In 2026, AI assistants collectively generate sessions equivalent to 56% of global search engine volume, but only 51.6% of those sessions involve search-style queries, putting AI's true search-competing footprint at approximately 28% of global search volume and 17% in the US (Graphite.io, March 2026; Harvard/NBER Working Paper 34255). ChatGPT alone holds 89% of all global AI sessions, equivalent to roughly 20% of total global search-related traffic, making it the world's second-largest information discovery platform by session volume. Google's absolute search volume has not fallen; total information discovery grew 26% between Q1 2023 and Q4 2025, but Google's share of that combined discovery landscape dropped from 89.3% to 57.6% between December 2022 and December 2025. The shift is structural, already embedded in how buyers research vendors, and invisible to most standard attribution models.

TL;DR

• AI assistants now generate sessions equal to 56% of global search volume, but corrected for task-execution and creative sessions, the true search-competing figure is ~28% globally and ~17% in the US

• ChatGPT holds 89% of all global AI sessions, roughly equivalent to 20% of total global search-related traffic, the world's second-largest information discovery platform

• Google's share of combined information discovery (search + AI) fell from 89.3% to 57.6% between December 2022 and December 2025 (not a prediction, already happened)

• Total search volume has not collapsed; the full discovery landscape grew 26% from Q1 2023 to Q4 2025

• 25% of B2B buyers now use generative AI for vendor research, typically before a shortlist is formed; brands invisible in AI may be losing deals they never knew existed

There's a number going viral in marketing right now: 56%.

AI assistants now generate sessions equivalent to 56% of global search engine volume, according to new research by Ethan Smith, CEO of Graphite.io. Published in March 2026, it's the most rigorous methodology applied to this question yet. Smith set out to stress-test the hype, not confirm it.

The finding landed exactly as you'd expect. One camp declared search dead. The other dismissed the whole thing. Both are working from incomplete information.

The "Google is dying" crowd treats the 56% figure as proof that search is being replaced. It isn't. The "nothing has changed" crowd treats the methodological corrections in this research as reasons to ignore it entirely. That's a bigger mistake.

This article pulls the raw dataset behind Ethan's findings, adds original analysis, and delivers a sharper picture: AI is genuinely transforming how buyers discover information, but several distinct stories are hiding inside the headline number. Some of those make the shift look less alarming than the hype suggests. Others make it considerably more urgent.

The Measurement Problem: Why Every AI vs. Search Comparison You've Read Is Probably Wrong

Most AI vs. Google comparisons that circulated in 2025 and early 2026 used a fatally flawed method: desktop web traffic to google.com versus chatgpt.com. By that measure, AI looks like a rounding error against search.

The problem is structural: 83% of AI usage happens in mobile apps, not on the web. Comparing desktop web sessions is like measuring the smartphone market by looking at Apple.com's desktop traffic. You'd conclude nobody uses iPhones.

Graphite's methodology corrects for this. Smith compared full web and app sessions across five major AI platforms (ChatGPT, Gemini, Claude, Perplexity, Grok) against six search engines (Google, Bing, Yahoo, Yandex, Baidu, DuckDuckGo). The result: 45 billion monthly AI sessions globally by Q4 2025. In the US specifically, AI accounts for 34% of search-equivalent traffic.

Previous estimates weren't slightly wrong. They were off by a factor of four to five.

The US-global gap deserves attention on its own. AI usage worldwide runs roughly 7x higher than in the US alone. Most marketing media is written by US-based authors covering US data. If your brand operates internationally, or if you're targeting markets like the UK, Germany, Brazil, or Southeast Asia, the inflection point has already passed. The strategic window you're waiting to open has been open for over a year.

ChatGPT's Real Market Share (and the Challenger Nobody Is Watching)

ChatGPT dominates. By Q4 2025, it held 89% of all global AI sessions and approximately 20% of total global search-related traffic. That makes it the second-largest information discovery platform in the world by session volume.

The most interesting platform story in the raw data isn't ChatGPT, though. It's Grok.

Grok barely registered in January 2025, running at 1.2 million sessions per month. By December 2025, it was generating 1.4 billion sessions monthly, a position that puts it ahead of Perplexity entirely. That's a 116,000% increase in twelve months. Marketing teams monitoring ChatGPT, Perplexity, and Gemini are already behind if Grok isn't on their radar.

The platform landscape is not stable. Build your strategy assuming it will keep shifting, because it will.

AI and Search Are Not Zero-Sum

The dominant narrative since Gartner's 2024 prediction ("search engine volume will fall 25% by 2026 due to AI chatbots") has been one of substitution: AI takes share, search loses it. Everything is a trade-off.

