The State of AI Search in 2026: 150+ Statistics on Adoption, Traffic, Citations, and What Actually Works

AI search has become a parallel discovery layer used by hundreds of millions of people every week, reshaping how content gets found, cited, and clicked. In the span of two years, we went from “Will anyone actually use ChatGPT for search?” to Google AI Mode crossing one billion monthly users and B2B buyers building vendor shortlists inside LLMs before they ever visit a website.

Yet the full picture is more nuanced than either the hype or the panic suggests. AI tools account for just 3.2% of all US desktop search activity. Roughly 95% of ChatGPT users still rely on Google. The click economy is being rewritten while the rules for who gets cited are still being discovered.

This post collects over 150 statistics on the state of AI search in 2026, sourced from original studies by Pew Research Center, Forrester, Gartner, McKinsey, Google, OpenAI, Adobe, Cloudflare, Ahrefs, Semrush, and dozens of others. Every figure is linked to the primary source with the study date. Where the data intersects with our own work, we also include findings from OrganikPI’s citation research: three studies analyzing over 196,000 AI citations to understand what actually drives visibility in AI search engines.

Every stat below links directly to its primary source. Let’s get into it.

AI search in 2026, in 10 numbers

1. 800 million weekly active users on ChatGPT as of December 2025 (OpenAI, 2025)

2. 1 billion monthly users on Google AI Mode, one year after launch (Google, 2026)

3. 3.2% of all US desktop search activity comes from AI tools combined (SparkToro / Datos, 2026)

4. 58% lower clickthrough rate for position-1 pages when an AI Overview appears (Ahrefs, 2026)

5. 76.95% of URLs cited by AI engines are NOT in the organic top 10 (OrganikPI original research, 2026)

6. 11.4% conversion rate for ChatGPT-referred ecommerce visitors, versus 5.3% from organic search (Similarweb, 2025)

7. 94% of B2B buyers used generative AI during their buying process in 2025 (Forrester, 2026)

8. 693.4% year-over-year growth in AI referral traffic to retail sites during the 2025 holiday season (Adobe, 2026)

9. 97% of llms.txt files received zero traffic in May 2026 (Ahrefs, 2026)

10. $2.52 trillion in worldwide AI spending forecast for 2026 (Gartner, 2026)

Adoption and usage

Half the US population now uses AI chatbots, and engagement keeps deepening. The growth trajectory is global, cross-generational, and accelerating fastest in developing markets.

• ~50% of US adults now use AI chatbots, up from a third in 2024 (Pew Research Center, 2026)
• 60% of US adults say they read AI search engine summaries at the top of search results (Pew Research Center, 2026)
• ~40% of US adults use chatbots for information searching, the most common use case (Pew Research Center, 2026)
• 800 million weekly active users on ChatGPT as of December 2025 (OpenAI, 2025)
• 2.5 billion prompts sent daily on ChatGPT globally, with 330 million daily prompts from US users alone (Axios / OpenAI, 2025)
• 50% more messages per day sent by ChatGPT users six months after signing up, along with double the number of distinct tasks tried (OpenAI, 2026)
• 26%+ of US workers use ChatGPT for work, rising to 45% among those with postgraduate degrees (OpenAI, 2026)
• 58% of employees globally reported using AI at work on a semiregular or regular basis in 2025, exceeding 80% in India, China, Nigeria, the UAE, Egypt, and Saudi Arabia (Stanford HAI, 2026)
• Weekly use of generative AI tools nearly doubled from 18% to 34% across six countries between 2024 and 2025 (Reuters Institute, 2025)
• 54% of respondents across six countries said they saw an AI-generated search answer in the past week (Reuters Institute, 2025)
• 10%+ increase in Google usage for query types that trigger AI Overviews, in the US and India (Google, 2025)
• 5 trillion searches processed annually by Google, equivalent to roughly 13.7 billion per day (Google, 2025)

The engine landscape and market share

ChatGPT dominates the standalone AI chatbot market, but the landscape is becoming more crowded. Google’s AI Mode is scaling rapidly inside Search, and Chinese platforms are carving out significant mobile audiences.

