We tracked AI adoption across four internet layers during Q1 2026 using Cloudflare Radar data, covering 67 million DNS queries per second and 81 million HTTP requests. ChatGPT hit global domain rank #11, AI bot traffic surged 34%, and Claude overtook Google Gemini for the #2 generative AI position. This May update cross-references our Q1 telemetry against two Statista publications — the Q2 2026 AI dossier (33 pages, study ID 38609) and the September 2025 in-depth AI market analysis (295 pages, study ID 50485) — to frame whether Q1's infrastructure curve is the beginning, middle, or end of the wave.

Key findings (updated May 19, 2026 with Statista's September 2025 in-depth AI market analysis):

• The "other" TLD category (which includes .ai domains) grew 20% within Q1, from 14.7% to 17.6% of DNS queries, while .com declined from 60.8% to 58.9%
• chatgpt.com ranked #11 globally, sitting between fbcdn.net and amazon.com — the first AI-native platform in the top 15
• AI bot traffic increased ~34% over the quarter, with training-purpose crawls growing from 42% to 52% of all bot activity
• Meta-ExternalAgent emerged as the fastest-growing AI crawler (+43% within Q1), while GPTBot traffic dropped 30%
• DeepSeek shot from rank #9 to #4 among GenAI services in just three weeks during late January
• Applebot saw a 5x traffic surge by late March
• Statista forecasts the global AI market to grow from $94.81B (2020) to $1.675T (2031) — a 17.7x expansion, with GenAI alone hitting $442B by 2031
• Global business AI adoption jumped from 20% (2017) to 72% (2024) — the +17-point gain between 2023 and 2024 is the steepest single-year jump in the IDC/McKinsey series (IDC; McKinsey via Statista)
• U.S. AI investment hit $471B in May 2025 — roughly 4x China's $119B and more than the next 8 countries combined (Spherical Insights via Statista)
• China (58%) and India (57%) are the only countries where AI deployment exceeds "still exploring" — the U.S. sits at 25% deployed / 43% exploring (Vention via Statista)
• The AI cybersecurity market is forecast to grow 5.5x ($24.3B → $133.8B, 2023-2030) — and organizations with extensive AI security automation save $1.9M per breach vs no automation (IBM; Ponemon Institute via Statista)
• 53% of new unicorn births in 2025 were AI startups, up from 6% in 2015 (CB Insights via Statista)
• Enterprise AI adoption jumped 23 points in a single year (55% → 78% from 2023 to 2024), with developing markets posting the biggest gain (+28 points)

How Did Q1 2026 AI Adoption Trends Evolve in April 2026?

According to Cloudflare Radar AI insights, Applebot held its late-Q1 surge at 9.23% of AI crawler traffic in April 2026, GPTBot's decline extended to 9.84% (-2.46 pts below Q1's average), and training-purpose crawls held the wave at 51.5% of all AI bot activity. We re-pulled all four layers for the full month of April 2026 (Q2 month one). The "other" TLD bucket (which we used as the proxy for the AI domain rush) contracted from 17.6% back to 15.0% — the first reversal in any of the four layers. Updated May 2, 2026.

I've been pulling these four-layer datasets every quarter for over a year now, and April was the first time I've seen a layer reverse rather than just decelerate. The "other" TLD bucket dropping back to 15% caught me off guard — speculative domain registration usually has more momentum than this, especially in a category as hot as AI. Watching whether May restores that trajectory is the most interesting open question I'm tracking for Q2.

Which Cloudflare Radar Metrics Changed Most Between Q1 2026 and April 2026?

Metric	Q1 2026 (post)	April 2026	Δ	Direction
Layer 1 — DNS
.com query share	58.9% (Q1 end)	60.22%	+1.3 pt	Bounced
.net query share	12.31% (avg)	11.63%	-0.68 pt	Slight slip
"Other" TLDs (incl. .ai/.app/.xyz)	17.6% (Q1 end)	15.0%	-2.6 pt	Reversed
.io query share	1.76% (Q1 end)	1.77%	+0.01 pt	Flat
.ru query share	2.16% (mid-Mar spike)	2.10%	settled at baseline	Spike resolved
NXDOMAIN rate	10.74%	10.48%	-0.26 pt	Slight improvement
Layer 3 — AI bot user agents
Googlebot	35.0%	30.28%	-4.72 pt	Eased
Meta-ExternalAgent	14.2% (avg, ended 15.7%)	14.91%	held the gain	Stabilized
ClaudeBot	11.3%	11.69%	+0.39 pt	Steady
GPTBot	12.3% (avg, ended 8.6%)	9.84%	decline extended	Confirmed pullback
Bingbot	9.1%	8.04%	-1.06 pt	Eased
Applebot	3.9% avg, peaked 10.2%	9.23%	held near peak	Surge consolidated
Bytespider	3.5%	5.73%	+2.23 pt	New riser
Amazonbot	4.8%	4.47%	-0.33 pt	Steady
OAI-SearchBot	2.2%	1.91%	-0.29 pt	Quiet
Layer 3 — crawl purpose
Training	45.4% (avg, ended 52%)	51.5%	held the surge	Confirmed
Mixed Purpose	44.1% (avg, ended 44%)	38.3%	-5.8 pt	Re-classified down
Search	8.0% (avg, ended 6.3%)	7.5%	mild rebound	Flat
User Action	2.1%	2.2%	flat	Flat
Layer 3 — bot industry mix
Retail	28.2%	28.9%	+0.7 pt	Continued lead
Computer Software	13.6%	12.96%	-0.64 pt	Slight pullback
IT & Services	5.8%	5.78%	flat	Flat
Internet	5.0%	5.20%	+0.2 pt	Flat
Telecommunications	(not top 5)	3.85%	new top 5	New entrant
Gambling & Casinos	2.8%	dropped out of top 9	exited	Q1 anomaly

Which Q1 2026 Narratives Did April Reverse, Confirm, or Extend?

1. The Applebot story has fully crossed from "trend" to "structural." Q1 closed with Applebot at a 10.2% peak (the late-March spike). April held that gain at 9.23% averaged across the full month — Applebot is now plausibly a top-three AI crawler, having overtaken Bingbot (8.04%) and closed within 0.6 points of GPTBot (9.84%). The post's read that "this crawl volume tells us they're training something far bigger than what shipped" looks correct one month later. The next Apple Intelligence announcement is the macro signal to watch.

2. GPTBot's Q1 decline wasn't noise — it extended into April. GPTBot fell from 12.3% (Q1 avg) → 8.6% (Q1 end) → 9.84% (April avg). The Q1 post offered two explanations (data sufficiency vs publisher resistance). April's data fits both: GPTBot rebounded slightly inside April (vs the 8.6% Q1 close) but is still 2.46 points below Q1's average, while Meta-ExternalAgent and Applebot held or grew their share. OpenAI's relative position in the crawl economy is structurally weaker than it was a quarter ago — that's now established, not a one-month read.

