The State of AI Adoption and AI Search

The short version

AI adoption crossed a line in the last two years. In McKinsey's annual State of AI survey, which tracks large companies, the share using AI in at least one part of the business sat near half for most of a decade. Then generative AI arrived. The number reached 78 percent by mid-2024 and 88 percent in 2025.¹ Among the organizations that survey represents, using AI in at least one function is now the norm; across all businesses, and small firms especially, adoption remains substantially lower.

The harder number is what companies get for it. In the same McKinsey survey, only about 6 percent tie a meaningful share of profit to AI, and just 7 percent say they have scaled it across the whole organization.¹ BCG's Widening AI Value Gap study reads the same picture from a different angle: roughly 5 percent of companies capturing value at scale, and about 60 percent seeing little material return.²⁷ Experimentation is widespread in that population; enterprise-wide scale and meaningful earnings impact remain a small minority. That gap between using AI and being paid by it is the real story of 2026, and it is where the next five years of work sits.

How big adoption looks depends heavily on who you count. McKinsey surveys large enterprises, so it reads near 90 percent. The US Census surveys every business including the smallest, so it reads closer to 20 percent.³¹ Both are right. Throughout this report we name the population behind each number, because a single headline figure hides more than it shows.

On the demand side, people moved faster than their employers. Pew Research Center's Americans and AI 2026 survey finds 49 percent of US adults have used an AI chatbot and 44 percent have used ChatGPT,³⁵ which by OpenAI's February 2026 count serves roughly 900 million weekly users worldwide,⁴² and the most common thing people do with AI is look up information, which used to mean a search engine. When Google shows an AI summary, users click a link about half as often, per Pew's 2025 browsing-data study,³⁷ and SparkToro's zero-click research finds 68 percent of Google searches now end without any click at all.⁴⁰ Business buyers moved the same way: Forrester's Buyers' Journey Survey put generative AI use at 94 percent of the business buyers it surveyed in 2025.⁵⁷

For a company, that reshapes the practical question. Not whether to adopt AI, which is nearly settled, but whether AI knows your business well enough to put you in the answer when a buyer asks. The rest of this report lays out the evidence behind both shifts, with every source attached.

At a glance

1. The enterprise curve

One definition before the numbers, because AI is an umbrella term. In the McKinsey series, using AI means the organization reports using AI tools, analytical or generative, in at least one business function. That is a deliberately low bar: a marketing team on an off-the-shelf copilot counts the same as a bank running custom risk models or a firm deploying autonomous agents, and the survey does not distinguish buying AI from building it. Adoption alone therefore says little about capability, which is why this report pairs it with the value and depth numbers throughout.

For most of the last decade, enterprise AI barely moved. McKinsey's annual State of AI survey put the share of organizations using AI in at least one function at 20 percent in 2017, rising to a peak of 58 percent in 2019, then settling back and hovering between 50 and 56 percent through 2022.⁵ The technology was real, but it lived in narrow, specialist corners of the business.

Generative AI broke that pattern. Under McKinsey's definition, regular generative AI use in at least one function went from 33 percent of organizations in 2023 to 71 percent in mid-2024 and 79 percent in 2025.¹³ Overall AI use followed, reaching 88 percent in 2025. Stanford's AI Index Report 2026 also carries the 88 percent figure,¹² which keeps it prominent in the broader evidence base; its public summary does not establish an independent measurement of the same population, so we do not count it as a second source. Other syntheses count generative AI use differently and land lower, so this report quotes McKinsey's definition throughout. The plainest way to say it: for six years AI was something most companies were trying, and in the last two it became something most surveyed organizations report using regularly in at least one function.

The workforce caught up on its own. Gallup finds the share of US employees who have used AI in their role roughly doubled in two years, from 21 percent in 2023 to 40 percent in mid-2025, reaching 50 percent in early 2026, with 28 percent using it at least weekly.³³³⁴ Controlled studies explain why it sticks. In the 2023 GitHub Copilot trial by Peng and colleagues, developers with an AI coding assistant finished a benchmark task 55.8 percent faster.⁷² In the NBER Generative AI at Work study by Brynjolfsson, Li and Raymond, support agents resolved 14 percent more issues per hour on average, and the least experienced gained 34 percent.⁷³ Cost helped too: Stanford's AI Index, drawing on Epoch AI data, counts a roughly 280-fold fall in the price of a query capable of GPT-3.5-level work between late 2022 and late 2024,¹¹⁷⁴ which lowered one more barrier to widespread employee access.

The money confirms the direction. Stanford's AI Index Report 2026 estimates global corporate AI investment, a broad measure spanning private investment, mergers and acquisitions, minority stakes and public offerings, at about 252 billion dollars in 2024 and 582 billion in 2025; under its narrower private-investment measure, the United States accounted for roughly 286 billion in 2025.¹² Gartner, drawing a wider boundary that includes AI servers, devices and services, put total worldwide AI spending near 1.5 trillion dollars in 2025.¹⁶ The far-out economic projections are larger and softer: McKinsey estimated in 2023 that generative AI could add 2.6 to 4.4 trillion dollars a year in economic value to the global economy, measured across the 63 business use cases it analyzed,⁷ and PwC's often-quoted figure of up to 15.7 trillion dollars added to global GDP by 2030 dates from 2017.²⁴ Treat those as direction, not precision.

2. Using AI is common. Getting paid for it is rare.

The adoption number and the value number have pulled apart. In McKinsey's 2025 global survey, published in November 2025 as The State of AI in 2025: Agents, Innovation, and Transformation, 88 percent of organizations use AI, 39 percent attribute any earnings impact to it, 7 percent have scaled it across the whole organization, and only about 6 percent, the group McKinsey calls AI high performers, attribute at least 5 percent of EBIT to it.¹

BCG finds the same shape twice. Its October 2024 study, Where's the Value in AI?, reported that 74 percent of companies had yet to show tangible value from AI, with only 4 percent running cutting-edge capabilities that consistently generate significant value.²⁶ A year later, in the Widening AI Value Gap survey of 1,250 executives, it found 5 percent capturing value at scale and about 60 percent seeing little to no benefit.²⁷ Accenture's Making Reinvention Real with Gen AI research puts the share reporting significant enterprise-level value from generative AI at 13 percent.²⁹ The definitions differ, so these are not one number, but they point the same way. The bottleneck is no longer access to the technology. It is rewiring how work actually gets done. Deloitte's State of AI in the Enterprise 2026 survey makes that point from the process side: 37 percent of organizations use AI only at a surface level, while about 30 percent are redesigning key processes around it.¹⁹

Agents are the next test. The same Deloitte survey finds 74 percent of organizations expect to be using agentic AI at least moderately within two years, up from 23 percent today.¹⁹ Gartner's June 2025 agentic AI release expects agentic AI inside a third of enterprise software by 2028, up from under 1 percent in 2024, with at least 15 percent of day-to-day work decisions made autonomously by then, and, in the same release, more than 40 percent of today's agentic AI projects canceled by the end of 2027.¹⁵ Enthusiasm is running well ahead of durable results, which is the normal shape of a technology that is genuinely useful and hard to operationalize.