• 76.87% worldwide AI chatbot market share held by ChatGPT as of June 2026, with Gemini at 7.94% and Perplexity at 7.91% (StatCounter, 2026)
• ChatGPT is 2.7x larger than #2 Gemini on web traffic and 2.5x larger on mobile by monthly active users (a16z, 2026)
• Gemini ranked #2 on web with approximately 12% of ChatGPT’s visits; Google AI Studio debuted in the web top 10 (a16z, 2025)
• Grok ranked #4 on web and reached over 20 million monthly active users on mobile after launching its app in late 2024 (a16z, 2025)
• Chinese AI products ranked prominently in the a16z Top 100: Quark at #9 on web, Doubao at #12 on web (#4 on mobile), and Kimi at #17 on web (a16z, 2025)
• AI platforms generated more than 1.1 billion referral visits in June 2025, up 357% year over year (Similarweb, 2025)
• Gen AI average monthly visits grew 76% year over year, and app downloads surged 319% YoY (Similarweb, 2025)
• 95% of ChatGPT users still rely on Google, so for most people AI layers on top of search (Similarweb, 2025)
• Google AI Mode surpassed 1 billion monthly users one year after launch, with queries more than doubling every quarter (Google, 2026)
• 73.7% of all US desktop web searches still go through Google, while all AI tools combined account for just 3.2% (SparkToro / Datos, 2026)
• Over 20% of Americans are heavy AI tool users (10+ times per month), yet 95%+ still use traditional search engines monthly (SparkToro / Datos, 2025)
• US AI tool adoption nearly quintupled from 8% to 38% over 2.5 years, while traditional search engine usage declined by less than 1 percentage point (SparkToro / Datos, 2025)
• Perplexity raised $100 million at an $18 billion valuation in July 2025 (Bloomberg, 2025)

AI chatbot use for news

A growing slice of the population is using AI for news, though trust remains low and the behavior skews young.

• Weekly use of AI chatbots for news grew from 7% to 10% globally between 2025 and 2026 (Reuters Institute, 2026)
• Only 1% of people globally say AI chatbots are their main source of news; trust in news from AI chatbots stands at just 20% (Reuters Institute, 2026)
• 17% of 18-to-24-year-olds use AI chatbots for news weekly, three times the rate of the oldest age group at 5% (Reuters Institute, 2026)
• Only 4% of respondents across 27 markets say they always or often click through to news sources from AI chatbots, compared with 19% from search and 17% from social media (Reuters Institute, 2026)

How common AI Overviews have become

AI Overviews are now a fixture in search results, though their prevalence varies wildly by query type, industry, and device. Google has also been experimenting with their length and format.

• AI Overviews appear for 21% of all keywords across Ahrefs’ 146 million SERP database (Ahrefs, 2025)
• AI Overview presence grew from roughly 30% to 48% of tracked queries from February 2025 to February 2026, a 58% increase (BrightEdge, 2026)
• 25.11% of 21.9 million Google searches analyzed by Conductor triggered an AI Overview result (Conductor, 2026)
• AI Overviews appeared in over 11% of Google queries as of May 2025, a 22% increase since their debut in May 2024 (BrightEdge, 2025)
• 30% of US desktop keywords triggered an AI Overview in September 2025, a roughly 492% surge from September 2024 (seoClarity, 2025)
• AI Overviews on mobile US searches increased nearly 475% year over year from September 2024 to September 2025 (seoClarity, 2025)
• In the US, 25% of keywords show the AI Mode option, while AI Overviews appear for 29% of the same keyword set (seoClarity, 2026)
• The Gemini app reached over 400 million monthly active users as of May 2025 (Google, 2025)
• The Gemini app grew to more than 450 million monthly active users by Q2 2025, with daily requests up over 50% from Q1 (Google, 2025)
• Google processed over 480 trillion tokens per month as of May 2025, a 50x increase year over year from 9.7 trillion (Google, 2025)
• Google’s monthly token processing doubled to over 980 trillion by July 2025 (Google, 2025)
• The average length of AI Overviews dropped 70%, from approximately 5,300 characters in July 2025 to just 1,600 characters in August 2025 (seoClarity, 2025)

AIO prevalence by query type

AIO prevalence by category

AIO prevalence over time (Semrush, 10M+ keywords)

Source: Semrush, 2025

The click and traffic impact

This is the most contested area in AI search data. Independent studies show significant CTR declines for queries where AI Overviews appear. Google disputes the severity, pointing to stable total click volume and improved click quality. Both sides deserve a fair read.

The decline side

• The presence of an AI Overview correlates with a 58% lower average CTR for position-1 pages as of December 2025, nearly double the 34.5% reduction measured eight months earlier (Ahrefs, 2026)
• When a Google AI Overview appeared, users clicked a traditional result link in only 8% of visits, compared to 15% on pages without an AI summary (Pew Research Center, 2025)
• Users clicked a link within the AI summary itself in just 1% of visits (Pew Research Center, 2025)
• After encountering an AI Overview, 26% of users ended their browsing session entirely, compared to 16% after standard search results (Pew Research Center, 2025)
• Organic CTR for AIO queries fell 61% from June 2024 (1.76%) to September 2025 (0.61%) across 3,119 informational search terms tracked across 42 client organizations (Seer Interactive, 2025)
• Paid CTR for AIO queries fell 68% from June 2024 (19.70%) to September 2025 (6.34%) (Seer Interactive, 2025)
• Google search impressions rose 49% year over year since the launch of AI Overviews, but click-throughs fell nearly 30% over the same period (BrightEdge, 2025)

The citation advantage

• Brands cited in an AI Overview earn 35% more organic clicks and 91% more paid clicks compared to brands not cited (Seer Interactive, 2025)

Google’s counter-narrative

Google has publicly pushed back on the third-party decline reports, calling them inaccurate.