3. The "other" TLD bucket reversed. This is the only Q1 narrative that materially flipped. Q1 said the "other" bucket (which includes .ai/.app/.xyz domains) grew from 14.7% to 17.6% — a 20% in-quarter gain we used as a proxy for the AI domain rush. April shows the bucket back at 15.0%, almost back to Q1's start. Two readings are plausible: (a) the Q1 surge was speculative registration that didn't sustain real query volume into Q2; (b) the late-March end-of-Q1 spike was anomalous (.ru spike-and-resolve dynamics support this). Either way, the "domain gold rush" Q1 framing should be tempered. NXDOMAIN at 10.48% (still high but slightly down from 10.74%) suggests speculative activity is cooling.

Which Q1 2026 AI Crawler Findings Held Through April?

Training-purpose crawls held at 51.5% in April, validating the Q1 close at 52%. The "more foundation models are coming" Q1 read remains the right framing — model builders did not pull back in April. Combined Training + Mixed Purpose stayed near 90% of all AI crawl activity, almost identical to Q1.

Retail kept its lead as the most-crawled industry at 28.9% (+0.7 pt). The "if you're in e-commerce, AI bots are your second-largest traffic source after Google" Q1 read continues to apply. Computer Software, IT, and Internet held their Q1 ranks. Telecommunications appeared in the April top five; Gambling & Casinos (Q1's #5 at 2.8%) dropped out — a weight-of-data shift that may simply be normalization noise quarter to quarter.

We don't have direct GenAI service ranking telemetry for April from the dimensions used in this report, so we're not refreshing Layer 4 here — but the GPTBot decline alongside Applebot and Bytespider rises is consistent with the consumer-side rankings continuing to consolidate at the top while the long tail churns.

What Should We Watch for in May 2026?

Will the "other" TLD bucket recover or stay reset? If May rebounds to 17%+, Q1's read about the domain gold rush is right and April was a calendar artifact. If May stays at 15%, the Q1 surge was end-of-quarter speculation that didn't bind to real usage.

Does Applebot pass GPTBot in May? The 0.6-point gap is small enough that one more Apple Intelligence-related crawl burst flips the order. If Applebot crosses GPTBot, that's a leadership change in the AI training data race.

Does Meta-ExternalAgent re-accelerate, or has it found its 2026 floor? Q1's +43% Meta surge predicted a Llama release within 2-4 months. April held the gain but didn't extend it. Continued plateau suggests the release is imminent; another surge would push the timeline back.

Every major AI adoption report published this quarter relied on surveys. McKinsey's State of AI surveyed executives. Microsoft's AI Economy Institute polled populations across countries. Deloitte's State of AI in the Enterprise asked CIOs about their roadmaps. Their numbers are useful. But they measure what people say they're doing with AI, not what's actually happening on the wire.

Our approach was different. We pulled passive internet telemetry from Cloudflare's DNS infrastructure, covering 92 days of actual internet behavior. No surveys. No self-reporting. Just raw traffic.

The AI hype cycle has 4 measurable internet layers

Technology hype doesn't just live in headlines and analyst decks. It propagates through the internet's infrastructure in a predictable sequence, and each stage leaves a fingerprint we can measure.

We identified four layers.

The first is the domain gold rush. When a technology trend heats up, entrepreneurs and speculators register domains. DNS query patterns shift. New TLDs gain traction. This tends to run 6-12 months ahead of actual product adoption.

Second comes traffic. As products launch and users show up, domain rankings change. A platform jumping from rank #500 to the top 15 means millions of daily users. That separates working products from press releases.

Third, the bot ecosystem. AI companies need training data and they deploy crawlers to get it. The volume, targeting patterns, and purpose mix of these bots reveal which capabilities are being built right now. When training-purpose crawls spike, new model releases follow within months.

Fourth is enterprise adoption. When services reach stable, high-frequency usage patterns (weekday/weekend cycles, consistent ranking positions), hype has converted to revenue. Deloitte's latest survey found that companies with 40% or more of their AI projects in production are expected to double in the next six months. Our traffic data shows which services those production workloads are running on.

Layer 1: The domain gold rush

Cloudflare's 1.1.1.1 resolver processes over 67 million DNS queries every second. That volume makes it one of the most reliable proxies for tracking where internet activity is shifting. During Q1 2026, we watched the DNS data for signals of AI's growing presence in the domain name system.

Here's what the TLD distribution looked like:

TLD	Query share (Q1 avg)	Q1 trend
.com	61.86%	Declined from 60.8% to 58.9%
.net	12.31%	Stable
.io	1.72%	Grew from 1.65% to 1.76% (+7%)
.dev	0.61%	Stable
"Other" (incl. .ai)	~16% avg	Grew from 14.7% to 17.6% (+20%)

April 2026 update — the "other" TLD rush reversed: April's full-month figures put .com back at 60.22%, .net at 11.63%, .io flat at 1.77%, and the "other" bucket back to 15.0% — almost back to where Q1 began (14.7%). The 20% in-quarter "other" surge that we read as a proxy for the AI domain rush did not bind in Q2. This is the only Q1 narrative that materially flipped. Two readings: the late-Q1 spike was speculative end-of-quarter registration that didn't sustain real query volume, or it was an anomalous weekend-spike pattern (similar to the .ru spike-and-resolve dynamic). NXDOMAIN eased slightly to 10.48%, consistent with speculative activity cooling.

The .com TLD still dominates, but its share dropped from 60.8% of queries in early January to 58.9% by late March. At 67 million queries per second, a 2-point shift represents billions of daily queries moving to alternative TLDs.

The real story is the "other" category. This bucket includes .ai, .app, .xyz, and dozens of newer TLDs. It surged 20% within a single quarter. Cloudflare doesn't break out .ai separately yet, but the growth pattern aligns with the AI domain registration boom we've been tracking through our own popular technology data by category..

Why NXDOMAIN rates matter more than registration counts

Surveys can tell you how many .ai domains were registered. They can't tell you how many are actually being used. DNS data can.

10.74% of all DNS queries in Q1 hit non-existent domains (NXDOMAIN responses). One in ten queries pointed to domains that don't resolve. That's a direct measure of speculative activity, typosquatting, and recently expired registrations. For AI-related TLDs, this rate ran even higher, which suggests aggressive domain parking and speculation around AI brand names.

Domain registrations get headlines. NXDOMAIN rates tell you how much of it is real.