3. Where each industry sits

The measure tracked here is the share of organizations in each industry using AI in at least one business function, the same yardstick McKinsey's State of AI survey applies economy-wide. No single survey measures every sector on that yardstick, so the middle column reports the strongest published figures directly, with the population and source named in each cell; these are reported values, not estimates. The all-firms column is the same measure across every business, including the smallest, from the Census Bureau's Business Trends and Outlook Survey or the Federal Reserve's Monitoring AI Adoption note; the gap between them is the small-firm lag. The standing column is Detectabli's qualitative reading of the evidence.

The denominator drives the spread. Technology reads above 90 percent among the large enterprises McKinsey surveys but 39.7 percent when the Census Bureau's Business Trends and Outlook Survey counts every information-sector firm,³¹ and manufacturing runs above 70 percent among Manufacturing Leadership Council members against 11 percent across all US manufacturers,⁸⁵ with worker-level generative AI use growing about 58 percent year over year, the fastest of any sector in the Federal Reserve's Monitoring AI Adoption note; the economy-wide rate in the same note is about 31 percent.³² Individuals often run ahead of institutions: 86 percent of higher-ed students use AI in their studies,⁹³ while only 39 percent of institutions have an acceptable-use policy per EDUCAUSE's 2025 AI Landscape Study.⁹¹

4. People adopted AI faster than their employers

Consumers did not wait for a rollout plan. In Pew Research Center's tracking, the share of US adults who have used ChatGPT went from 18 percent in 2023 to 34 percent in early 2025 and 44 percent in 2026, and the Americans and AI 2026 survey puts any-chatbot use at 49 percent.³⁵ By OpenAI's February 2026 count, ChatGPT hit roughly 900 million weekly users worldwide in about three years, an exceptionally fast adoption curve for a consumer technology.⁴² Globally, KPMG and the University of Melbourne's 47-country Trust, Attitudes and Use of AI study found 66 percent of people use AI intentionally and 38 percent use it weekly or daily.⁵²

Use skews young and is evening out on gender. By early 2025, 58 percent of US adults under 30 had used ChatGPT against 10 percent of those 65 and older, per Pew;³⁶ OpenAI's How People Use ChatGPT paper finds roughly 46 percent of ChatGPT messages come from users under 26;⁴¹ and Pew's Teens, Social Media and AI Chatbots survey has 64 percent of US teens already using them.³⁸ The gender gap that existed at launch has largely closed.³⁵⁴¹

What people do with it matters more than how many do. The single most common use is looking up information: among US chatbot users in Pew's Americans and AI 2026 survey, 42 percent use AI to search for information, ahead of work tasks at 38 percent of employed users, entertainment at 25 percent, and image or video creation at 24 percent, with medical advice and diet or fitness at 20 percent each.³⁵ OpenAI's own analysis of ChatGPT usage found non-work messages rose from 53 percent to 73 percent of the total in a single year, with practical guidance and information-seeking dominating the mix.⁴¹ When the top job people bring to AI is the job they used to bring to a search engine, the competition for attention moves from the results page into the model's answer.

Business buyers moved hardest of all

The consumer numbers understate what happened on the business side. Forrester's Buyers' Journey Survey found 89 percent of B2B buyers had brought generative AI into their buying process by 2024⁵⁶ and 94 percent by 2025, and in the 2025 Buyers' Journey Survey more buyers named generative AI and conversational search their most meaningful information source than any other option, ahead of vendor websites and sales teams.⁵⁷ 6sense's 2025 Buyer Experience Report, a study of roughly 4,000 buyers, reads the same way: 94 percent used large language models during a purchase, most heavily in the middle of the journey to compare vendors.⁵⁸

The shortlist math is what makes this matter. Bain's Losing Control research finds 85 percent of B2B buyers purchase from the list of vendors they had in mind on day one,⁶² and 6sense finds the eventual winner was on that first list 95 percent of the time.⁵⁸ Shaping the day-one list is a major advantage, won before a salesperson is ever contacted, and AI is now doing the shaping. G2's The Answer Economy study surveyed over a thousand software buyers in early 2026.⁵⁹ It found 51 percent now start research with AI chatbots more often than Google, up from 29 percent a year earlier. The same buyers named AI chatbots the top influence on which vendors made the shortlist, and a third had bought from a vendor they had never heard of before AI surfaced it. Bain also measures the cost of staying invisible: click-through rates in some categories, B2B software among them, have fallen by as much as 30 percent since AI summaries arrived in search.⁶²

Two grains of salt. Several of the firms publishing these numbers sell tools that benefit from the trend, and the most independent reads run cooler: Gartner's May 2026 release on its survey of 645 B2B buyers, fielded in late 2025, finds 45 percent of them using generative AI for vendor research, and 69 percent preferring to validate what it tells them with a sales rep.⁶¹ But buyer skepticism cuts the other way too. People are checking AI's claims about you, not skipping them.

5. Search is changing shape

The measured pieces are these. SparkToro and Datos's 2024 Zero-Click Search Study found 58.5 percent of US Google searches and 59.7 percent in the EU end without any click at all; counted against the open web, about 360 of every 1,000 US Google searches produce a click that leaves Google's ecosystem, compared with about 374 in the EU.³⁹ SparkToro's June 2026 update, run on Similarweb's US panel of browser-based searches, puts zero-click at 68 percent for the first four months of 2026, against 60.45 percent on the same basis in 2024, and finds AI Overviews cut click-through by nearly 60 percent when they appear in that panel.⁴⁰ The two studies use different panels, so the levels are not perfectly comparable, but the direction is steep and consistent. Zero-click and AI-led are different measures, which is why the scenarios in figure 3 sit far below the zero-click rate. Google's AI Overviews now reach more than 2 billion monthly users, per Alphabet's second-quarter 2025 earnings call,⁴³ and Pew's browsing-data study measured what they do to behavior: when a summary appears, users click a traditional result 8 percent of the time, against 15 percent without one.³⁷ Bain's February 2025 consumer survey finds about 80 percent of search users rely on AI summaries at least 40 percent of the time,⁴⁴ that about 60 percent of searches end without the user reaching a third-party site, and that 44 percent of online buyers now start product journeys in an AI tool or split between AI and classic search, per its 2026 buyer research.⁴⁵ For news the drop is starker in the Reuters Institute's Digital News Report 2026: 4 percent of people who read news through an AI chatbot click through to the source, against 19 percent from search.⁵⁰

Not all of that traffic vanished. Some is absorbed by answers and features on the results page, AI summaries among them, some shifts to other discovery channels, and the part that still reaches websites converts unusually well. Salesforce's holiday shopping data⁴⁷ puts AI-influenced online holiday sales in late 2025 at 262 billion dollars, about a fifth of the total, with shoppers arriving from AI search converting roughly nine times more often than those from social media.⁴⁷ Adobe Analytics measured generative AI referral traffic to US retail sites up 1,200 percent in early 2025 versus the prior July,⁴⁸ and by the 2025 holidays those visitors converted 31 percent above non-AI traffic.⁴⁹ Two honest caveats: the absolute base is still small, near 1 percent of web traffic, and in Adobe's early-2025 reading AI visitors actually converted about 9 percent below average. The two readings cover different periods and traffic mixes, so the swing is direction, not proof the same traffic improved.