• Google stated that total organic click volume from Search to websites has been “relatively stable year-over-year” as of August 2025 (Google / Liz Reid, 2025)
• Google said average click quality from AI Overview pages has increased, and Google is sending “slightly more quality clicks” to websites than a year ago (Google / Liz Reid, 2025)
• Google explicitly disputed third-party reports of dramatic organic traffic declines, calling them based on “flawed methodologies, isolated examples, or traffic changes that occurred prior to the roll out of AI features in Search” (Google / Liz Reid, 2025)

Desktop vs mobile CTR divergence

• In Q1 2026, desktop CTRs for the top five positions rose by a combined 10.54 percentage points quarter over quarter, while position-1 mobile CTR fell 2.20 percentage points, deepening a device-type split in click behavior (Advanced Web Ranking, 2026)

The honest read: CTR is declining on queries where AI Overviews appear. Google’s claim that total organic clicks are stable is not necessarily contradictory: if AI Overviews drive more total searches (which Google confirms), the per-query CTR can drop while aggregate clicks hold. Both things can be true simultaneously.

Who gets cited, and why

Citation behavior varies dramatically across platforms. Wikipedia dominates everywhere, but each engine has its own source preferences, and the overlap between them is surprisingly low.

• Wikipedia is the most cited domain across all three major AI assistants: ChatGPT at 16.3%, Perplexity at 12.5%, and Google AI Overviews at 8.4% (Ahrefs, 2025)
• Only 14% of the top 50 most-cited sources are shared across ChatGPT, Perplexity, and Google AI Overviews. 86% of top sources are platform-exclusive (Ahrefs, 2025)
• Wikipedia is the most cited domain in Google AI Mode with over 1.1 million mentions, followed by YouTube with 961,938 mentions, across 5.5 million AI Mode queries (Ahrefs, 2025)
• Within Perplexity’s top 10 cited sources, Reddit accounts for 46.7% of citation share; within ChatGPT’s top 10, Wikipedia accounts for 47.9%; within Google AI Overviews’ top 10, Reddit leads at 21% (Profound, 2025)
• ChatGPT search routes queries through Bing (and Shopify) as its named third-party search providers, making Bing index visibility directly relevant to ChatGPT citation exposure (OpenAI, 2024)
• AI assistants cite content that is on average 25.7% fresher than pages in organic Google search results, based on 17 million citations analyzed across seven platforms (Ahrefs, 2025)
• ChatGPT shows the strongest freshness preference, citing URLs that are on average 458 days newer than pages appearing in organic search (Ahrefs, 2025)
• Google AI Overviews are the outlier on freshness: they prefer to cite content that is on average 16 days older than the standard organic SERP (Ahrefs, 2025)
• AI citation sets are highly stable week over week: 96.8% of cited domains and 97.2% of mentioned brands saw zero change (BrightEdge, 2026)
• When citation changes do occur, 87% are declines rather than gains, and over 51% of all citation volume was associated with declining domains (BrightEdge, 2026)
• Citation volume is heavily concentrated: the top 1% of domains capture 64% of all citations, and the top 10% capture 84% (BrightEdge, 2026)
• Across six AI platforms, AI assistants cited a link when mentioning the Ahrefs brand only 28% of the time on average, meaning brands are mentioned without a hyperlink roughly 72% of the time (Ahrefs, 2025)
• When weighted by impressions, linked mentions account for 78.4% of impressions on Perplexity and 71.3% on Gemini, because citation links cluster on higher-traffic queries (Ahrefs, 2025)

How AIO citations overlap with organic rankings

Two different measurements get conflated here, and separating them matters. The share of AI Overview citations that overlap with organic rankings *anywhere* has risen. The share that comes from the organic *top 10* specifically has fallen. Both are true at once.

Overlap with any organic ranking, one provider tracking the same way over 16 months:

Overlap with the organic top 10 specifically:

Put the two together and the story is clear: AI Overviews increasingly cite pages that rank organically somewhere, but pull less and less from page 1 specifically. Ranking in the top 10 is neither necessary nor sufficient to get cited. Absolute levels differ by provider and method, so read each series for its trend rather than as one continuous line.

