutes) at lower resolution (480p–720p) versus hours of high-resolution content on Netflix.

The best available estimates put pornography at 4–8% of total internet traffic.

Sport, paradoxically, dominates not historically but only now. Sandvine 2025 revealed that all ten days with the highest internet traffic in the U.S. in 2024 coincided with a live sports broadcast — nine NFL games and the Paul vs. Tyson boxing match on Netflix. The share of U.S. households with a sports streaming subscription rose from 4% (2019) to 38% (2025). The market reached $33.9 billion, with a forecast of $75 billion by 2030.

Internet traffic has gone through three phase transitions.

In the file-sharing era (2004–2010), BitTorrent dominated with 70% of global traffic (in some regions as much as 83%). The pivotal year 2010 brought a historic turn — for the first time in a decade, video overtook direct sharing. BitTorrent fell from 44% (2008) to 17% (2011), 9% (2013), below 5% (2015), and to today's 3–4%.

The COVID period (2020) caused a one-off jump of 20–50% in a single week — Zoom exploded from 10 million to 300 million daily users in four months. The surge became a permanent baseline. Total volume grew fortyfold over 14 years: from 20–25 EB per month (2010) to roughly 1,000 EB per month (2024).

A study by Rand Fishkin (SparkToro), using user-behavior data from the firm Datos (Semrush) from July 2024, established the key figure: 58.5% of American and 59.7% of European Google searches end without any click-through. For every 1,000 searches there are only 360 clicks to the open web in the U.S. and 374 in the EU. A further roughly 30% of clicks go to Google-owned platforms (YouTube, Maps, Flights).

On mobile, the share of zero-click queries reaches 77.2%; on desktop, 46.5%. Given that mobile searches make up 63% of all queries, the overall share continues to rise. SimilarWeb recorded a jump in news queries from 56% to 69% within a single year (May 2024 → May 2025). Bain & Company found that 80% of consumers rely on zero-click results for at least 40% of their searches.

The trend has been deepening since at least 2016, when the share of zero-click queries was roughly 44%. In 2019 it reached 50%, and in 2020 roughly 65%. Methodological differences between panels (Jumpshot, SimilarWeb, Datos) complicate direct comparison, but the direction is unambiguous.

Google's AI Overviews feature (formerly SGE), launched in May 2024, represents the key mechanism accelerating the zero-click trend. Its share of queries is growing rapidly: from 6.49% (January 2025) through 13.14% (March) to roughly 30% (September 2025, seoClarity). On mobile it recorded a 474.9% year-over-year increase in frequency. In health and education, AI Overviews cover up to 90% of queries, whereas in e-commerce Google has protectively held their occurrence down to 4–18.5%.

The impact on click-through rate is fundamental. The Pew Research Center (68,879 searches, 900 adults, March 2025) found that the click-through rate with an AI Overview present is 8%, versus 15% without one — a relative drop of 46.7%. Only 1% of visits led to a click on a link directly within the summary.

Seer Interactive, in a longitudinal study (3,119 informational queries, 25.1 million impressions, 42 organizations, June 2024 – September 2025), measured more dramatic declines: the organic click-through rate fell from 1.76% to 0.61% (a 61% drop), and the paid click-through rate from 19.7% to 6.34% (a 68% drop).

A critically important finding: even queries without an AI Overview show a 41% drop in click-through rate, suggesting a broader behavioral change — users are adapting to finding answers directly on the results page, regardless of the specific format.

Physically, AI Overviews take up on average more than 1,200 pixels of height — more than the entire visible area of a desktop screen. The first organic result is pushed to roughly 1,674 pixels below the top edge. Concrete effects: HubSpot lost roughly 50% of its organic traffic (from 13.5 million to under 7 million per month in November–December 2024), Chegg recorded a 49% drop and filed an antitrust lawsuit, and Forbes lost roughly 50% of its traffic.

Running parallel to the zero-click trend is the shift of the first query from Google to specialized platforms. Amazon today captures 54–66% of product searches — JungleScout reports that 66% of American consumers begin a product search directly on Amazon rather than Google. Amazon's conversion rate (12.7%) dramatically exceeds the e-commerce average (2.3%). Amazon does not generate traffic from Google — users bypass the search engine entirely.

Google found internally (2022) that nearly 40% of young people (18–24) look for restaurants via TikTok or Instagram instead of Google Maps. But the preference for TikTok over Google among Generation Z has since fallen from 8% to 4% (2024–2026) — the real new threat is becoming ChatGPT, which 14% of consumers across age groups prefer.

Reddit has undergone an unprecedented rise in organic visibility — from roughly position 80 on Google (July 2023) to 3rd–5th place (July 2024), with a visibility increase of roughly 1,300%. It appears in 97.5% of review queries and is the most-cited source in Google's AI Overviews. Google is de facto strengthening the platform that users prefer for an authentic opinion.

In local search, Google Maps maintains dominance with 2 billion monthly users. "Near me" queries grew by 900% over two years. Google's overall market share in global search nonetheless fell below 90% for the first time since 2015 — to 89.34% (StatCounter, Q4 2024).