Trust sets the ceiling. Bain finds half of shoppers trust generative AI for initial research and product comparisons, while half remain cautious about letting AI complete a purchase on its own.⁴⁵⁴⁶ Edelman's November 2025 Trust Barometer flash poll measured trust in AI at 87 percent in China and 67 percent in Brazil, against 39 percent in Germany, 36 percent in the UK and 32 percent in the US.⁵¹ People are willing to let AI help them decide long before they let it decide for them, so the near-term battleground is influence over the recommendation, not the checkout.

The answers are often wrong

The uncomfortable part of AI-led discovery is the error rate. The Tow Center at Columbia, in its AI Search Has a Citation Problem study, ran 1,600 news-attribution queries through eight AI search tools in early 2025 and got incorrect answers on more than 60 percent of them, with the tools rarely admitting uncertainty; the paid tiers were more confidently wrong than the free ones, and more than half the citations from two of the tools pointed to fabricated or broken links.⁶⁴ The BBC's February 2025 study found 51 percent of AI answers to news questions, judged against its own reporting, had significant issues,⁶⁵ and the EBU and BBC's News Integrity in AI Assistants follow-up across 22 public broadcasters in 18 countries and 14 languages put the rate at 45 percent.⁶⁶

The engines are improving, and still wrong at a scale that matters. A New York Times analysis with the AI lab Oumi measured Google's AI Overviews at roughly 91 percent accuracy on a factual benchmark in early 2026, up from 85 percent a few months before, but a single-digit error rate against billions of queries a day is still an enormous volume of wrong answers, and 56 percent of the correct ones were not fully supported by the sources they cited.⁶⁷ For businesses specifically the data is younger but points the same way: one vendor study by Searchable of 165 London businesses across more than 13,000 queries, reported by Search Engine Journal in mid-2026, found 93 percent had at least one basic fact wrong or missing in AI answers, with fabricated details hitting small firms hardest: half saw at least one made-up fact, against a third of large firms.⁶⁹ Ahrefs's September 2025 link study measured AI assistants sending clicks to dead pages nearly three times as often as Google.⁶⁸ A model can misstate your prices, your services or your location for months, and unlike an error on a page you control, it may never surface in your analytics or publishing workflow.

For a business, the takeaway is concrete. Ranking a page is no longer the whole game. Being the source an AI cites, and being described accurately when it does, is the new first page. If the model lacks a clear, well-corroborated picture of what you do, it fills the gap with whatever it found, which may be a competitor.

6. The next five years

Enterprise: depth, not breadth

Adoption of AI somewhere in the business is near its ceiling. A saturating curve fitted to the 2023 to 2025 McKinsey series produces an illustrative ceiling of roughly 96 percent, though the result is sensitive to which 2024 survey wave feeds the fit; across the ceilings we test, 2030 lands between about 92 and 97 percent. Either way, there is not much room above 88. The action moves to depth. Gartner's June 2026 data and analytics outlook predicts more than one in ten enterprises will operate as AI-first businesses by 2030,¹³ and its July 2025 forecast has 80 percent of enterprise software multimodal by then, up from under 10 percent in 2024.¹⁴ AI-first, in Gartner's usage, means the operating model itself is built around AI, with agents and a converged data and analytics platform at the core of products and decisions, rather than AI bolted onto existing workflows. The agent numbers from section 2 sit alongside those, cancellation caveat included. Expect a messy few years in which the winners compound and the average company churns through pilots. The phrase to watch is not adoption but value at scale, and it climbs from a single-digit base.

Consumers: from app to default

Counterpoint's Global AI Consumer Spending Forecast, released in December 2025, projects consumer spending on generative AI rising from 225 billion dollars in 2023 to 699 billion by 2030, with monthly active users of AI conversational platforms passing 5 billion and generative AI smartphone shipments growing at a 26 percent compound rate. Counterpoint quotes the spending growth as 21 percent compounded annually over its 2024 to 2030 forecast window; measured from the 2023 base it works out to 17.6 percent a year.³⁰ The direction is that generative AI stops being an app people choose and becomes a feature of the devices and services they already own. Younger cohorts who default to AI today age into their peak spending years within this window. The real uncertainty is trust and regulation, which could slow autonomous AI commerce even as AI-assisted discovery keeps growing.

Search: three ways it could go

Conservative: AI answers stay a strong complement, zero-click rates rise modestly, and classic results keep most commercial intent. Central: AI becomes the default interface for research, how-to and complex purchase queries by 2030 while navigational and transactional queries stay with classic search. Aggressive: platforms push AI-first experiences across most query types, which would also invite heavy regulatory scrutiny. Worth remembering on the skeptical side: Gartner's February 2024 call that classic search volume would fall 25 percent by 2026 is widely disputed. We show all three as curves in figure 3 rather than pick a favorite. In the aggressive scenario, AI-led queries pass 50 percent during 2028; the central scenario reaches the 50 percent line right around 2030; the conservative scenario, whose ceiling sits at 40 percent, cannot cross it by construction. Our own read is that the evidence in section 5 sits between the central and aggressive paths, and we hold that as a judgment, not a data-supported point estimate.

7. What it means for being found

Put the two shifts together. Large organizations have largely adopted AI somewhere in the business, and their customers now begin or shape a growing share of buying journeys through AI tools, often alongside traditional search. The scarce thing is no longer access to the technology. It is being understood by it.

The evidence that visibility can be engineered

Two research results make AI-answer visibility a discipline rather than a hope. The GEO benchmark by Aggarwal and colleagues, presented at KDD 2024, tested content rewrites across 10,000 queries and found that adding citations, quotations and statistics can lift a source's visibility in generative answers by up to about 40 percent, with the largest gains for mid- and lower-ranked sources,⁷⁰ while old-style keyword stuffing did nothing or hurt. The follow-on e-commerce testbed E-GEO, from Bagga and colleagues in late 2025, showed optimized content and prompts moving products up 0.7 to 1.6 positions on average in an AI shopping assistant's rankings.⁷¹ Neither replaces classic SEO. They stack on top of it, and the brands that work both layers gain compounded visibility while the practice is still young.

That is the problem Detectabli works on. We measure how most AI engines, including ChatGPT, Gemini, Claude and Perplexity, describe and recommend a business, find where they get it wrong or leave it out, and ship the specific fixes that make them treat it as a reliable source. Two honest caveats belong here. Technical hygiene like schema markup helps machines read a page but is not by itself a proven lever on how often a model cites you. And no one controls a model's output outright. The realistic goal is to shape the signals the model synthesizes, consistently and across enough independent sources that the answer comes out accurate and in your favor.

The numbers in this report are the case for doing that now rather than later. Adoption is done arguing with itself. The next competition is over who the answer names.

Methodology and how to read the numbers

Reported versus modeled. Every figure attributed to a named source is reported survey or market data. The year-by-year curves between reported points, all values beyond the latest reported observation in each series, and the scenarios in figure 3 are Detectabli estimates. Several series include reported 2026 observations; only values after the last observation are modeled. The industry table in section 3 contains only reported survey values. Projections appear only as dashed, shaded, labeled curves; they are never blended with reported survey values or given the same standing.