• Overlap varies sharply by industry: Healthcare sits at 75.3% and Education at 72.6%, while E-commerce is just 22.9% (BrightEdge, 2025)
• Across ChatGPT, Gemini, and Copilot, only about 12% of cited URLs rank in Google’s top 10, while Perplexity’s overlap is 28.6% (Ahrefs, 2025)

Original research: what OrganikPI found about AI citations

Our own citation studies add a dimension most third-party research does not cover: what the cited content actually looks like at the sentence level.

Study A: We Decoded 42,971 AI Citations (March 2026)

Our analysis of 42,971 AI citations found that cited sentences cluster in the top 35% of the page, with a mean position of 34.9% down the page (median 31.2%) across 1,719 matched sentences. The typical cited sentence is 6 to 17 words long. Citation word-length split: 7.6% at 1-5 words, 43.0% at 6-10 words, and 49.4% at 11-20 words. Pages with structured content had a 91.3% sentence-match rate versus 39.3% for unstructured pages.

Study B: Positional Bias in AI Citations (May 2026, 153,425 citations)

In a larger follow-up study of 153,425 citations, the mean cited-sentence position was 37% through the document. 74.9% of cited sentences appeared in the first half of the page. The bottom quarter of a page received only 7.4% of citations. Sentences of 6-10 words accounted for 45.2% of all citations. And the headline finding: 76.95% of cited URLs were NOT in the organic top 10 (only 23.05% were).

Study C: Google Killed Text Fragments in AI Mode (May 2026, 9,968 matched citations)

From the same 153,425-citation dataset, we matched 9,968 citations to specific pages. 84.1% of Gemini citation URLs now carry a text fragment (up from 51.8%). The platforms share only 4.66% of cited domains with each other. Cited sentences split bimodally: 20.5% very hard to read, 22.9% very easy, only 2.6% mid-range. Perplexity had 50.3% domain overlap with the organic top 10, while only 4.2% of ChatGPT citations ranked in Google’s top 10. AI Mode produced 57.6% of citations and Grok 20.0%. YouTube alone drew 9,868 citations.

AI referral traffic quality and conversion

AI referral traffic is still tiny in absolute terms, but it converts at a rate that demands attention. The visitors arriving via AI chatbots behave differently from organic search traffic: they stay longer, browse more, and buy more.

• Ecommerce visits referred by ChatGPT convert at 11.4%, more than double the 5.3% conversion rate from organic search (Similarweb, 2025)
• In a single-client case study, ChatGPT traffic converted at 15.9% compared to Google Organic at 1.76% (Seer Interactive, 2025)
• The average AI search visitor is 4.4x as valuable as the average traditional organic search visitor based on conversion rate (Semrush, 2025)
• Traffic to US retail websites from generative AI sources increased 1,200% between July 2024 and February 2025, doubling roughly every two months (Adobe, 2025)
• Generative AI traffic to US retail sites surged 4,700% year over year in July 2025, with AI-driven revenue per visit rising 84% from January to July 2025 (Adobe, 2025)
• During the 2025 holiday season, AI referrals converted 31% more than other traffic sources, with AI-driven revenue per visit up 254% year over year (Adobe, 2026)
• Retail traffic from generative AI tools grew 693.4% year over year during the 2025 holiday season. Travel was up 539%, financial services up 266% (Adobe, 2026)
• Shoppers from AI sources spent 45% more time on-site and viewed 13% more pages per visit during the 2025 holiday season (Adobe, 2026)
• AI referral visitors to retail sites had a 23% lower bounce rate, viewed 12% more pages, and spent 41% longer on site compared to non-AI traffic as of February 2025 (Adobe, 2025)
• By May 2025, AI referral traffic had a 27% lower bounce rate, 38% longer time per visit, and 10% more page views per visit (Adobe, 2025)
• ChatGPT referral visitors averaged 2.3 pages per session versus 1.2 pages for Google Organic, with a 62% engagement rate versus 60% for organic (Seer Interactive, 2025)
• AI makes up just 0.1% of total web referral traffic; Google sends 345x more traffic than the three main AI platforms combined (Ahrefs, 2025)
• AI represents 0.13% of total sessions site-wide across 1.96 million LLM sessions analyzed, but reaches 0.46% penetration on pricing pages (3.5x the site average) (Previsible, 2025)
• ChatGPT owned 84.2% of AI referral sessions, growing 3.26x year over year. Copilot grew 25.2x and Claude grew 12.8x (Previsible, 2025)
• AI traffic accounted for just 0.07% of one client’s organic traffic, yet drove 100% more attributable conversions year over year (Seer Interactive, 2025)
• E-commerce AI traffic penetration jumped 67% month over month in November 2025 during peak holiday shopping, growing 10.5x year over year (Previsible, 2025)
• AI referral visitors view 4 pages per session on average versus 5.2 for search visitors, and have a 67.8% bounce rate versus 63.7% for search, but spend 86 seconds on site versus 78 for search (Ahrefs, 2025)
• Web traffic from AI-driven referrals increased more than tenfold in the US between July 2024 and February 2025 (Adobe, 2025)

Who is using AI search (demographics and behavior)

AI search adoption is broad but not uniform. Age is the strongest predictor of both usage intensity and attitude.