A geopolitical footnote

One anomaly worth flagging: .ru domains spiked to 2.16% of queries in mid-March before settling back to baseline. We've documented similar patterns in our cloud provider traffic share analysis. These spikes typically correlate with geopolitical events driving infrastructure changes or traffic rerouting.

For sales teams prospecting into AI-adjacent markets, the DNS layer is your earliest warning system. Domain speculation in a technology category tends to precede the buying cycle by 6-12 months.

Layer 2: ChatGPT enters the top 15 global domains

Domain rankings based on actual HTTP traffic paint the clearest picture of real adoption. And the Q1 2026 rankings delivered one result that would've seemed absurd two years ago.

chatgpt.com. Global rank #11.

Rank	Domain	Category
1	google.com	Search
3	cloudflare.com	Infrastructure
5	apple.com	Hardware/services
6	microsoft.com	Enterprise software
7	facebook.com	Social media
10	fbcdn.net	CDN (Meta)
11	chatgpt.com	Generative AI
12	amazon.com	E-commerce
22	bing.com	Search (AI-powered)
51	sentry.io	Developer tooling

ChatGPT now generates more traffic than Amazon. It sits between Meta's CDN and the world's largest e-commerce platform. No other AI-native platform appears anywhere in the global top 100.

Microsoft's AI Economy Institute found that generative AI adoption reached 16.3% of the world's population, up from 15.1% in the first half of 2025. Our traffic data puts a concrete shape on that number: ChatGPT's position at #11 globally requires hundreds of millions of monthly active users, and it's not sharing the spotlight. OpenAI confirmed 900 million weekly active users in February 2026 — our updated ChatGPT statistics report tracks the full revenue, country, and crawl breakdown behind that number. The next AI-native domain (claude.ai, perplexity.ai, deepseek.com) doesn't crack the top 100.

Where that traffic is coming from

Global domain ranks obscure geography. A consumer AI usage survey fielded by KPMG and the University of Melbourne (48,000 respondents) makes the source of ChatGPT's volume clearer — and it flips the usual narrative about where AI lands first.

Every country at the top is an emerging market. India and Nigeria tie at 92%, UAE and Egypt clear 90%, and China sits at 89%. The first advanced economy on the list, Singapore at 73%, is 19 points behind the leaders. That matches what we see in Cloudflare's DNS data where the "other" TLD bucket (including newer markets) grew 20% in Q1 — the fastest layer in the stack. The population adopting AI aggressively is not where the English-language AI press is written.

What's missing from the trending lists tells a story too

No AI platforms appear in the "trending rise" domain lists. That's not because growth has stalled. It's because ChatGPT has moved past "trending" entirely. When a domain stops rising fast and holds steady in the top 15, it's no longer a trend. It's infrastructure.

Bing at #22 is worth watching separately. As the backend for Microsoft Copilot, its steady ranking reflects AI-powered search growth rather than traditional Bing usage.

One more signal. Developer tooling platform sentry.io climbed to #51 globally. That's a second-order effect of AI adoption: AI-assisted coding drives more deployments, more error tracking, more developer tool usage. Harvard Business Review's February 2026 analysis found 88% of companies report regular AI use. Our traffic data shows what that 88% is actually using: overwhelmingly ChatGPT, with everything else fighting for a distant second.

Government capacity is clustered tight — except at the top

Consumer adoption tells you where demand shows up. Government AI readiness tells you where public-sector AI policy has the infrastructure to absorb it. Oxford Insights' 2025 Government AI Readiness Index (published via Statista's Q2 2026 AI dossier) scores the top 15 countries on their capacity to deploy AI across healthcare, education, and administration.

The United States sits almost ten points ahead of the UK, the largest single-country gap anywhere in the top 15. Outside that lead, everyone is bunched: 13 of the top 15 score between 69 and 78. China is the only non-OECD country in the top five. Pair this against the consumer usage chart and a picture forms — emerging markets lead on consumer uptake, but advanced economies still dominate the government capacity to build AI into public services.

Layer 3: A three-tier AI crawler economy is forming

Every major AI company needs web data. They need it to train models, power search features, and serve real-time user queries. We tracked nine major AI bot user agents across Q1 2026, and the traffic patterns reveal an ecosystem that's maturing fast.

Total AI bot traffic increased roughly 34% over the quarter, measured by a normalized index rising from 0.71 to 0.95.

Bot	Traffic share	Q1 trend	Primary purpose
Googlebot	35.0%	Stable	Mixed (search + training)
Meta-ExternalAgent	14.2%	+43%, from 11% to 15.7%	Training
GPTBot	12.3%	-30%, from 12.4% to 8.6%	Training
ClaudeBot	11.3%	Stable at 11-13%	Training
Bingbot	9.1%	Stable	Search + Copilot
Amazonbot	4.8%	Stable	Alexa AI / training
Applebot	3.9% avg, peaked 10.2%	5x surge by late March	Apple Intelligence
Bytespider	3.5%	Grew to 4.5%	Training (ByteDance)
OAI-SearchBot	2.2%	Stable	SearchGPT

April 2026 update — Applebot consolidated, GPTBot decline extended, new ranking order forming: Applebot held its late-Q1 surge at 9.23% of AI bot traffic — now within 0.6 points of GPTBot (9.84%) and ahead of Bingbot (8.04%). GPTBot extended its Q1 decline (-2.46 pts vs Q1 average). Bytespider continued to climb to 5.73% (+2.23 pts). The Q1 read "OpenAI's share of the crawl economy is shrinking while Meta's and Apple's are growing" is now an established Q2 fact, not a one-quarter signal.

Three things jumped out at us.

Meta is crawling the web harder than anyone expected

Meta-ExternalAgent was the quarter's biggest gainer. From 11% to 15.7% of AI bot traffic in 92 days. Meta has been relatively quiet about its training data strategy compared to OpenAI or Google. The crawl data tells a different story. They're consuming web content at an accelerating rate, which aligns with their Llama model development timeline. When a company's crawling ramps like this, a major model release is usually 2-4 months away.

OpenAI pulled back. The question is why.

GPTBot traffic dropped 30% across the quarter, from 12.4% to roughly 8.6%. This was the single most significant shift we observed.

Two explanations. OpenAI may have accumulated enough training data for their current model pipeline. Or the growing publisher resistance to AI crawlers, which we documented in our robots.txt blocking analysis, is having a measurable effect. Probably both. The practical result is the same: OpenAI's share of the crawl economy is shrinking while Meta's and Apple's are growing.

Apple's crawl volume says more than any WWDC keynote

Applebot surged from 2% of AI bot traffic to 10.2% by late March. A 5x increase in a quarter. Apple Intelligence launched in late 2025 with limited capabilities, and the reviews were mixed. But this crawl volume tells us they're training something far bigger than what shipped. When Apple moves quietly and at scale, you pay attention.