The projection models. Every dashed curve in the charts is generated by a fitted equation rather than drawn by hand. Enterprise AI follows a saturating curve fit to the post-2023 regime, A(t) = 95.7 − 7.7·e^−0.83(t − 2025), so the ceiling of just under 96 percent comes out of the fit rather than being set by hand. Three observations give the curve enough freedom to fit them almost perfectly, though, so that ceiling should be read as illustrative rather than statistically determined; the band in figure 1 spans declared ceilings of 92 to 97 percent, with 2030 values of roughly 92 to 97. For each declared ceiling, the rate is re-estimated by unweighted least squares against the 2023 to 2025 points, with the curve anchored at the reported 88 percent in 2025, so only the asymptote is assumed. The enterprise fit treats McKinsey's waves as annual index points at 2023, 2024 and 2025, the spacing the survey's yearly rhythm approximates; the stress-test section shows what changes when the wave choice changes. No single curve is fit across the full 2017 to 2025 history, because the survey definition changed twice and generative AI broke the trend in 2023.

Generative AI use is modeled as a constant share of overall AI use, A_gen(t) = 0.904·A(t): that share jumped once, from 60 percent of AI adopters in 2023 to 91 percent in 2024, and has held near 90 percent since. Two years of a stable ratio is an assumption carried forward, not an established constant: generative AI could keep converging toward all AI users, or some organizations could keep running conventional predictive AI without deploying it broadly. The band in figure 1 spans those paths, and by 2030 it runs from the low 80s to the mid 90s of surveyed large organizations.

Consumer adoption follows a logistic, F(t) = L / (1 + e^−k(t − m)). The ChatGPT curve is fit by unweighted least squares to Pew's four fieldwork-dated observations, July 2023, February 2024, February to March 2025 and February 2026, giving k = 0.55 per year with midpoint late 2025. The February to March 2025 fieldwork date matters: that survey published in June 2025, and dating the observation by publication instead of fieldwork shifts the fit visibly, which is why fieldwork dates are used throughout. The any-chatbot curve is calibrated exactly through Pew's two dated observations and gives k = 0.44 with midpoint late 2025; Pew reports the earlier observation by year only, as 2024, so we place it at mid-2024, and the calibration inherits that assumption. With two points and a fixed ceiling that is calibration, not statistical fitting. The ceilings are declared rather than estimated, 80 and 90 percent of US adults, because every observed point sits before the curve's midpoint, where data cannot identify a ceiling; no fit residual exceeds half a point. The chart bands span the ceiling assumptions, 70 to 90 percent for ChatGPT and 80 to 95 for any chatbot, the functional-form spread across logistic, Gompertz and saturating-exponential fits, and plus-or-minus 1.5-point sensitivity perturbations of the observations, which puts 2030 at roughly 59 to 80 and 68 to 85 percent respectively. The any-chatbot band is the widest because a curve calibrated through only two points amplifies any change in them. Two caveats travel with these curves: they model ever-used, which counts someone who tried a chatbot once the same as a daily user, and the any-chatbot series rests on just two reported observations, so it is an illustrative scenario built on limited data and an assumed ceiling, not a fit with statistical confidence.

The search scenarios use the same logistic form with declared ceilings of 40, 55 and 65 percent. The 65 percent upper bound reflects navigational queries staying classic-led, roughly a third of volume in SparkToro and Datos's 332-million-query dataset, though it does not follow that everything else becomes AI-led, which is what the lower ceilings represent.¹⁰³ For each declared ceiling, the rate and midpoint are re-estimated by unweighted nonlinear least squares against the three composite annual anchors; the aggressive case lands at k = 0.52 with midpoint early 2026. The aggressive curve crosses 50 percent during 2028, the central curve reaches 50 percent right around 2030, and the conservative curve stays below it through the window. These are strategic scenarios, not predictive forecasts: no directly measured historical series exists for the share of queries resolved AI-led, and the anchors are composites built from related but non-interchangeable measures, zero-click rates, AI Overview coverage and AI referral traffic among them. One model was tried and rejected: a Bass diffusion specification for consumer adoption, because the short, unusually rapid launch history produced unstable and implausible parameter estimates, and we say so rather than force it.

Stress tests on the models. Three checks worth disclosing. The enterprise ceiling depends on which 2024 survey wave feeds the fit. The model uses the mid-2024 wave (78 percent) for two stated reasons, neither of them preference: it is the value McKinsey itself uses as the 2024 point in its published trend series, where the 2025 report frames the result as 88 percent against 78 a year earlier, and its July 2024 fieldwork sits closest to the annual spacing the fit assumes against the 2023 and 2025 fieldwork dates. With the early-2024 wave (72) the three points run nearly linear and identify no ceiling in range, which we disclose rather than average away; the Census evidence that remaining non-adopters are mostly small, resource-constrained firms supports the saturation reading but is not an input to the fit. A leave-one-out test on the ChatGPT model, trained without the newest Pew observation, over-predicts that observation by about a point, while the linear baseline lands within half a point below it; two folds cannot rank the models, so we report both and draw no bias conclusion. And the declared ceilings move the endpoints roughly linearly: ten points of consumer ceiling shifts the 2030 value by about six, and the three search scenarios in figure 3 span ceilings of 40 to 65 percent, which puts the 2030 AI-led share anywhere between 39 and 57 percent and the crossover anywhere from 2028 to not at all. We use scenario bands rather than formal confidence intervals throughout because the dominant uncertainty is structural, definitions, ceilings and survey design rather than sampling error, and intervals computed from three or four observations would manufacture false precision. One outside check: Semrush, projecting from its own traffic data for the digital-marketing and SEO topics it tracks, also lands AI search visitors passing traditional search by early 2028.¹⁰²

The denominator. AI adoption runs from about 20 percent to about 90 percent depending on who is asked. McKinsey and most industry surveys sample large enterprises and read high. The US Census and Federal Reserve sample every firm, including the very small, and read low. Both are accurate for what they measure, and each figure in this report names its population.

Definitions drift. Survey wording changes over time, from adopted AI to uses AI to regularly uses AI. McKinsey fielded two survey waves in 2024, which is why 72 and 78 percent both circulate for that year. The US Census widened its question in late 2025, which by itself lifted its headline from single digits to about 20 percent. Where a jump reflects a change in definition rather than real growth, we say so. Where sources disagree, such as market-size boundaries or the fate of Gartner's 2024 search-volume prediction, we show the range rather than pick a favorite.

Vintage. Figures are point-in-time and reflect the best available data as of July 2026. Some widely quoted projections are old: PwC's 15.7 trillion dollar GDP figure dates from 2017 and IDC's 632 billion dollar spending forecast from August 2024. We keep the dates attached so the reader can weigh them. All third-party data belongs to its original publisher and is cited below.

Appendix: model inputs and validation

This appendix publishes what the models actually consume and how they hold up against simpler alternatives: every dated observation, rolling backtests against two baselines, the spread across functional forms, and the forecasts we are locking in now so the next edition can score them.