• Adults under 50 are about twice as likely as those 50+ to use ChatGPT: 57% vs. 28% (Pew Research Center, 2026)
• About a third of US adults ages 30 to 49 use AI chatbots daily, compared with just 7% of those 65 and older (Pew Research Center, 2026)
• One in five US adults under 30 (20%) have used AI chatbots for emotional support or advice (Pew Research Center, 2026)
• Two thirds of US teens ages 13 to 17 (64%) say they use AI chatbots (Pew Research Center, 2025)
• About three in ten US teens use AI chatbots every day, including 16% who do so several times a day or almost constantly (Pew Research Center, 2025)
• More than half of US teens have used chatbots to search for information (57%) and over half to get help with schoolwork (54%) (Pew Research Center, 2026)
• About half of US Gen Z (ages 14 to 29) use generative AI at least weekly (51%), with 22% using it daily (Gallup, 2026)
• Gen Z excitement about AI dropped 14 percentage points to 22% in 2026, while anger increased 9 points to 31% (Gallup, 2026)
• The gender gap in US chatbot use has closed: men and women now use AI chatbots at nearly equal rates (50% vs. 47%), compared with an 11-point gap in 2024 (Pew Research Center, 2026)
• More than half (52%) of 18-to-24-year-olds globally say social media, video networks, and AI chatbots are now their main way of getting news (Reuters Institute, 2026)
• AI chatbot news users are disproportionately heavy news consumers: 38% are classified as news lovers, compared with 22% of all respondents (Reuters Institute, 2026)
• 7% of online news consumers globally use AI chatbots for news weekly, rising to 15% among those under 25 (Reuters Institute, 2025)

How B2B buyers use AI search

This is where AI search stops being a consumer curiosity and becomes a revenue problem (or opportunity). B2B buyers are building vendor shortlists inside these tools before they ever contact sales.

• 94% of B2B buyers used generative AI or conversational search during their buying process in 2025, up from 89% the prior year (Forrester, 2026)
• 61% of business buyers report using private AI tools provided by their organization, and they are twice as likely to use ChatGPT as consumers (Forrester, 2026)
• 95% of B2B buyers plan to use generative AI in at least one area of a future purchase, and over half say it led them to consider more or different vendors (Forrester, 2025)
• 87% of B2B buyers who used generative AI in their purchasing process agreed it helped them create a better business outcome (Forrester, 2024)
• 45% of B2B buyers used generative AI during a recent purchase, primarily to gather information on vendors and products (Gartner, 2026)
• 69% of B2B buyers prefer to validate AI-generated insights with sales reps rather than acting on them directly (Gartner, 2026)
• 51% of B2B software buyers now start their research with an AI chatbot more often than with Google, up from 29% seven months earlier (G2, 2026)
• 71% of B2B software buyers rely on AI chatbots at some point in their research process (G2, 2026)
• 69% of B2B software buyers chose a different vendor than initially planned because that vendor was part of an AI chatbot’s recommendation (G2, 2026)
• 85% of B2B buyers think more highly of a vendor when an AI chatbot includes them in a recommended answer (G2, 2026)
• 53% of B2B software buyers say research done with AI chatbots is more productive than traditional search, up from 36% seven months earlier (G2, 2026)
• 94% of B2B buyers in 6sense’s 2025 study used LLMs during their buying journey, yet those buyers still had an average of 16 interactions with the winning vendor (6sense, 2025)
• 72% of B2B tech buyers encountered Google AI Overviews during their product research, and 90% clicked through to one of the cited sources (TrustRadius, 2025)
• 44% of US online buyers start their purchase journey in an LLM or split between AI tools and traditional search engines (Bain, 2026)
• SMB buyers have already started building vendor shortlists inside LLMs, with AI constructing the consideration set before any website validation (Bain, 2026)
• Gartner forecasts that by 2028, 90% of B2B buying will be AI agent intermediated, pushing over $15 trillion of B2B spend through AI agent exchanges (Gartner, 2025)
• Bain estimates fully agentic commerce could reach $300 billion to $500 billion in US revenues by 2030, representing up to one-quarter of total e-commerce (Bain, 2026)

The impact on publishers and media

Publishers are caught between rising AI-driven content consumption and declining referral traffic. The fear is existential, even if the timeline is uncertain.