Training, search, and user action: the three tiers

The bot traffic breaks into three purpose tiers that reveal what the industry is building:

Crawl purpose	Share (Q1 avg)	Q1 trend
Training	45.4%	Grew from 42% to 52%
Mixed purpose	44.1%	Declined from 48% to 44%
Search	8.0%	Declined from 7.4% to 6.3%
User action	2.1%	Stable

April 2026 update — training-crawl wave held: April locked in the late-Q1 training surge. Training crawls held at 51.5% (vs Q1's 52% close), Mixed Purpose at 38.3%, Search at 7.5%, User Action at 2.2%. The "more foundation models are coming" Q1 read remains correct one month into Q2. Combined Training + Mixed crawl share stayed near 90% — the model-builder demand is the dominant signal in the AI bot economy, not search-side or agent-side activity.

Training crawls grew from 42% to 52% of all AI bot activity during Q1. More foundation models are coming. The industry hasn't finished building. McKinsey's State of AI survey found that 80% of high-performing companies set growth and innovation as AI objectives, not just efficiency. The crawl data shows model builders responding to that demand in real time.

Meanwhile, search-purpose crawls actually declined. AI-powered search (OAI-SearchBot, parts of Bingbot) is still a small slice of the bot economy. The real volume is in training.

What the bots are after

Industry	Bot traffic share
Retail	28.2%
Computer software	13.6%
Information technology	5.8%
Internet	5.0%
Gambling & casinos	2.8%

Retail dominates at 28.2%. Product catalogs, reviews, pricing pages, comparison content: it's structured, factual data that AI models need. If you're in e-commerce, AI bots are your second-largest traffic source after Google. The gambling number (2.8%) is surprising until you realize that casino and sports betting sites are among the most structured, data-rich websites on the internet.

Vention's 2026 AI report found that 93% of companies are already using AI in some capacity. Our bot data adds color to that number: the AI tools those companies use are themselves consuming web content at an accelerating pace, creating a feedback loop. More adoption means more crawling means more capable models means more adoption.

Layer 4: The generative AI rankings tell us who's winning

Cloudflare Radar tracks generative AI service usage daily. These rankings reflect actual user traffic, not app store downloads or press mentions. The Q1 2026 picture is a market consolidating fast at the top and churning hard below.

Rank	Service	Q1 movement
1	ChatGPT / OpenAI	#1 for 92/92 days
2	Claude / Anthropic	Rose from #4 to #2 by mid-Jan, held 85+ days
3	Perplexity	Stable #3
4	DeepSeek	Rose from #9-10 to #4 in 3 weeks (late Jan)
5	Google Gemini	Fell from #2 to #5 by mid-Feb
6	Character.AI	Stable
7	Grok / xAI	Stable
8	Suno AI	Stable
9	GitHub Copilot	Volatile, #5-9 weekday/weekend pattern
10	Doubao	Intermittent new entrant

The market size behind the rankings

Service rankings show who is winning. They don't show what the prize is worth. Statista Market Insights forecasts that the generative AI segment alone will grow from roughly $5.5 billion in 2020 to $442 billion by 2031.

The 2026 mark is projected at $108 billion — roughly 3x the 2024 actual of $37.87 billion. When ChatGPT holds #1 for 92 consecutive days in a market growing at that pace, the durability of its lead compounds with every quarter it doesn't lose ground. Peak growth year is 2024 at +85%, and by 2031 growth decelerates to ~20% as the base scales. The window for displacing incumbents closes fastest between now and 2028.

ChatGPT's lock is tighter than you think

Number one every single day. 92 out of 92 days. No wobbles. When a product holds rank #1 in its category and ranks #11 among all global domains, displacement becomes a structural problem for competitors, not just a product quality question. You'd need to change user habits at a scale that hasn't happened in consumer tech since the smartphone transition.

Claude's rise isn't noise. It's an enterprise signal.

Anthropic's Claude moved from #4 to #2 by mid-January and held that position for 85+ consecutive days. Not a spike. A sustained shift.

Claude overtook Google Gemini, which held #2 for most of late 2025. The reason matters: Claude's focus on safety, long-context windows, and business-facing use cases is winning the segment where revenue concentration is highest. Deloitte found that only 34% of enterprises are truly reimagining their businesses around AI. The ones that are tend to pick tools built for enterprise workflows, not consumer chatbots.

DeepSeek's three-week rocket

From rank #9-10 to #4 in three weeks. That happened in late January, driven by DeepSeek's open-source model releases. The open-source angle matters here. According to NVIDIA's State of AI report, North America leads with 70% actively using AI, 27% assessing, and just 3% not using it at all. Open-source models give the "assessing" cohort a way to experiment without procurement cycles.

Whether DeepSeek sustains this position depends on trial-to-retention conversion. The daily data suggests they're holding, but the signal is noisier than Claude's steady line.

What happened to Gemini?

From #2 to #5 in about a month. Despite massive cloud infrastructure, deep integration across Workspace, Android, and Search, and distribution advantages no competitor can match, Gemini lost ground to Claude and DeepSeek simultaneously.

The market is choosing specialized AI tools over bundled ones. That's an unusual dynamic. Google has never lost a market where distribution was the deciding factor, but generative AI is apparently not that kind of market. Product quality, or at least perceived product quality, is winning.

The weekday/weekend pattern and what it reveals

GitHub Copilot bounces between #5 and #9, with clear weekday peaks and weekend drops. That's the fingerprint of enterprise adoption. A tool deeply embedded in work routines but ignored on Saturday.

Compare that to Character.AI and Suno AI. Their rankings are flat across weekdays and weekends. Consumer products. The weekday/weekend split is how you identify real B2B adoption from traffic data alone, without asking anyone a survey question.

Two new categories are forming

Windsurf AI (an AI coding assistant) and Manus (an AI agent platform) both entered the top 20 during Q1. They represent something the rankings haven't shown before: category fragmentation.

McKinsey found that 62% of organizations are at least experimenting with AI agents. Manus's appearance in the rankings is the traffic-level confirmation of that survey data. AI is fragmenting from "general chatbot" into specialized vertical tools, and the ranking data captures this as it happens. But Deloitte's enterprise survey adds a caution: only one in five companies has a mature governance model for autonomous AI agents. The adoption is outrunning the guardrails.

The macro AI market behind the Q1 traffic data

The four internet layers measure current behavior. Statista's Q2 2026 AI dossier measures the trajectory — where the market was, where it's forecast to go, and which cohorts are driving the curve. The telemetry tells you what is happening now. The forecast tells you whether now is early, peak, or late.