Table A1. Every observation the models use

Rolling backtests against simple baselines

A curve earns its keep only if it predicts withheld data better than something simpler. For each series with enough history, we refit on the early observations and predicted the next one: last value carried forward, a straight line, and the Detectabli curve with its declared ceiling.

In the two available rolling folds, both trend models substantially outperformed the no-growth baseline: mean absolute error is 1.0 point for the linear baseline and 0.8 for the central logistic, against 10.5 for last value carried forward. The logistic had the slightly lower error, but two folds cannot rank models reliably, and linear growth is not viable long run because adoption is bounded at 100 percent; the enterprise series makes that concrete, with the linear baseline extrapolating to an infeasible 101. The enterprise curve itself cannot be tested out of sample on three points, so it is an illustrative saturation scenario, not an empirically validated forecast. The any-chatbot series has two observations, so no fold exists at all. And the AI-led row is a holdout check against our own composite anchor, a mechanical consistency test, not validation against measured AI-led query share. Across the two available ChatGPT folds, the central logistic missed by less than a point on average; that is a useful preliminary reference, not a reliable estimate of expected one-year forecast error.

One curve is a choice: the functional-form spread

Three enterprise points are matched exactly by a monomolecular, a logistic and a Gompertz curve alike, so in-sample fit cannot choose. They land at 95.6, 92.4 and 93.8 percent in 2030, much of the reason the published band spans 92 to 97. For ChatGPT at the declared 80 percent ceiling, the logistic, Gompertz and saturating-exponential forms fit the four points within 0.5, 0.3 and 1.0 points respectively but diverge to 73.1, 68.7 and 63.3 percent by 2030: the choice of curve moves the endpoint by about ten points, and the 70-to-90 ceiling span moves it by about twelve, so neither assumption dominates and both feed the band. In-sample fit is too close to choose among the forms; we keep the logistic as the central case as the standard adoption form, and the spread is part of the structural uncertainty.

Validation at a glance

Table A2. Forecasts locked for scoring

These values are locked as of this edition; the next edition fills in the actuals and scores them, misses included. A reported value outside its scenario range triggers a re-examination of that model's form and ceiling, reported here rather than adjusted away. The ranges are scenario ranges, not statistical prediction intervals: each is the minimum to maximum across model forms, declared ceilings and plus-or-minus 1.5-point observation-value sensitivity perturbations, rounded outward. The perturbations are sensitivity assumptions applied uniformly, not the original publishers' confidence intervals. The narrower within-one-family bands remain in the published results files for reference. The ChatGPT targets are restated from the first edition after correcting the third observation's date from its June 2025 publication to its February to March 2025 fieldwork; the first edition's central value was 53.1.

Reproducibility. Estimation is unweighted nonlinear least squares (SciPy curve_fit), with exact closed forms where a system is exactly identified; rates and midpoints are estimated, ceilings fixed where declared. Consumer-series fieldwork dates convert to decimal years programmatically, never by hand; the enterprise fit uses annual index points as stated in the methodology. A separately implemented calculation path recomputes every published value by grid search and exact fractions, guarding against optimizer and starting-value artifacts. Python 3.11, SciPy 1.17; data cutoff July 20, 2026.

Second edition changelog

1. Section 5: the SparkToro open-web click figures were checked twice. The first edition's US figure of about 360 per 1,000 was correct; an interim draft of this audit reversed it after trusting the study's URL slug, which binds 374 to the US, but the published headline and body both read US 360, EU 374. The sentence now states both figures, sourced to the article body.
2. Industry table, Real Estate & Construction: all-firms figure corrected from 24% to under 10% (BTOS puts construction among the lowest sectors; no Census figure supports 24%). RICS citation replaced: the Q1 2025 Global Construction Monitor contains no AI question; the row now cites RICS's AI in Construction 2025 (45% report no AI use, 1% scaled). Standing reworded to match.
3. Industry table, Energy & Utilities: the 17–26% genAI figure could not be located in the cited Deloitte 2026 outlook and is removed; the row now carries the outlook's actual statistic, a projection of nearly 40% of control rooms using AI by 2027, labeled as a projection.
4. Industry table, smaller fixes: telecom row now says generative AI (NVIDIA's measure); professional services all-firms widened to 33–40% (BTOS readings vary by survey week; the Minneapolis Fed's 2026 read is near 40%); education row clarified to teachers who used AI during the school year; agriculture row corrected to farmers using AI weekly or more (MorganMyers's actual measure); government row reworded to the NASCIO finding (employees in 82% of state offices use AI daily).
5. Section 2: the two June 2025 Gartner agentic citations are one press release (June 25, 2025) and are now cited as one. Deloitte agentic figure reworded to the source's formulation: 74% expect to be using agentic AI at least moderately within two years, not 'deploy across multiple operational areas'.
6. Section 3 closing paragraph: the Federal Reserve 58% growth figure is now explicitly scoped to manufacturing (the fastest sector in the FEDS note) with the economy-wide rate, about 31%, added alongside. The first edition's sentence was accurate but easy to misread as economy-wide.
7. Section 1: the claim that Stanford's AI Index independently corroborates the McKinsey adoption trend is removed. The AI Index 2026 reports the same 88 percent figure, but its public summary does not establish an independent measurement of the same population, so the sentence no longer counts it as a second source.
8. Section 4: Gartner B2B buyer survey now notes its late-2025 fieldwork behind the May 2026 release, per the report's own vintage rule.
9. Section 6: the conservative search scenario's inability to cross 50% is now labeled as true by construction (its ceiling is 40%), a definition rather than a finding.
10. Model correction, the substantive one: Pew's 34% ChatGPT observation was dated to its June 2025 publication; fieldwork was February 24 to March 2, 2025. Refit on fieldwork dates: k moves from 0.54 to 0.55, midpoint stays late 2025 (2025.75), and no residual exceeds half a point (previously 1.6).
11. Downstream of that refit: locked ChatGPT central for Pew early 2027 moves from 53.1 to 54.3 (range 47–57 unchanged after recomputation); the 2030 form spread becomes 73.1 / 68.7 / 63.3 (was 72.7 / 68.0 / 62.3); the ChatGPT 2030 band widens slightly to roughly 59–80.
12. Backtest table recomputed with corrected dates: fold 1 becomes linear 32.4 / curve 34.2 against actual 34; fold 2 becomes linear 43.7 / curve 45.3 against actual 44. MAEs: logistic 0.8, linear 1.0, last value 10.5. This reverses the first edition's finding that the linear baseline had the lower error; the text now says the logistic edges it and that two folds cannot rank models either way.
13. Stress tests: the leave-one-out result flips sign. With corrected dates the model over-predicts the newest Pew observation by about a point (first edition: under-predicts by about three). The 'central curve was conservative' sentence is removed.
14. Functional-form paragraph: with corrected dates the ceiling span (about twelve points at 2030) slightly exceeds the form spread (about ten); the first edition said form mattered more. Gompertz now fits marginally best in-sample; the logistic is kept central as the standard adoption form and the text says the in-sample fit cannot choose.
15. Methodology: the enterprise fit's annual-index-point spacing is now stated plainly in both the methodology and the reproducibility note, which previously implied all series used decimal fieldwork dates. The any-chatbot mid-2024 placement is now labeled as an assumption in the methodology and Table A1 (Pew reports that observation by year only).
16. Table A1: ChatGPT 34% fieldwork corrected to Feb–Mar 2025; the any-chatbot 2024 row is labeled 'year only; mid-2024 assumed'.
17. Table A2: ChatGPT central restated to 54.3 with a note explaining the restatement; enterprise (92.3, 89–95), genAI (83.5, 80–88) and any-chatbot (58.5, 54–63) targets unchanged, since their inputs did not change.
18. Sources: [67] outlet corrected to Search Engine Land; [95] replaced with RICS AI in Construction 2025; [102] date corrected to June 2025; [15], [22], [61] and [98] annotated as above.
19. Charts: figure 2 is regenerated for this edition (the 34% point moves to its early-2025 fieldwork position; dashed curves and bands come from the refit; caption updated). Figures 1 and 3 are unchanged. Verified in recomputation: figure 3's stated crossings hold (aggressive crosses 50% at 2028.6, central at 2029.9).
20. Verified and left alone, among others: the full McKinsey series, Stanford HAI and Gartner spending figures, BCG, Accenture, all Pew figures and the usage breakdown, the 900M weekly users count, Counterpoint (including the 17.6% recomputation), Forrester, 6sense, Bain, G2, Tow Center, BBC and EBU error rates, Reuters Institute, Salesforce, Adobe, Ahrefs, GEO and E-GEO, Gallup, Census BTOS headline and the information, finance and retail rows, and the Edelman country numbers.