ChatGPT reached 900 million weekly active users (February 2026), processes over 2.5 billion queries per day, and ranks among the five most-visited websites in the world. Perplexity is growing to 780 million queries per month (May 2025, up from 230 million in August 2024). Nonetheless, SparkToro/Datos (March 2025) put ChatGPT's share of the classic search market for 2024 at just 0.25% — Google processed 373× more searches.

A Semrush study (260 billion rows of user-behavior data, August 2025) produced the key conclusion: adopting ChatGPT does not reduce Google usage, it expands overall search behavior. A full 95–99% of conversational-assistant users continue to use Google. Artificial intelligence therefore so far represents an expansion of the information market, not a replacement.

Data from NP Digital and First Page Sage make it possible to distinguish which types of queries are at risk.

Creative and generative tasks (writing copy, brainstorming ideas, analyses, code) — ChatGPT 64%, Google 29%. This is the category where conversational assistants have already won. These queries essentially never appeared in the classic search engine — they are a market expansion.

Informational queries (factual questions, explanations, definitions) — ChatGPT 23%, Google 71%. A full 93.7% of queries to ChatGPT are informational in nature (NP Digital). AI Overviews trigger on Google for 88–96.5% of informational queries, showing that even Google sees this segment as AI terrain. Queries of eight or more words have a 57% chance of triggering an AI Overview — corresponding precisely to complex informational queries where conversational assistants excel.

Commercial investigation (product comparisons, reviews) — 5–10% ChatGPT, 15% Google. A growing segment for artificial intelligence, but users still prefer authentic opinions (Reddit) or product databases (Amazon, price comparison engines).

Transactional queries (purchase, order, booking) — ChatGPT 5%, Google 90%. Google protects this category with limited deployment of AI Overviews (4–18.5% in e-commerce). The infrastructure for transactions — product inventory, payments, logistics — remains out of reach for conversational assistants.

Navigational queries (finding a specific website or service) — ChatGPT 3%, Google 93%. AI Overviews trigger for only 1.2–1.7% of navigational queries. There is no point offering an AI answer when the user wants to go to facebook.com.

Local queries (restaurants, shops, services nearby) — de facto zero share for conversational assistants. Google Maps, with 2 billion users and 50% of voice queries carrying local intent, remains an unthreatened domain.

Google Search revenue is accelerating: from $198.1 billion (2024) to an estimated $232–235 billion (2025), with year-over-year growth of 13–17%. Quarterly revenue for Q4 2025 reached a record $63.07 billion (up 17% year-over-year). Google is monetizing AI Overviews "at approximately the same rate" as classic results (Chief Business Officer Philipp Schindler).

But under the hood, a structural shift is underway. The cost per click in the Google Ads system is rising continuously — from an average of $4.66 (2024) to $5.26 (2025), an increase of 12.88% (WordStream, 17,000 campaigns). In the B2B purchasing segment, cost per click jumped by 29% while the click-through rate simultaneously fell by 26% (Dreamdata). Advertisers are paying more for less effective results. Google compensates for the loss of clicks by raising the price of the remaining ones — a model that works as long as advertisers have worse alternatives.

eMarketer predicts that in 2025 Google will, for the first time, lose its majority share of the U.S. search-advertising market — falling to 48%, while Amazon reaches 24%. The AI-driven advertising market is only just starting up: $1.1 billion (2025), with a forecast of $26 billion by 2029.

ChatGPT launched ads in February 2026 at roughly $60 per thousand impressions and a minimum budget of $200,000. Perplexity, by contrast, opted for a revenue-sharing model with publishers (the Comet Plus program: an 80/20 split). The monetization model for conversational assistants remains an open question.

The economic data, however, favor artificial intelligence: traffic referred by conversational assistants converts at 14.2%, versus 2.8% for Google (Exposure Ninja). Ahrefs reports 23× better conversion for visitors from conversational assistants. The economic value of such traffic is 4.4× higher per visit. The problem is volume — AI-referred traffic still makes up just 0.18% of total traffic (SE Ranking, August 2025), even though it is growing 357% year-over-year (SimilarWeb).

Meanwhile, traffic from Google to publishers fell by 33% globally and 38% in the U.S. (Chartbeat/Reuters Institute, November 2024–2025). Traffic from Facebook fell by 68% from its peak in March 2020. The open-web ecosystem is losing sources of visitors from all sides.

Navigational and local queries are the fortress. Google Maps with 2 billion users, "near me" searches with 900% growth, and 50% of voice queries carrying local intent show that local search remains an unthreatened domain. For regional search engines, this confirms the strategic value of their own mapping services — local search is the segment where conversational assistants cannot compete without their own geolocation infrastructure.

Transactional queries protect the ecosystem. Google deliberately limits AI Overviews in e-commerce to 4–18.5%. Amazon captures 54–66% of product searches through direct navigation. For regional search engines this means that vertical services (price comparison engines, real-estate portals, business directories) play the role of capturers of purchase intent — analogous to what Amazon does to Google in the American context.