AI market size 2020-2031: 17.7x over the decade

The worldwide AI market is modeled to grow from $94.81 billion in 2020 to $1.675 trillion by 2031 — an addition of $1.3 trillion between 2026 and 2031 alone. The 2022 dip to $126.78B from the 2021 surge to $206.34B is the only contraction year in the series, and it resolved the following year. Everything after 2024 is forecast, but the 2020-2024 actuals match the trajectory the infrastructure layer confirms: usage is not decelerating. The 34% quarterly growth in AI bot traffic we measured in Q1 aligns with the dossier's implied annualized growth rate through 2028.

User base crosses 500 million by 2028

Revenue forecasts get challenged. User count forecasts are cleaner. Statista's model grows the global AI user base from 115.9 million in 2020 to 729.1 million by 2030 — a 6.3x expansion with no contraction year. The 451.8 million projected for 2026 is roughly the monthly active base ChatGPT alone commands today. Which means either the forecast is conservative, or a meaningful share of ChatGPT users aren't being counted as "AI users" yet by this methodology. Both readings support the infrastructure pattern.

The unicorn signal: from 6% to 53%

In 2015, 6% of new unicorn startups worldwide were AI companies. In 2025, 53% were — more than half of all new billion-dollar valuations in a single year. The inflection point is 2023-2025, which maps cleanly to the post-ChatGPT funding wave. CB Insights also notes that one in five new 2025 unicorns are AI agents specifically — a sub-category inside a category that didn't meaningfully exist two years ago. The same signal that made Manus and Windsurf appear in our Cloudflare Layer 4 rankings is showing up in the VC data.

Enterprise adoption: the 23-point year

McKinsey's Global Survey found that organizational AI use jumped from 55% in 2023 to 78% in 2024 — a 23-point year-over-year gain globally. The biggest regional move came in developing markets, where AI adoption rose 28 points (49% → 77%), likely because procurement friction and regulatory overhead are lower. North America still leads in absolute terms at 82%, but the growth rate is actually faster elsewhere. The production gap remains: using AI in "at least one function" is a low bar, and Deloitte's finding that only 20% of organizations are generating revenue from AI still holds underneath these numbers.

Where AI becomes a required skill

A World Economic Forum employer survey (1,000 respondents) asked which industries expect AI and big data to be core worker skills between 2025 and 2030. Information technology services and telecommunications tie at 66% — effectively, AI skill is no longer optional in those verticals. Financial services (61%) and insurance (58%) are compliance-and-risk-driven: regulation plus model risk management plus fraud detection all push AI deep into day-to-day work. Even the lowest-ranked industry (professional services at 37%) sits above a third of employers. There is no AI-exempt industry in the 2025-2030 outlook.

The agentic AI bottleneck is regional

A Dynatrace/Qualtrics study (919 IT and line-of-business leaders) asked about the hardest parts of moving agentic AI into production. The answer depends on where you are. APAC (60%) and EMEA (50%) name managing and monitoring agents at scale as their #1 barrier. The Americas flip the order: 51% cite a shortage of skilled staff versus 45% citing operational complexity. This is the clearest regional divergence in the dossier. In Asia and Europe, the agent problem is operational. In the Americas, it is labor-market. Either way, it lines up with the weekday/weekend Copilot pattern we saw in Layer 4 — enterprise AI adoption is hitting limits, and those limits are different in different places.

What the September 2025 Statista in-depth report adds

The Q2 2026 dossier we leaned on earlier is a 33-page summary. The September 2025 in-depth report (study ID 50485, 295 pages) is the full long-form analysis behind it — and it surfaces five datasets we didn't have at our April refresh. They sharpen the picture of where the AI adoption curve actually bends.

Enterprise AI adoption: a 52-point expansion in 7 years

The IDC/McKinsey survey series is the longest comparable enterprise-AI-adoption benchmark we have. From 20 percent of businesses using AI in 2017 to 72 percent in 2024, the curve isn't smooth — 2020 and 2022 were both flat or slight retreats, tied to macroeconomic budget cycles. What stands out in the 2024 number is the +17-point jump from 2023's 55 percent. That's the largest single-year gain in the entire series and maps cleanly to the generative AI productionisation wave we see in our Layer 3 bot data (training crawls went from 42 percent to 52 percent in the same window).

The 72 percent number is also where most "AI adoption" coverage stops. The next chart is the more useful one.

The deployment gap: 72% adoption hides who's actually in production

Vention's 2024 country-level survey splits AI use into "deployed" (in production) and "exploring" (considering or piloting). Only two countries — China (58 percent deployed) and India (57 percent) — have more organizations running AI in production than evaluating it. Every other country in the survey has a higher exploring share than deployed share. The U.S. sits twelfth out of fourteen at 25 percent deployed versus 43 percent exploring.

That's a meaningful counterweight to the headline 72 percent global adoption number from McKinsey. Yes, businesses have AI somewhere in their stack. No, most of it isn't in production in most countries. The "experimentation-to-production gap" we already flagged from Deloitte's data shows up sharply when you split deployment status by geography.

For GTM teams, this reverses a common assumption: the U.S. is not the easiest market to sell production AI tools into. India and China are. The U.S. is the easiest market to sell pilots into.

Investment doesn't track deployment

Spherical Insights' May 2025 cut shows the U.S. ($471B) ahead of China ($119B) by roughly 4x in total AI investment — and ahead of the next eight countries combined. That number includes VC funding, corporate R&D, government programs (notably the CHIPS and Science Act's $280B commitment), and infrastructure spending.

What's interesting is the disconnect with deployment. The U.S. invests 4x more than China but deploys at less than half the rate. Two readings: (a) U.S. AI investment is overwhelmingly going into infrastructure and model R&D rather than enterprise deployment, which matches NVIDIA's market position (almost 100 percent share of AI training compute); (b) Chinese enterprises are deploying off the back of a leaner, more domestically-focused AI stack (Baidu's ERNIE, Alibaba's Qwen, Huawei's Ascend chips). Either reading supports our Layer 3 finding that the U.S.-headquartered crawlers (GPTBot, ClaudeBot) are crawling for future models, not current production workloads.

AI cybersecurity is the fastest-growing AI sub-market

Techopedia's AI cybersecurity forecast (via Statista) projects the segment from $24.3B in 2023 to $133.8B by 2030 — a 5.5x expansion. That growth rate is faster than most AI sub-segments in the same window, including the broader machine learning market (32.4 percent CAGR per Statista Market Insights) only because the cybersecurity base is smaller. What matters more than the trajectory is the reason behind it.

The $1.9M reason

IBM/Ponemon Institute's breach-cost benchmark is the clearest ROI signal in the entire Statista in-depth report. In 2024, organizations with extensive AI security automation paid an average of $3.8M per data breach. Organizations with no security automation paid $5.7M. That $1.9M delta isn't a 2024 anomaly — the gap has been positive every single year since 2018, peaking at $3.8M in 2021 (pandemic-era manual operations were the costliest).