Sources

Superscript numbers in the text link to the entries below. Third-party forecasts belong to the named publishers. Detectabli's own modeled scenarios are labeled as such in the charts and methodology.

1. [1]McKinsey & Company. The State of AI in 2025: Agents, innovation, and transformation. Nov 2025. mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
2. [2]McKinsey & Company. The state of AI: How organizations are rewiring to capture value. Mar 2025. mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
3. [3]McKinsey & Company. The state of AI in early 2024. May 2024. mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-2024
4. [4]McKinsey & Company. The state of AI in 2023: Generative AI's breakout year. Aug 2023. mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2023-generative-ais-breakout-year
5. [5]McKinsey & Company. The state of AI in 2022 and a half decade in review. Dec 2022. mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2022-and-a-half-decade-in-review
6. [6]McKinsey & Company. AI at work but not at scale (Week in Charts). 2025. mckinsey.com/featured-insights/week-in-charts/ai-at-work-but-not-at-scale
7. [7]McKinsey Global Institute. The economic potential of generative AI: The next productivity frontier. Jun 2023. mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
8. [8]McKinsey & Company. Superagency in the workplace. Jan 2025. mckinsey.com/featured-insights/week-in-charts/leaders-underestimate-employees-ai-use
9. [9]McKinsey & Company. Generative AI in healthcare: Current trends and future outlook. 2025. mckinsey.com/industries/healthcare/our-insights/generative-ai-in-healthcare-current-trends-and-future-outlook
10. [10]McKinsey & Company. Telcos' AI inflection point. Dec 2025. mckinsey.com/industries/technology-media-and-telecommunications/our-insights/telcos-ai-inflection-point-what-leaders-do-to-capture-value
11. [11]Stanford HAI. AI Index Report 2025, Chapter 4: Economy. Apr 2025. hai.stanford.edu/assets/files/hai_ai-index-report-2025_chapter4_final.pdf
12. [12]Stanford HAI. AI Index Report 2026, Chapter 4: Economy. Apr 2026. hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf
13. [13]Gartner. Gartner Identifies the Top Trends for Data and Analytics (AI-first by 2030). Jun 2026. gartner.com/en/newsroom/press-releases/2026-06-16-gartner-identifies-the-top-trends-for-data-and-analytics
14. [14]Gartner. 80% of Enterprise Software and Applications Will Be Multimodal by 2030. Jul 2025. gartner.com/en/newsroom/press-releases/2025-07-02-gartner-predicts-80-percent-of-enterprise-software-and-applications-will-be-multimodal-by-2030-up-from-less-than-10-in-2024
15. [15]Gartner. Over 40% of Agentic AI Projects Will Be Canceled by End of 2027. Jun 25, 2025. One press release; it also carries the third-of-enterprise-software and 15-percent-of-decisions forecasts. gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027
16. [16]Gartner. Worldwide AI Spending Will Total $1.5 Trillion in 2025. Sep 2025. gartner.com/en/newsroom/press-releases/2025-09-17-gartner-says-worldwide-ai-spending-will-total-1-point-5-trillion-in-2025
17. [17]Gartner. Search Engine Volume Will Drop 25% by 2026. Feb 2024. gartner.com/en/newsroom/press-releases/2024-02-19-gartner-predicts-search-engine-volume-will-drop-25-percent-by-2026-due-to-ai-chatbots-and-other-virtual-agents
18. [18]IDC. Worldwide AI spending forecast to reach $632 billion in 2028. Aug 2024. businesswire.com/news/home/20240819177906/en/
19. [19]Deloitte. State of AI in the Enterprise 2026: The Untapped Edge. Jan 2026. deloitte.com/us/en/about/press-room/state-of-ai-report-2026.html
20. [20]Deloitte. State of Generative AI in the Enterprise, Wave 4. Jan 2025. deloitte.com/us/en/about/press-room/state-of-generative-ai.html
21. [21]Deloitte. 2026 Manufacturing Industry Outlook. 2026. deloitte.com/us/en/insights/industry/manufacturing-industrial-products/manufacturing-industry-outlook.html
22. [22]Deloitte. 2026 Power and Utilities Industry Outlook. 2026. Cited for its control-room projection only; the outlook publishes no current genAI adoption rate for utilities. deloitte.com/us/en/insights/industry/power-and-utilities/power-and-utilities-industry-outlook.html
23. [23]Deloitte Center for Government Insights. Scaling AI in government. 2025. deloitte.com/us/en/insights/industry/government-public-sector-services/government-trends/2025/scaling-ai-in-government.html
24. [24]PwC. Sizing the Prize: PwC's Global AI Study. 2017. pwc.com/gx/en/issues/data-and-analytics/publications/artificial-intelligence-study.html
25. [25]PwC. 2025 Global AI Jobs Barometer. Jun 2025. pwc.com/gx/en/news-room/press-releases/2025/ai-linked-to-a-fourfold-increase-in-productivity-growth.html
26. [26]Boston Consulting Group. Where's the Value in AI? Oct 2024. bcg.com/publications/2024/wheres-value-in-ai
27. [27]Boston Consulting Group. The Widening AI Value Gap (Build for the Future 2025). Sep-Oct 2025. media-publications.bcg.com/The-Widening-AI-Value-Gap-Sept-2025.pdf
28. [28]Boston Consulting Group. AI at Work 2025. Jun 2025. bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain
29. [29]Accenture. Making Reinvention Real with Gen AI. Mar 2025. accenture.com/us-en/insights/consulting/making-reinvention-real-with-gen-ai
30. [30]Counterpoint Research. Global Generative AI Consumer Spending to Approach $700 Billion by 2030. Dec 2025. counterpointresearch.com/en/insights/global-generative-ai-consumer-spending-to-skyrocket
31. [31]US Census Bureau. Business Trends and Outlook Survey: AI use by businesses. May 2026. census.gov/library/stories/2026/05/ai-use-businesses.html
32. [32]US Federal Reserve. Monitoring AI Adoption in the U.S. Economy (FEDS Note). Apr 2026. federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.html
33. [33]Gallup. Rising AI Adoption Spurs Workforce Changes. 2026. gallup.com/workplace/704225/rising-adoption-spurs-workforce-changes.aspx