• 74% of media leaders said they were worried about a potential decline in referral traffic from search engines in 2025 (Reuters Institute, 2025)
• Publishers expect traffic from search engines to almost halve, forecasting a 43% decline over the next three years (Reuters Institute, 2026)
• About 58% of US adults conducted at least one Google search that produced an AI-generated summary in March 2025 (Pew Research Center, 2025)
• 60% of people do not yet regularly encounter audience-facing AI features on news websites and apps, despite 77% consuming news daily (Reuters Institute, 2025)

Named cases: publishers putting numbers on the damage

The macro anxiety is now showing up in named companies’ own financial disclosures.

• Chegg told investors its non-subscriber traffic fell 8% year over year in Q2 2024, 19% in Q3, and 37% in October 2024, and named Google AI Overviews directly: the company said it “would not need to review strategic alternatives if Google hadn’t launched AI Overviews.” Q4 2024 revenue fell 24% (Chegg SEC filing, 2024)
• At Dotdash Meredith (People Inc.), parent company IAC told investors that Google Search’s share of its sessions fell from 52% to 28% over three years (IAC Q2 2025 earnings call)
• Across 2,500+ publisher sites, Google search referrals fell 33% globally in the year to November 2025, and 38% for US organic search specifically (Chartbeat via Reuters Institute / Press Gazette, 2026)
• Business Insider’s organic search traffic fell 55% between April 2022 and April 2025; HuffPost’s fell by more than half over the same period (WSJ citing Similarweb, 2025)
• A survey of 19 premium publishers by Digital Content Next found median Google referral traffic down 10% in just eight weeks (May to June 2025), with declines outnumbering gains two to one (Digital Content Next, 2025)

The crawl-to-refer imbalance

AI platforms crawl far more content than they send back in referral traffic. The imbalance is measured in crawl-to-refer ratios: how many pages a platform crawls for every single visitor it refers back to a website.

• Anthropic has been the most crawl-heavy platform all year: its ratio was as extreme as 286,930:1 back in January 2025 before falling over the following months (Cloudflare, 2025)
• Training traffic accounts for nearly 80% of all AI bot crawling activity, up from 72% a year earlier (Cloudflare, 2025)
• Google referrals to news sites were 15% lower in April 2025 than in January, as AI Overviews expanded (Cloudflare, 2025)

The licensing and legal economy

As referral traffic falls, two new markets have sprung up around AI and publisher content: one where AI companies pay to license it, and one where publishers sue over it.

On the licensing side:

• Reddit disclosed in its IPO filing that it signed data-licensing deals worth $203 million in aggregate contract value in January 2024, with a minimum $66.4 million to be recognized in 2024 alone (Reddit SEC S-1, 2024). It remains the only major AI-content deal with a dollar figure confirmed by a party rather than the press.
• News Corp and OpenAI signed a multi-year global partnership in May 2024, reported at more than $250 million over five years (terms not officially disclosed by either party) (News Corp, 2024)
• Amazon licensed New York Times content in 2025, reported at $20 to $25 million a year (the Times stated terms were not disclosed) (The New York Times, 2025)
• Perplexity’s Publishers Program shares a double-digit percentage of revenue with cited publishers (Perplexity, 2024)
• OpenAI alone has signed content deals with Axel Springer, the Financial Times, Associated Press, Condé Nast, Hearst, Vox Media, The Atlantic, Le Monde, Prisa Media, and Guardian Media Group, most with undisclosed terms (OpenAI, 2024)

On the legal side:

• More than 70 copyright lawsuits had been filed against AI companies in US courts by the end of 2025, up from around 30 a year earlier (Copyright Alliance, 2026)
• Anthropic agreed to a $1.5 billion settlement covering roughly 482,460 books, about $3,000 per title, the largest copyright settlement of the AI era, preliminarily approved in September 2025 (Settlement administrator, 2025)
• The landmark New York Times v. OpenAI and Microsoft case (filed December 2023) is active, with the court narrowing it to fair-use questions in April 2025 (CourtListener docket, 2025)
• In Thomson Reuters v. Ross Intelligence, a federal court ruled in February 2025 that Ross’s copying of 2,243 Westlaw headnotes was not fair use (case analysis, 2025). In the UK, Getty Images largely lost its case against Stability AI in November 2025 after dropping its main training-copyright claims (case analysis, 2025)

AI crawlers and the new bot economy

The infrastructure layer of AI search is an arms race. Crawling volumes are surging, bot populations are proliferating, and publishers are starting to fight back.