The data doesn't support this framing.

Total information discovery sessions (search engines plus AI Asking prompts combined) grew 26% worldwide from Q1 2023 to Q4 2025. Search volume hasn't declined in absolute terms. The pie expanded. AI took a large slice of the new growth without eating the existing pie.

There's a useful historical parallel here. In 2010, Wired published its "The Web Is Dead" cover story as mobile app usage surged. Web traffic in absolute terms kept growing for the following decade. New categories of digital behavior emerged without simply replacing old ones. Consumers don't delete old behaviors. They stack new ones on top.

AI is following a similar pattern. It's not cannibalizing search sessions in aggregate. It's creating new types of information-seeking behavior that didn't exist three years ago: multi-turn research conversations, real-time synthesis of multiple sources, task completion inside a single interface. These are distinct from search. They coexist with it.

Why Is My Organic Traffic Dropping If Overall Search Demand Is Stable?

Total volume holding doesn't mean nothing changed for marketers. Something changed significantly. Just not the thing most people are measuring.

According to the Datos/SparkToro Q4 2025 State of Search Report, Google searches per US user fell approximately 20% year-over-year, even as total volume held flat. Rand Fishkin described the mechanism directly: "AI answers have dramatically altered the way many users engage with Google, answering their questions before they ever need to click on an organic result or perform a second, third, or fourth search."

The buyer journey got shorter. A user who would have run five Google searches to research a vendor decision in 2023 is now running one or two, having resolved most of their questions in a ChatGPT or Perplexity session. The final click still happens through Google. But most of the discovery work happened elsewhere and left no trace in your attribution model.

If your organic traffic is down while pipeline is holding, that's the most likely explanation. Demand didn't disappear. Attribution broke.

This divergence between total search volume and per-user frequency is exactly what Answer Engine Optimization (AEO) audits are built to surface: where your brand appears (and where it doesn't) in the shortened, AI-assisted buyer journey.

Not All AI Usage Is Search: The Breakdown That Changes Your Numbers

The 56% figure deserves one more correction before you base strategy on it.

A Harvard/NBER analysis of more than one million ChatGPT prompts (Working Paper 34255) mapped user intent across session types:

• Asking (search-equivalent queries): 51.6%

• Doing (task execution: coding, drafting, analysis): 34.6%

• Expressing (conversational, creative): 13.8%

Only the Asking category competes with search. The other 48.4% of AI sessions represent a genuinely new behavior category: AI as a productivity layer, not a discovery engine. Someone using ChatGPT to draft a contract isn't replacing a Google search. Someone asking it to debug Python code isn't in a purchase journey. These sessions have no search equivalent.

When corrected for non-search behavior, AI's actual search-equivalent footprint drops to approximately 28% of global search volume and 17% in the US (roughly half the headline 56% figure).

Apply that correction to Graphite's figures and AI's actual search-equivalent footprint is still a massive, strategy-changing shift. But it's materially different from the headline number, and the difference matters for how you prioritize your response.

If you're cutting SEO investment because AI is supposedly replacing search, you're reacting to a figure that overstates the substitution effect by roughly 2x.

The B2B Purchase Decision Shift Is Where It Gets Urgent

Session volume data tells one story. What B2B buyers are actually doing inside those sessions tells a different one.

According to the Eight Oh Two / Search Engine Land 2026 AI and Search Behavior Study (January 7, 2026):

• 47% of buyers say AI influences which brands they trust

• 37% now start product research with AI instead of Google

• 85% still verify AI answers elsewhere, meaning SEO remains critical in the validation stage

• 63% expect to use AI more in 2026 for finding products and services

For B2B challenger brands specifically, Demandbase and Column Five data makes the implication concrete: 25% of B2B buyers now use generative AI over traditional search for vendor research. That's happening in the category awareness and vendor shortlisting phase, before decision-making criteria are even set. Being invisible in AI-powered discovery doesn't just reduce traffic. It means being off the shortlist before an RFP is ever written.

The revenue implication is asymmetric. Missing a Google click costs you a potential visitor. Missing AI discovery costs you the deal before you knew the opportunity existed.

Read our full breakdown of AI marketing strategy for challenger brands for the broader B2B implications.

Three Things the Raw Data Shows That the Headlines Are Missing

The Graphite study released headline figures. The underlying dataset contains three additional stories that shift how you should set strategy in 2026. These aren't widely covered because most coverage summarized the press release rather than working through the raw numbers.

1. The AI Growth Plateau That Started in July 2025

The 56% figure doesn't describe a rising trend. It describes a plateau.