34. [34]Gallup. AI Use at Work Has Nearly Doubled in Two Years. Jun 2025. gallup.com/workplace/691643/work-nearly-doubled-two-years.aspx
35. [35]Pew Research Center. Americans and AI 2026: Chatbots, smart devices, and views on impact. Jun 2026. pewresearch.org/internet/2026/06/17/americans-and-ai-2026-chatbots-smart-devices-and-views-on-impact/
36. [36]Pew Research Center. 34% of US adults have used ChatGPT. Jun 2025. pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023/
37. [37]Pew Research Center. Google users are less likely to click on links when an AI summary appears. Jul 2025. pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
38. [38]Pew Research Center. Teens, Social Media and AI Chatbots 2025. Dec 2025. pewresearch.org/internet/2025/12/09/teens-social-media-and-ai-chatbots-2025/
39. [39]SparkToro (Rand Fishkin). 2024 Zero-Click Search Study. Jul 2024. Caution: the URL slug reverses the US/EU open-web pairing; the published headline and body read US 360, EU 374 per 1,000 searches. sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/
40. [40]SparkToro (Rand Fishkin). In 2026, Less Than One-Third of Google Searches Still Send a Click. Jun 2026. sparktoro.com/blog/in-2026-less-than-one-third-of-google-searches-still-send-a-click/
41. [41]OpenAI / NBER (Chatterji et al.). How People Use ChatGPT. Sep 2025. cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f142e/economic-research-chatgpt-usage-paper.pdf
42. [42]TechCrunch. ChatGPT reaches 900M weekly active users. Feb 2026. techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users/
43. [43]TechCrunch / Alphabet. Google AI Overviews have 2B monthly users, AI Mode 100M. Jul 2025. techcrunch.com/2025/07/23/googles-ai-overviews-have-2b-monthly-users-ai-mode-100m-in-the-us-and-india/
44. [44]Bain & Company. Consumer reliance on AI search results signals new era of marketing. Feb 2025. bain.com/about/media-center/press-releases/20252/consumer-reliance-on-ai-search-results-signals-new-era-of-marketing--bain--company/
45. [45]Bain & Company. Your Next Customer Will Find You Using AI. Now What? Apr 2026. bain.com/insights/your-next-customer-will-find-you-using-ai-now-what/
46. [46]Bain & Company. Agentic AI Poised to Disrupt Retail. Nov 2025. bain.com/about/media-center/press-releases/20252/agentic-ai-poised-to-disrupt-retail-even-with-50-of-consumers-cautious-of-fully-autonomous-purchasesbain--company/
47. [47]Salesforce. How AI Influenced $262 Billion in Online Holiday Sales. Jan 2026. salesforce.com/news/stories/2025-holiday-shopping-data/
48. [48]Adobe Analytics. Traffic to US retail websites from generative AI sources jumps 1,200%. Mar 2025. blog.adobe.com/en/publish/2025/03/17/adobe-analytics-traffic-to-us-retail-websites-from-generative-ai-sources-jumps-1200-percent
49. [49]Adobe Digital Insights. AI Traffic Report: AI-sourced traffic insights 2025. Jan 2026. business.adobe.com/assets/pdfs/resources/reports/ai-traffic-report/ai-sourced-traffic-insights-2025.pdf
50. [50]Reuters Institute. Digital News Report 2026: AI chatbots and news. Jun 2026. reutersinstitute.politics.ox.ac.uk/digital-news-report/2026/emerging-uses-ai-chatbots-news-and-what-it-means-journalism
51. [51]Edelman. Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads. Nov 2025. edelman.com/trust/2025/trust-barometer/flash-poll-trust-artifical-intelligence
52. [52]KPMG / University of Melbourne. Trust, attitudes and use of AI: A global study (47 countries). Apr 2025. kpmg.com/xx/en/media/press-releases/2025/04/trust-of-ai-remains-a-critical-challenge.html
53. [53]Ipsos. Ipsos AI Monitor 2025 (30 countries). Jun 2025. ipsos.com/en/ipsos-ai-monitor-2025
54. [54]Menlo Ventures. 2025: The State of Consumer AI. Jun 2025. menlovc.com/perspective/2025-the-state-of-consumer-ai/
55. [55]eMarketer. US GenAI Consumer Adoption Forecast H1 2026. 2026. emarketer.com/content/us-genai-consumer-adoption-forecast-h1-2026
56. [56]Forrester. B2B Buyer Adoption of Generative AI (Buyers' Journey Survey 2024). Nov 2024. forrester.com/report/b2b-buyer-adoption-of-generative-ai/RES181769
57. [57]Forrester. B2B Buyers Make Zero-Click Buying Number One (Buyers' Journey Survey 2025). 2025. forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/
58. [58]6sense. 2025 B2B Buyer Experience Report: The Science of B2B. Nov 2025. 6sense.com/science-of-b2b/buyer-experience-report-2025/
59. [59]G2. The Answer Economy: How AI Search Is Rewiring B2B Software Buying. Apr 2026. company.g2.com/news/g2-research-the-answer-economy
60. [60]TrustRadius. 2026 B2B Buying Disconnect Report. Jul 2026. prnewswire.com/news-releases/trustradius-2026-b2b-buying-disconnect-report-reveals-ai-has-changed-how-buyers-research-but-not-what-they-trust-302825792.html
61. [61]Gartner. Survey Finds 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights. May 2026 (fieldwork Aug-Sep 2025). gartner.com/en/newsroom/press-releases/2026-05-20-gartner-survey-finds-sixty-nine-percent-of-b-two-b-buyers-turn-to-sales-reps-to-validate-ai-generated-insights
62. [62]Bain & Company. Losing Control: How Zero-Click Search Affects B2B Marketers. Mar 2025. bain.com/insights/losing-control-how-zero-click-search-affects-b2b-marketers-snap-chart/
63. [63]Bain & Company / Google (HBR). What B2Bs Need to Know About Their Buyers. Sep 2022. hbr.org/2022/09/what-b2bs-need-to-know-about-their-buyers
64. [64]Columbia Journalism Review / Tow Center. AI Search Has a Citation Problem. Mar 2025. cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php
65. [65]BBC. Representation of BBC News Content in AI Assistants. Feb 2025. bbc.co.uk/aboutthebbc/documents/bbc-research-into-ai-assistants.pdf
66. [66]EBU / BBC. News Integrity in AI Assistants. Oct 2025. ebu.ch/Report/MIS-BBC/NI_AI_2025.pdf
67. [67]The New York Times / Oumi. How Accurate Are Google's AI Overviews? Apr 2026. Reported at searchengineland.com/google-ai-overviews-accuracy-wrong-answers-analysis-473837