• AI and search crawler traffic grew 18% from May 2024 to May 2025 across a fixed cohort of Cloudflare customers, with GPTBot raw requests surging 305% (Cloudflare, 2025)
• GPTBot’s share of all AI-only crawler traffic grew from 11.9% to 28.1% from July 2024 to July 2025, making it the top AI-only crawler (Cloudflare, 2025)
• Googlebot’s crawling volume grew 96% year over year, and its share of all AI and search crawler traffic rose from 30% to 50% (Cloudflare, 2025)
• Google’s crawl-to-refer ratio worsened from 3.8:1 in January 2025 to a peak of 22.5:1 in April 2025, a nearly 6x deterioration in a single quarter (Cloudflare, 2025)
• Among the top 10,000 domains with a robots.txt file, only about 14% had directives specifically targeting AI bots (Cloudflare, 2025)
• GPTBot is the most blocked AI bot, with 5.89% of approximately 140 million analyzed websites blocking it. ClaudeBot saw the fastest growth in block rates at 32.67% year over year (Ahrefs, 2025)
• The number of active AI bots more than doubled in just over a year, growing from 10 to 21 bots between August 2023 and December 2024 (Ahrefs, 2025)
• Automated traffic grew 8x faster than human traffic in 2025, with AI-driven traffic volumes growing 187% from January to December 2025 (HUMAN Security, 2026)
• Agentic AI traffic grew 7,851% year over year in 2025, with 77% of agentic activity concentrated on product and search pages (HUMAN Security, 2026)
• OpenAI’s bots accounted for approximately 69% of all observed AI-driven traffic in 2025, with Meta contributing 16% and Anthropic 11% (HUMAN Security, 2026)
• AI scraper traffic grew 597% from January to December 2025, with scrapers representing 31.9% of all observed AI bot traffic (HUMAN Security, 2026)
• Cloudflare launched Pay Per Crawl in private beta on July 1, 2025, enabling content owners to charge AI crawlers a per-request price using HTTP 402 Payment Required responses (Cloudflare, 2025)

Enterprise adoption and GEO/AEO spend

Enterprises are not waiting for the data to settle. The shift in budgets toward AI search optimization is already visible in survey data and spending forecasts.

• 88% of organizations regularly used AI in at least one business function as of mid-2025, up from 78% a year prior (McKinsey, 2025)
• 71% of organizations regularly used gen AI in at least one business function as of mid-2024, up from 65% in early 2024 (McKinsey, 2025)
• 97% of CMOs and digital leaders at enterprise companies reported that AEO had a positive impact on their marketing funnel in 2025 (Conductor, 2026)
• 94% of CMOs and digital leaders plan to increase their AEO investment in 2026 (Conductor, 2026)
• Enterprises allocated an average of 12% of their digital budgets to AEO in 2025 (Conductor, 2026)
• 68% of organizations are actively changing their search strategies in response to AI search (BrightEdge, 2025)
• 54% of organizations have tasked their SEO or digital marketing teams to lead AI search initiatives (BrightEdge, 2025)
• 78% of surveyed companies fund GEO programs, nearly matching the 77.5% that invest in traditional SEO and PPC, marking budget parity for the first time (Clutch, 2025)
• 63% of marketers plan to increase their GEO investment in the next 12 months, including 26% who expect significant increases (Clutch, 2025)
• Gartner predicted in February 2024 that traditional search engine volume will drop 25% by 2026 as AI chatbots displace queries (Gartner, 2024)
• Semrush projects AI search visitor volumes will surpass traditional organic search visitors for digital marketing topics by early 2028 (Semrush, 2025)
• Worldwide spending on AI is forecast to total $2.52 trillion in 2026, a 44% increase year over year, with AI infrastructure alone accounting for $1.37 trillion (Gartner, 2026)
• Goldman Sachs projects $765 billion in annual AI capital expenditure in 2026, scaling to $1.6 trillion by 2031 and totaling roughly $7.6 trillion cumulatively from 2026 to 2031 (Goldman Sachs, 2026)
• Wall Street consensus for 2026 capex by the four largest AI hyperscalers (Google, Microsoft, Amazon, Meta) reached $527 billion (Goldman Sachs, 2025)

What actually drives AI citations

This is where the signal separates from the noise. The biggest misconceptions in GEO are about schema markup and llms.txt. The data is clear: neither has a measurable causal effect on AI citations.

The schema reality check

• 53% of AI-cited pages carry schema markup, but a controlled study of 1,885 pages tracked over 7 months found adding JSON-LD schema produced no statistically meaningful uplift in AI citations on any platform (Ahrefs, 2026)
• Specifically: Google AI Mode showed a +2.4% change and ChatGPT a +2.2% change, both indistinguishable from zero (Ahrefs, 2026)

The takeaway: schema correlates with citation because well-built sites tend to have both schema and good content. Adding schema to a mediocre page does not make AI engines cite it.