This explains why PwC's 2024 survey found 77 percent of organizations expect their cybersecurity budget to grow in 2025, with 20 percent expecting double-digit-percent increases (PwC via Statista). CFOs don't usually fund "AI strategy." They fund "$1.9M of avoided cost per breach." The framing is what gets the budget approved.

For our Layer 4 service rankings, this means AI cybersecurity platforms — Darktrace, CrowdStrike, Microsoft Sentinel, Vectra, Cynet 360 — are likely to push into the top 20 over the next 18 months as adoption scales. We've already seen Microsoft Security Copilot show up in enterprise-pattern traffic (weekday peaks, weekend drops) on Cloudflare Radar.

What this layer adds to the four-layer model

The September 2025 in-depth report doesn't change our four-layer model. It adds a fifth signal underneath it: the dollars that justify the deployment. Layer 4 told us which services are winning user traffic. The cybersecurity breach-cost gap and the investment-vs-deployment country split tell us why those services will keep winning — and where the next 18 months of enterprise budget is actually going to land.

What the data actually tells sales and GTM teams

Cloudflare Radar shows the macro wave. DNS patterns, traffic rankings, bot behavior, service usage. But sales teams don't sell to waves. They sell to companies.

The question isn't "is AI growing?" It's: which specific companies adopted which AI tools, when they adopted them, and what those tools replaced. That's where technographic data platforms like TechnologyChecker come in, detecting 40,000+ technologies across 50M+ domains with real-time adoption and churn signals.

Here's how we'd translate the four layers into action.

If you're an SDR, the bot traffic data gives you targeting intelligence. Retail (28.2% of bot traffic) and computer software (13.6%) are the most crawled industries. Those verticals will see the most AI tool launches in the next 6-12 months. Companies already seeing heavy AI bot activity on their sites are thinking about AI strategy right now, whether they're blocking crawlers or building AI features. Digital Third Coast found that 52% of large organizations have dedicated AI adoption teams compared to 23% of small organizations. Enterprise prospects have buyers specifically tasked with AI procurement. That's who you're looking for.

If you lead GTM or RevOps, the four-layer model gives you timing. Domain speculation in a category (Layer 1) means start building content. Traffic ranking shifts (Layer 2) mean launch prospecting campaigns. Bot traffic targeting an industry (Layer 3) means those companies are 3-6 months from evaluating vendors. Stable service rankings (Layer 4) mean you're selling into an established market with clear incumbents. GitHub Copilot's weekday/weekend pattern is your template for identifying enterprise adoption in any vertical. Our technology lookup checker shows how real-time detection data can pinpoint companies at each stage.

If you run marketing, the DeepSeek and Claude stories prove that distribution doesn't guarantee market position. Google had every advantage and still dropped three positions. Content strategy should track these shifts. The audience searching for AI adoption data wants to understand dynamics, not just confirm AI is popular. Harvard Business Review found that 88% of companies now use AI regularly. Your marketing should assume AI familiarity and focus on differentiation.

The experimentation-to-production gap is real

One pattern we noticed across all four layers: there's a wide gap between AI experimentation and AI in production. Deloitte's latest data makes this concrete. Worker access to AI rose 50% in 2025, and 66% of organizations achieved productivity improvements. But only 20% are seeing actual revenue growth from AI, despite 74% aspiring to it. Only 42% believe their strategy is highly prepared.

Our traffic data maps to this gap. ChatGPT and Claude have stable, enterprise-grade traffic patterns. But below the top 5, the rankings churn weekly. Tools appear, spike, and fade. The long tail of AI services is still in experimentation mode. Only 27% of organizations review 100% of AI outputs before acting on them, per Digital Third Coast. The rush to adopt is outpacing the infrastructure to govern.

Q1 2026 AI adoption benchmarks at a glance

Benchmark	Value	Source
ChatGPT global domain rank	#11	Cloudflare Radar (our analysis)
AI bot traffic growth (Q1)	+34%	Cloudflare Radar (our analysis)
Training-purpose crawl share	45.4% → 52%	Cloudflare Radar (our analysis)
.ai-adjacent TLD growth	+20% in Q1	Cloudflare Radar (our analysis)
NXDOMAIN rate	10.74%	Cloudflare Radar (our analysis)
Most crawled industry	Retail (28.2%)	Cloudflare Radar (our analysis)
GenAI #1 consistency	92/92 days	Cloudflare Radar (our analysis)
Claude #2 streak	85+ days	Cloudflare Radar (our analysis)
Enterprise AI adoption rate	72%	McKinsey Global Survey
Global GenAI population reach	16.3%	Microsoft AI Economy Institute
Worker AI access increase	+50% in 2025	Deloitte State of AI
Orgs achieving AI revenue growth	20% (74% aspire)	Deloitte State of AI
Companies with AI agent governance	1 in 5	Deloitte State of AI
Companies regularly using AI	88%	Harvard Business Review
AI market size 2026 (forecast)	$347B	Statista Market Insights
AI market size 2031 (forecast)	$1.675T	Statista Market Insights
GenAI market size 2031 (forecast)	$442B	Statista Market Insights
AI users worldwide 2026 (forecast)	451.8M	Statista Technology Market Insights
New unicorns that are AI (2025)	53%	CB Insights via Statista
Enterprise AI adoption YoY gain	+23 points (55% → 78%)	McKinsey Global Survey
Top consumer-AI country (2025)	India, Nigeria (92%)	KPMG / Univ. of Melbourne
#1 government AI readiness	United States (87.2)	Oxford Insights
Global business AI adoption 2017 → 2024	20% → 72%	IDC; McKinsey via Statista
U.S. AI investment (May 2025)	$471B	Spherical Insights via Statista
China AI investment (May 2025)	$119B	Spherical Insights via Statista
#1 AI-deployed country (2024)	China (58%)	Vention via Statista
U.S. AI deployment vs exploring (2024)	25% / 43%	Vention via Statista
AI cybersecurity market 2030 (forecast)	$133.8B	Techopedia via Statista
Breach cost savings (extensive AI vs none)	$1.9M per breach	IBM; Ponemon Institute via Statista
Orgs expecting cybersecurity budget growth 2025	77%	PwC via Statista

Our methodology

We collected data from four Cloudflare Radar API endpoints covering January 5 through April 6, 2026 (92 days, full Q1 plus the first week of April).

Cloudflare Radar draws on traffic flowing through Cloudflare's global network. That includes DNS resolution data from the 1.1.1.1 public resolver (67M+ queries/sec), HTTP request volumes, AI bot classification by user agent and declared purpose, and generative AI service rankings by traffic volume.