68. [68]Ahrefs. How Often Do AI Assistants Hallucinate Links? Sep 2025. ahrefs.com/blog/how-often-do-ai-assistants-hallucinate-links/
69. [69]Searchable (via Search Engine Journal). AI Answers About Your Locations Are Often Wrong. Jul 2026. searchenginejournal.com/ai-answers-about-your-locations-are-often-wrong-check-before-customers-do/582205/
70. [70]Aggarwal et al. GEO: Generative Engine Optimization. ACM SIGKDD 2024. arxiv.org/abs/2311.09735
71. [71]Bagga et al. E-GEO: A Testbed for Generative Engine Optimization in E-Commerce. Nov 2025. arxiv.org/abs/2511.20867
72. [72]Peng et al. Impact of AI on Developer Productivity (GitHub Copilot RCT). 2023. arxiv.org/abs/2302.06590
73. [73]Brynjolfsson, Li & Raymond. Generative AI at Work (NBER WP 31161). 2023. nber.org/papers/w31161
74. [74]Epoch AI. LLM inference price trends. 2025. epoch.ai/data-insights/llm-inference-price-trends
75. [75]Backlinko (OpenAI-reported milestones). ChatGPT user statistics. 2026. backlinko.com/chatgpt-stats
76. [76]NAIC. AI/ML insurer surveys. 2022-2025. content.naic.org/insurance-topics/artificial-intelligence
77. [77]NVIDIA. State of AI in Financial Services 2026. 2026. blogs.nvidia.com/blog/ai-in-financial-services-survey-2026/
78. [78]NVIDIA. State of AI in Healthcare and Life Sciences. 2025. blogs.nvidia.com/blog/ai-healthcare-life-sciences-survey-2025/
79. [79]NVIDIA. State of AI in Telecommunications 2026. 2026. blogs.nvidia.com/blog/ai-in-telco-survey-2026/
80. [80]Evident Insights. Evident AI Index 2025 (banking). 2025. evidentinsights.com/bankingbrief/heres-the-2025-evident-ai-index/
81. [81]Thomson Reuters. Generative AI in Professional Services 2025. Apr 2025. thomsonreuters.com/en/press-releases/2025/april/from-incubation-to-integration-generative-ai-adoption-nearly-doubles-as-professional-services-reach-crossroads
82. [82]Clio. 2025 Legal Trends Report. 2025. clio.com/resources/legal-trends/read-online/
83. [83]American Bar Association. 2024 AI TechReport. 2024. americanbar.org/groups/law_practice/resources/tech-report/2024/2024-artificial-intelligence-techreport/
84. [84]MHI / Deloitte. 2026 MHI Annual Industry Report. 2026. dcvelocity.com/technology/ai-continues-to-drive-major-disruptions-in-supply-chain-field-according-to-mhis-annual-industry-report
85. [85]Manufacturing Leadership Council (NAM). GenAI Adoption Surges. 2026. manufacturingleadershipcouncil.com/survey-genai-adoption-surges-as-manufacturers-continue-to-grapple-with-data-skills-issues-39942/
86. [86]Honeywell. 2025 retail research. Jan 2025. honeywell.com/us/en/news/press-releases/2025/01/honeywell-research-reveals-more-than-80-of-retailers-plan-to-increase-ai-capabilities-in-their-operations-in-2025
87. [87]IBM / NRF. State of Retail: consumer AI. Jan 2026. newsroom.ibm.com/2026-01-07-ibm-nrf-study-brands-and-retailers-navigate-a-new-reality-as-ai-shapes-consumer-decisions-before-shopping-begins
88. [88]Salesforce. State of Marketing 2026. 2026. salesforce.com/news/stories/state-of-marketing-2026/
89. [89]HubSpot. State of AI for Marketers. 2025. blog.hubspot.com/marketing/state-of-ai-report
90. [90]Content Marketing Institute. B2B Content Marketing Research. 2026. contentmarketinginstitute.com/b2b-research/b2b-content-marketing-trends-research
91. [91]EDUCAUSE (via GovTech). 2025 AI Landscape Study. 2025. govtech.com/education/higher-ed/survey-higher-ed-ai-adoption-faces-financial-policy-hurdles
92. [92]Gallup / RAND. Three in 10 Teachers Use AI Weekly. 2025. news.gallup.com/poll/691967/three-teachers-weekly-saving-six-weeks-year.aspx
93. [93]Digital Education Council. Global AI Student Survey. 2024. digitaleducationcouncil.com/post/what-students-want-key-results-from-dec-global-ai-student-survey-2024
94. [94]JLL. Global Real Estate Technology Survey. 2025. jll.com/en-us/insights/global-real-estate-cre-technology-survey
95. [95]RICS. AI in Construction Report 2025. Sep 2025 (2,200+ professionals; replaces the Q1 2025 Global Construction Monitor, which contains no AI measure). rics.org/news-insights/optimism-high-for-ai-in-construction-but-skills-shortages-and-integration-challenges-adoption
96. [96]NASCIO. 2025 State CIO Survey. 2025. nascio.org/resource-center/the-2025-state-cio-survey/
97. [97]Brookings. Where does federal AI spending stand in 2026? 2026. brookings.edu/articles/where-does-federal-ai-spending-stand-in-2026/
98. [98]MorganMyers. 2026 Farmer Trust in AI Report. 2026. The 48 percent figure is farmers using AI weekly or more. morganmyers.com/ai-agriculture-report-2026/
99. [99]Eliciting Insights (via Fierce Healthcare). 2nd Annual AI Adoption Survey. 2026. fiercehealthcare.com/ai-and-machine-learning/75-us-healthcare-systems-use-plan-use-ai-platform-2026
100. [100]Stack Overflow. 2025 Developer Survey. 2025. survey.stackoverflow.co/2025/ai
101. [101]Similarweb. Gen AI Stats 2026. 2026. similarweb.com/blog/marketing/geo/gen-ai-stats/
102. [102]Semrush. AI Search Traffic Study (AI visitors projected to pass traditional search for its tracked topics by early 2028). Jun 2025. semrush.com/blog/ai-search-seo-traffic-study/
103. [103]SparkToro / Datos. How People Use Google: 332M queries over 21 months. Oct 2024. sparktoro.com/blog/new-research-we-analyzed-332-million-queries-over-21-months-to-uncover-never-before-published-data-on-how-people-use-google/
104. [104]IBM. Global AI Adoption Index 2023. Jan 2024. newsroom.ibm.com/2024-01-10-Data-Suggests-Growth-in-Enterprise-Adoption-of-AI-is-Due-to-Widespread-Deployment-by-Early-Adopters
105. [105]Goldman Sachs. Generative AI could raise global GDP by 7%. Mar 2023. goldmansachs.com/insights/articles/generative-ai-could-raise-global-gdp-by-7-percent
106. [106]International Monetary Fund. AI will transform the global economy. Jan 2024. imf.org/en/blogs/articles/2024/01/14/ai-will-transform-the-global-economy-lets-make-sure-it-benefits-humanity
107. [107]Grand View Research / Fortune Business Insights / Precedence Research. AI market size reports. 2025.