AI "Asking" sessions peaked in July 2025 at 23 billion per month globally. By December 2025, that figure measured at 22.5 billion, a 2.3% decline over five months. The global acceleration that produced the 56% headline happened between late 2022 and mid-2025. That wave has crested.

The strategic implication: the right question isn't "how do we respond to an AI explosion?" The explosion has already happened. The right question is "how do we get positioned in a landscape that has already structurally shifted?" Those questions require different urgency levels and different resource allocations.

One important exception: the US. American AI usage grew 300% year-over-year by December 2025 and is still accelerating. If your primary market is the US, you're in the earlier part of the adoption curve. The global plateau tells you where the US is heading. Use it to get ahead.

2. Google Lost Nearly a Third of Its Discovery Gravity in Three Years

Google's share of total information discovery (search and AI sessions measured together) fell from 89.3% in December 2022 to 57.6% in December 2025: a 31.7 percentage point drop in three years that has already happened, not a projection.

The platform share numbers in the raw data are starker than the session growth figures suggest.

Google held 89.3% of total information discovery (search plus AI combined) in December 2022. By December 2025, that figure was 57.6%. A 31.7 percentage point collapse in three years.

Most market share analyses separate AI sessions from search sessions, comparing them as distinct categories. That framing misses the point. They're competing for the same attention window: the moment a buyer wants to find something out. Measured across the full discovery landscape, Google lost nearly a third of its gravitational pull in 36 months. That's not a prediction. It already happened.

For challenger brands, this isn't cause for panic about Google. It's a reframing of the opportunity. The discovery landscape expanded significantly, and Google's dominance within it weakened. First-mover positioning in AI-powered discovery is worth more than it looks when measured against traditional search volume alone.

Understanding the difference between GEO and SEO gives you the strategic frame for building a content system that performs across both environments.

3. Long-Tail Traffic Is Absorbing the AI Impact While Total Demand Holds

This is the most practically important insight for challenger brand content teams, and it's rarely framed this way.

The plateau in global AI growth doesn't mean the impact on your content is plateauing. Long-tail informational queries, the foundation of most B2B challenger brand content programs, are precisely the type that AI resolves before users click. A buyer researching "what is answer engine optimization" or "how does CAC attribution work" is getting that question answered inside ChatGPT now. Your blog post ranking third for that query isn't generating the click it used to.

Traffic to long-tail informational pages can fall 40-60% while actual commercial demand for your product holds constant. If you're only measuring organic traffic as a proxy for brand health, you're looking at a broken instrument.

The buyers are still out there. They're just completing more of their research before you ever see them.

What This Means for Your Strategy: Three Decisions to Make Now

1. Audit Your AI Search Visibility Before You Do Anything Else

Most challenger brands are invisible in ChatGPT, Perplexity, and Gemini. Not because their content isn't strong enough to be cited there. Because they've never checked.

Exposure Ninja's 2026 research found that AI search traffic converts at 14.2% versus Google organic's 2.8%. Visitors who arrive via AI citations are deep in consideration mode: they've already had the AI answer their research questions and they're ready to evaluate vendors. Only 25.7% of marketers are actively developing content for AI citation today.

That's a first-mover window. It's open now. It won't stay open at current valuations.

DOJO AI's AEO Agent audits your brand's visibility across the major AI answer engines, identifying where you're cited, where your competitors appear instead of you, and what specific content changes move the needle. Start there before you make any budget or content strategy decisions.

Our comprehensive AEO guide breaks down the mechanics and what to optimize for across each platform.

2. Stop Treating SEO and AEO as an Either/Or Decision

The biggest strategic mistake following data like this is binary thinking: pivot everything to AI, or stay the course on SEO and ignore the shift.

Total search volume is growing. 85% of buyers still verify AI answers through subsequent searches. If you abandon SEO investment now, you're exiting a channel that still drives a significant part of the discovery and validation journey. The correct question isn't "SEO or AEO?" It's "how do I make sure my content performs across every discovery surface my buyers use?"

The answer turns out to be a single coherent content strategy, not two separate ones.

The structural content signals that improve AI citation rates overlap significantly with strong SEO fundamentals: answer-first paragraphs, clear H1/H2 hierarchy, FAQ schema markup, cited statistics, and direct definitional content all serve both goals. You're building one system that performs in two environments, not maintaining two separate content programs.

Our guide on how to rank in AI search covers the specific structural content changes that move the needle most.

3. Fix Your Attribution Model Before You Change a Single Budget Line

If organic clicks are declining but pipeline is stable, don't cut the SEO budget.