We used four endpoint groups:

• DNS query distribution by TLD (summary and weekly time-series)
• Global domain rankings (top 100 popular and top 50 trending)
• AI/ML bot traffic by user agent, crawl purpose, and target industry
• Internet services rankings filtered to "Generative AI"

All trend percentages compare the first week of Q1 to the last available data point. Bot traffic was analyzed by both user agent identity and declared crawl purpose. GenAI service rankings used daily position data to calculate streak lengths and transition points. We cross-referenced signals across layers to identify patterns that wouldn't be visible in any single data set.

What Cloudflare doesn't see: traffic that doesn't flow through its network, DNS queries to other resolvers (Google Public DNS, ISP resolvers), and .ai TLD breakouts (we track it within the "other" bucket). Bot user agents can be spoofed, though major AI companies have strong incentives to identify their crawlers honestly for robots.txt compliance. Our SSL certificate transparency research follows a similar methodology, using infrastructure-level data to draw adoption conclusions.

Statista Q2 2026 AI dossier (added April 2026)

For the long-view forecast layer, we cross-referenced our Q1 telemetry with Statista's 33-page Q2 2026 artificial intelligence dossier (study ID 38609). The dossier pulls together 27 curated datasets from Statista Market Insights (AI market sizing and segment forecasts through 2031), CB Insights (AI share of unicorn births), KPMG/University of Melbourne (consumer AI usage by country, 48,000 respondents), Oxford Insights (Government AI Readiness Index), McKinsey & Company / Stanford University (enterprise AI adoption by region), the World Economic Forum (AI as a core worker skill by industry), Dynatrace/Qualtrics (agentic AI barriers), Deloitte (talent strategy adjustments), and the Digital Education Council (faculty views on AI in education).

Attribution format in the dossier: the chart header lists the original data provider as "Source(s)"; Statista is the aggregator. Each chart embedded in this post preserves that attribution via the chart card's Source: X via Statista header line. "Read more" links in the chart panels point back to this article; the underlying Statista chart URLs are available via the dossier.

Statista in-depth AI market analysis, September 2025 (added May 2026)

The May 2026 refresh adds five additional datasets sourced from Statista's 295-page in-depth artificial intelligence market analysis (study ID 50485, published September 2025). The in-depth report is the long-form source behind the Q2 2026 dossier and includes data dimensions the dossier summarises but doesn't reproduce in full:

• IDC; McKinsey via Statista — Global AI adoption rate in businesses, 2017-2024 (20% → 72%)
• Vention via Statista — AI deployed vs exploring by country, 2024 (China and India are the only countries past the deployment threshold)
• Spherical Insights via Statista — Leaders in AI investment by country, as of May 2025 (U.S. $471B; China $119B)
• Techopedia via Statista — AI cybersecurity market value, 2023-2030 ($24.3B → $133.8B)
• IBM; Ponemon Institute via Statista — Average breach cost by security automation level, 2018-2024 ($1.9M gap between extensive AI use and no automation)

Each chart card embedded in this post uses the original source as the primary attribution, with "via Statista" appended to mark the aggregation chain. The September 2025 in-depth report shares some sources with the Q2 2026 dossier (CB Insights, McKinsey, WEF), so where a chart in this post draws from both we cite the primary research source.

Frequently asked questions

What are the AI adoption trends in 2026?

Based on our Cloudflare Radar analysis covering Q1 2026, AI adoption has moved past experimentation for the top platforms. ChatGPT ranks #11 globally among all domains. AI bot traffic grew 34% in a single quarter. The GenAI service market has consolidated into a clear top 5: ChatGPT, Claude, Perplexity, DeepSeek, and Google Gemini. Training-purpose crawls grew from 42% to 52%, which means model development is accelerating, not slowing. McKinsey's survey data (72% enterprise adoption) and Deloitte's findings (worker AI access up 50%) align with what we're seeing in the traffic.

What are AI adoption trends by industry?

Our AI bot traffic analysis shows retail leads all industries at 28.2% of crawler traffic, followed by computer software (13.6%), information technology (5.8%), and internet services (5.0%). Gambling and casinos round out the top 5 at 2.8%, likely because their sites are unusually structured and data-rich. These numbers reflect where AI companies are directing their training resources. The industries getting crawled most heavily will see the most AI-powered products entering the market in the next 6-12 months.

How does AI adoption vary by country?

Our Cloudflare Radar analysis focuses on global aggregates, but two geographic signals stood out in Q1. The .ru TLD spiked to 2.16% of DNS queries in mid-March, reflecting geopolitical infrastructure shifts. And Doubao's intermittent appearance in the GenAI top 10 signals Chinese AI platforms beginning to expand globally. NVIDIA's State of AI report found North America leads with 70% actively using AI, 27% assessing, and just 3% not using it. Most top-ranked AI platforms (ChatGPT, Claude, Perplexity) are headquartered in the US with predominantly North American traffic patterns, but DeepSeek and Doubao suggest that's changing.

What are the latest AI growth statistics?

From our Q1 2026 data: AI bot traffic grew 34% in 92 days. Training-purpose crawls reached 52% of all AI bot activity. ChatGPT holds global rank #11 with more traffic than Amazon. Meta-ExternalAgent grew 43% within the quarter. Applebot surged 5x. On the service side, Claude took #2, DeepSeek reached #4, and Gemini dropped to #5. MIT Sloan Management Review's Thomas Davenport and Randy Bean identified potential AI bubble deflation as a 2026 theme. Our traffic data shows the opposite at the infrastructure level: usage is still accelerating. The disconnect might be that public market sentiment is cooling while actual adoption keeps climbing.

Where can I find an AI adoption graph for 2026?

This article contains six data visualizations covering Q1 2026: DNS query share by TLD, global domain rankings, AI bot traffic by user agent, crawler purpose distribution, generative AI service rankings, and a GTM strategy summary. Each uses Cloudflare Radar data from January 5 to April 6, 2026. For additional technology tracking data, see our industry statistics insights and technology category breakdowns.

What do recent reports say about AI adoption trends in 2026?

The major reports this cycle include McKinsey (72% enterprise adoption, 62% experimenting with AI agents), Deloitte (worker AI access up 50%, but only 20% achieving revenue growth), Microsoft (16.3% global population using GenAI), NVIDIA (70% of North American companies actively using AI), and HBR (88% regular use). Our research adds the infrastructure dimension. Real traffic flows, actual crawler behavior, daily service rankings. Where surveys tell you what companies plan to do, traffic data shows what they're actually doing today.

What percentage of companies are using AI?