Informational queries are systematically losing value. AI Overviews cover 88–96.5% of informational queries, and the click-through rate is falling by 61%. This is the category where the classic search engine functions not as an intermediary but as an endpoint — the user gets an answer and leaves. The click-based monetization model is eroding.

Invest in optimization for all channels. Brands today optimize simultaneously for Google, YouTube, TikTok, Amazon, Reddit, and conversational assistants. The new discipline of optimization for generative systems (generative engine optimization) — focused on being cited in AI answers — has reached a market size of $886 million (2024). Brands cited in Google's AI Overviews gain 35% more organic clicks. A search engine that is a source for artificial intelligence (being cited in its answers) has higher strategic value than a search engine that competes with it.

Creative tasks are a battle already won. With a 64% share in creative and generative tasks, conversational assistants dominate a category that essentially did not exist in classic search. This is pure market expansion, not cannibalization.

Informational queries are the main battleground. 93.7% of queries to ChatGPT are informational in nature. This is where direct competition with Google takes place, but the data show that users are not leaving Google — they are adding artificial intelligence as a complementary channel. The key differentiator: complex, multi-word queries (8 or more words = a 57% chance of an AI Overview) are the natural domain of conversational assistants.

Transactional and local queries remain out of reach. What is missing is the infrastructure for up-to-date local data, product inventory, maps, and user trust for financial transactions. ChatGPT Shopping is in the experimental phase, Perplexity is attempting a direct-purchase button, but conversion stays with traditional platforms.

Conversion value is the main argument for monetization. With a 14.2% conversion rate (5× Google) and 4.4× higher value per visit, AI-referred traffic is extremely valuable — there just isn't enough of it yet. Strategically, this is about building volume while preserving the quality of recommendations.

Multimodal search is an opportunity. Google Lens processes 20 billion visual searches per month (a fourfold increase since 2021). The Circle to Search feature captures 10% of searches among users who try it. Conversational assistants with the ability to recognize images, accept voice input, and analyze documents have a natural advantage in these new modes of interaction.

The most likely scenario for the coming 3–5 years is neither "artificial intelligence will replace search" nor "search will absorb artificial intelligence." The data consistently point to the coexistence of two complementary paradigms.

The classic search engine is transforming into a navigational and transactional tool — a router that guides the user to the right service (a map, a shop, a website) and captures advertising revenue from intent close to conversion. The zero-click trend paradoxically strengthens this role: the search engine becomes an endpoint for simple factual queries and a dispatcher for complex actions.

The conversational assistant is establishing itself as an analyst and creative partner — a tool for complex information synthesis, idea-finding, writing, and analysis. Its economic value per session is higher, but its volume remains an order of magnitude smaller. Monetization will probably not follow the path of classic display advertising, but rather subscriptions, commission models, and revenue sharing with data providers.

For regional search engines, the implication is this: invest in vertical services (maps, price comparison engines, real-estate portals, business directories) as irreplaceable sources of structured data that conversational assistants need but cannot create themselves, and at the same time build an artificial-intelligence layer on top of that data. Not a horizontal conversational bot competing with ChatGPT, but a vertically integrated assistant that knows where in your area you can buy the cheapest washing machine and what apartment you can afford for that price.

This report synthesizes data from the following primary sources: the Sandvine/AppLogic Networks Global Internet Phenomena Report (2023, 2024, 2025), the Cisco VNI and Annual Internet Report, the Cloudflare Radar Year in Review (2023, 2024, 2025), the Imperva Bad Bot Report 2024/2025, the SparkToro/Datos Zero-Click Search Study (2024), the Pew Research Center (2025), the Seer Interactive AIO Impact Study (2025), Semrush — user-behavior study (2025), Alphabet SEC filings Q4 2025, WordStream Google Ads Benchmarks (2025), BrightEdge AI Overviews Research, and SimilarWeb AI Referral Traffic Data.

Limitations of the analysis. User-behavior data from various panels (Jumpshot, SimilarWeb, Datos) use differing methodologies, which complicates direct comparison of trends. The zero-click figures also include queries where the user deliberately sought a quick answer (weather, time, exchange rate) — not every zero-click query represents a "stolen" visit. Data on AI-referred traffic are at an early stage and may be underestimated due to imprecise tracking of such sessions. The figures on ChatGPT's market share versus Google depend on the definition of "search" — if we include creative and generative tasks, the AI share is substantially higher.

Transparency of creation

The conception, structure, and editorial line of the article are the work of the author, who prepared the content outline, set the key theses, and directed the entire creation process. Generative AI (Claude Opus 4.6, Anthropic) was used as a tool for research, fact-checking, and expanding the author's draft.

The author verified the key findings and approved the final wording. No part of the text was published without conscious authorial oversight. Factual data were verified against the publicly available sources cited in the text.

This process complies with the transparency principles of EU Regulation 2024/1689 (the AI Act). #poweredByAI