The llms.txt reality check

• 97% of llms.txt files across 137,000 domains received zero bot or human traffic in May 2026, and AI retrieval bots accounted for just 1.1% of the small fraction of requests that did reach any llms.txt file (Ahrefs, 2026)
• An XGBoost model trained on 300,000 domains found that removing the llms.txt variable actually improved prediction accuracy for AI citation frequency, indicating the file has no positive effect (SE Ranking, 2025)

What does correlate with AI visibility

• YouTube mentions show the strongest correlation with AI visibility across ChatGPT, AI Mode, and AI Overviews, with a correlation coefficient of approximately 0.737 (Ahrefs, 2025)
• Branded web mentions correlate with AI visibility at 0.66 to 0.71 across platforms, the second strongest factor after YouTube mentions (Ahrefs, 2025)

What the citation probability curve looks like

• 94% of Google AI Overviews include at least one URL from the top 20 organic results (seoClarity, 2025)
• GEO content methods (Cite Sources, Quotation Addition, Statistics Addition) improve AI search visibility by 30-40% compared to unmodified content (Princeton / IIT Delhi, 2023)
• The best GEO optimization methods improve AI visibility by up to 41% on Position-Adjusted Word Count (Princeton / IIT Delhi, 2023)
• Applying the Cite Sources method to rank-5 websites produces a 115% visibility uplift in generative engine responses (Princeton / IIT Delhi, 2023)

The actionable core: OrganikPI’s positional and structural findings

Our own research points to what actually matters at the content level:

• Cited sentences cluster in the top 35% of the page. The bottom quarter receives only 7.4% of citations. If your key claims are buried below the fold, AI engines will likely never cite them.
• Structured content had a 91.3% sentence-match rate versus 39.3% for unstructured pages. By structure, I mean clear headings, short paragraphs, and definitive statements near the top.
• The typical cited sentence is 6 to 17 words long: a clean, quotable statement of fact.
• 76.95% of cited URLs were not in the organic top 10. You do not need to rank on page 1 to get cited by AI engines, but you do need content that is structured, quotable, and factually specific.

What this means for your GEO strategy

Every takeaway here traces directly to a number above.

1. Front-load your key claims. 74.9% of cited sentences come from the first half of the page, and the bottom quarter gets just 7.4% of citations. Put your most quotable, factually specific statements above the fold.

2. Prioritize content structure over schema markup. Pages with structured content (clear headings, short paragraphs, definitive statements) had a 91.3% citation match rate. Adding JSON-LD schema to unstructured content produced zero measurable uplift. Spend the effort on content architecture.

3. Optimize for multiple engines separately. Only 14% of the top 50 most-cited sources are shared across ChatGPT, Perplexity, and Google AI Overviews. Each platform has its own source preferences. A one-size-fits-all approach will underperform.

4. Judge AI referral traffic by how it converts. AI traffic is still small (0.1% of total referrals), but ChatGPT visitors convert at 11.4% on ecommerce sites versus 5.3% for organic search. The ROI calculation favors treating AI traffic as a high-intent channel.

5. Accept that ranking on page 1 is not required for AI citation. 76.95% of cited URLs in our dataset were not in the organic top 10. Citation depends on content quality and structure, not SERP position alone.

6. Don’t waste time on llms.txt. 97% of llms.txt files received zero traffic, and citation-prediction models get more accurate when the llms.txt signal is removed. Focus your efforts elsewhere.

7. Monitor the B2B buyer shift. 51% of B2B software buyers now start research in an AI chatbot before Google, and 69% chose a different vendor because it appeared in a chatbot’s recommendation. If you sell to businesses, your AI visibility is already affecting pipeline.

8. Keep content fresh. AI assistants cite content that is 25.7% fresher than organic SERP results on average. Regular updates to core pages improve your citation odds, especially on ChatGPT (which shows the strongest freshness preference at 458 days newer than organic).

About this data

This post includes more than 150 distinct statistics, each linked to the original study that published it. The data spans late 2023 to June 2026, with the majority drawn from 2025 and 2026 studies.

OrganikPI’s original figures come from three proprietary citation studies conducted in March and May 2026, analyzing a combined dataset of over 196,000 AI citations across ChatGPT, Perplexity, Google AI Mode, Google AI Overviews, Gemini, and Grok. These studies are the basis for the positional bias, word-length, structure, and platform-overlap findings featured throughout this post.

Sources include: Pew Research Center, Forrester, Gartner, McKinsey, Google, OpenAI, Adobe Digital Insights, Similarweb, Cloudflare, Ahrefs, Semrush, seoClarity, BrightEdge, Reuters Institute, Stanford HAI, Bloomberg, Goldman Sachs, a16z, SparkToro/Datos, Conductor, Clutch, G2, 6sense, TrustRadius, Bain, Seer Interactive, Previsible, Advanced Web Ranking, HUMAN Security, Profound, SE Ranking, Gallup, StatCounter, and Princeton University/IIT Delhi.