It depends on how you define "using." Vention's 2026 report puts it at 93% of companies using AI in some form, with 80% using it directly in operations. McKinsey's Global Survey says 72% have adopted at least one AI capability. HBR says 88% report regular use. Our Cloudflare Radar data adds a different dimension: ChatGPT alone ranks #11 globally for web traffic, ahead of Amazon. But Deloitte's data reveals the nuance beneath those headline numbers. Only 20% of organizations are actually generating revenue from AI, despite 74% aspiring to. The gap between "using AI" and "getting business value from AI" remains wide.

How fast is the AI market growing?

Our infrastructure-level data shows AI-related internet activity growing at roughly 34% per quarter (Q1 2026), measured by AI bot traffic volume. The "other" TLD bucket that includes .ai domains grew 20% within Q1 alone. Training-purpose crawls jumped from 42% to 52% of all AI bot activity, signaling that model development is still accelerating. Microsoft's AI Economy Institute found that global generative AI adoption reached 16.3% of the world's population, up from 15.1% in H1 2025. On the services side, our data shows the top 5 GenAI platforms consolidating their positions while the long tail churns rapidly.

Which AI has the highest market share?

By actual traffic volume, ChatGPT is the clear leader. It held the #1 position among generative AI services for all 92 days of Q1 2026, and ranked #11 among all global domains. No other AI service comes close. Claude (Anthropic) held a stable #2 for 85+ consecutive days. Perplexity locked in #3. DeepSeek rose to #4. Google Gemini fell to #5. The important detail: no other AI-native domain besides chatgpt.com appears in the global top 100. The gap between #1 and #2 in generative AI is far wider than the ranking numbers suggest.

Is ChatGPT losing market share?

Not according to our traffic data. ChatGPT held the #1 position among generative AI services every single day of Q1 2026, and its global domain ranking at #11 puts it ahead of amazon.com. What's shifting is the competition below it. Google Gemini fell from #2 to #5 during Q1, losing ground to Claude and DeepSeek. But ChatGPT's position hasn't weakened. If anything, its move into the top 15 global domains suggests continued growth at a scale the other AI services haven't reached.

Why do 85% of AI projects fail?

The often-cited failure rates (85%, 90%) come from Gartner estimates about enterprise AI projects not reaching production. Our traffic data offers a partial explanation. Below the top 5 GenAI services, the rankings churn weekly: tools spike, plateau, and fade. This pattern mirrors the experimentation-to-production gap that Deloitte quantified: 66% of organizations achieved productivity improvements from AI, but only 20% are generating revenue. Only one in five companies has a mature governance model for AI agents. The problem isn't that AI doesn't work. It's that most organizations adopt faster than they can operationalize. They're running before they've figured out how to walk.

How big will the AI market be by 2031?

Statista Market Insights forecasts the worldwide AI market to grow from $94.81 billion in 2020 to $1.675 trillion by 2031 — a 17.7x expansion. The 2026 mark is $347 billion, roughly the combined revenue of the world's top five SaaS companies. The generative AI segment alone is projected to grow 80x over the same period, from $5.5 billion to $442 billion. Our Q1 2026 infrastructure telemetry (AI bot traffic +34% in a single quarter, training crawls jumping from 42% to 52% of all AI crawler activity) is consistent with that trajectory. Growth rates decelerate after 2027 as the base scales, but absolute dollar additions peak late: $1.3 trillion in new market value is forecast between 2026 and 2031 alone.

Which country uses AI the most?

By consumer usage rate, India and Nigeria tied for #1 in 2025 at 92%, according to a KPMG and University of Melbourne survey of 48,000 respondents. The UAE (91%), Egypt (90%), China (89%), and Saudi Arabia (88%) follow. Every country above 85% is classified as an emerging market — the highest-ranked advanced economy, Singapore, sits at 73%. By government AI readiness, the United States leads at 87.2, nearly ten points ahead of the UK. The consumer-usage picture and the government-capacity picture tell different stories about where AI's center of gravity lives.

What share of new unicorns are AI companies?

According to CB Insights (via Statista's Q2 2026 AI dossier), 53% of new unicorn startups worldwide in 2025 were AI companies — up from 6% in 2015 and 44% in 2024. The inflection point is 2023-2025, mapping directly to the post-ChatGPT funding wave. One in five 2025 unicorns are specifically AI agents, a category that didn't meaningfully exist two years ago. The appearance of Windsurf AI and Manus in our Q1 2026 Cloudflare Layer 4 rankings is the traffic-level confirmation of this fundraising data.

Which countries have actually deployed AI vs are still exploring it?

Vention's 2024 country survey (via Statista's September 2025 in-depth AI report) is the clearest split available. Only China (58% deployed, 30% exploring) and India (57% / 27%) have more organizations running AI in production than evaluating it. Every other country in the survey — including the U.S. (25% / 43%), UK (26% / 47%), Germany (34% / 44%), and France (31% / 44%) — has a higher exploring share than deployed share. That's a meaningful counterweight to the global 72% adoption number, which combines both groups. The world averages are 34% deployed and 42% exploring.

How much do AI cybersecurity tools actually save companies?

IBM and the Ponemon Institute (via Statista's September 2025 in-depth report) put the 2024 average breach cost at $3.8M for organizations with extensive AI security automation and $5.7M for organizations with no automation — a $1.9M gap per breach. The gap has been positive every year since 2018, peaking at $3.8M in 2021. The forecast for the AI cybersecurity market is $24.3B in 2023 growing to $133.8B by 2030, a 5.5x expansion driven primarily by this documented ROI. PwC found that 77% of organizations plan to increase their cybersecurity budget in 2025.

How much is the U.S. actually investing in AI compared to China?

As of May 2025, Spherical Insights (via Statista) tracks total U.S. AI investment at $471 billion versus China at $119 billion — roughly a 4x gap. The U.S. invests more than the next eight countries combined. The figure includes VC funding, corporate R&D, government programs (notably the CHIPS and Science Act's $280B commitment), and infrastructure spending. The disconnect: the U.S. invests 4x more than China but has less than half the AI deployment rate (25% deployed vs China's 58%), suggesting U.S. AI capital is going more into infrastructure and R&D than into enterprise production deployments.

Is AI more hype than reality?

Both, depending on which layer you're measuring. At the DNS level (Layer 1), there's genuine speculation: 10.74% of DNS queries hit non-existent domains, and the "other" TLD bucket (including .ai) grew 20% in a quarter. That's hype activity. But at the traffic level (Layer 2), ChatGPT at global rank #11 is not hype. That requires hundreds of millions of active users. At the bot level (Layer 3), training crawls growing from 42% to 52% shows real infrastructure investment, not just announcements. And at the service ranking level (Layer 4), the top 5 GenAI platforms held stable positions for months. MIT Sloan's Thomas Davenport flagged "AI bubble deflation" as a 2026 theme. Our data says the bubble may be deflating in public markets, but the underlying usage is still accelerating.