That pattern (volume down, pipeline holding) is the signature of AI search interception. Buyers are finding your brand through AI-assisted research, resolving their questions in ChatGPT or Perplexity, and arriving at your website far later in their journey with a much higher intent signal. The tracked click happens. But most of the brand consideration work happened before it, in places your current attribution model can't see.

Cutting SEO investment because attributed traffic looks worse, while buyers are still discovering and trusting your brand through AI-cited content, is a measurement problem masquerading as a strategic problem. Fix the measurement first.

Moving beyond last-click attribution is the starting point for getting an honest read on what's actually driving pipeline in an AI-assisted buyer journey.

The Bottom Line

Gartner's "25% search volume decline by 2026" prediction was wrong on the volume call. Total search isn't declining. But it was right on the market share call: Google's share of combined information discovery fell from 89% to below 58% in three years. That's not a prediction that came true late. It already happened.

The AI shift is real, significant, and already structurally embedded in how buyers research vendors. It's also not the binary replacement story that generates the most engagement on LinkedIn.

What's actually happening: the discovery landscape expanded considerably. New behavior categories emerged alongside traditional search. AI captured most of the growth while traditional search held its absolute volume. And the brands that built AI visibility early are now appearing in buyer journeys their competitors don't know exist.

For challenger brands, that's the real risk. Not that search is dying. That you're invisible in an expanded discovery landscape while better-resourced competitors are already building presence there.

The window to be a first mover in AI-powered discovery is still open. It won't stay open at current valuations indefinitely.

Run your AEO audit first. DOJO AI gives you a clear, unified picture of where your brand stands across ChatGPT, Perplexity, Gemini, and the full AI discovery landscape, alongside your traditional search position. Start your free trial and run your first audit in under 24 hours.

Frequently Asked Questions

Q: How does ChatGPT compare to Google search in terms of volume in 2026?

As of Q4 2025, ChatGPT generates sessions equivalent to approximately 20% of total global search-related traffic and holds 89% of all global AI sessions, according to Graphite.io research published in March 2026. AI assistants combined account for sessions equivalent to 56% of global search engine volume, but that headline figure includes task-completion and creative sessions that don't compete directly with search. The true search-equivalent share is closer to 28% globally. Google remains the world's largest information discovery platform; ChatGPT is now second.

Q: Is the 56% AI vs Google search figure accurate?

The figure is real but requires two corrections before you use it for strategy. First, it's based on full session methodology that accounts for mobile app usage, where 83% of AI activity happens; previous comparisons that used only desktop web traffic understated AI volume by a factor of four to five. Second, only 51.6% of AI sessions are search-equivalent "Asking" queries; the remaining 48.4% are task execution (coding, drafting, analysis) and creative or conversational sessions that have no search equivalent. Once you apply that correction, AI's actual search-competing footprint is approximately 28% of global search volume and 17% in the US, still large enough to reshape your strategy, but materially different from the headline.

Q: Has Google actually lost market share to AI search in 2025-2026?

Yes, and the scale is larger than most coverage reflects. Google held 89.3% of combined information discovery (search and AI sessions measured across the same landscape) in December 2022. By December 2025 that figure was 57.6%, a drop of 31.7 percentage points in three years. Absolute search volume did not decline; the discovery landscape expanded by 26% overall, and AI captured most of that new growth. But Google's gravitational pull within the full discovery ecosystem weakened significantly. This is historical data, not a forecast.

Q: Why is my organic traffic down if total search volume hasn't fallen?

AI assistants are now resolving most of the research phase of the buyer journey before users reach a search result. A buyer who ran five Google searches to evaluate a vendor in 2023 now resolves most of those questions inside a single ChatGPT or Perplexity conversation, then performs one or two confirmatory searches at the end. The final click still registers in Google, but the discovery work happened elsewhere and left nothing in your attribution model. Google searches per US user fell approximately 20% year-over-year in Q4 2025, according to the Datos/SparkToro State of Search Report, even as total volume held flat. If your pipeline is stable while organic traffic falls, the measurement is broken. Demand didn't disappear.

Q: What happened to Grok in 2025, and why does it matter for AI search strategy?

Grok went from 1.2 million monthly sessions in January 2025 to 1.4 billion by December 2025, a 116,000% increase in twelve months, putting it ahead of Perplexity in session volume. Most AI visibility monitoring focuses on ChatGPT, Gemini, and Perplexity; Grok is being missed almost entirely by marketing teams. The platform landscape is not stable. Grok's trajectory shows how fast a new entrant can become strategically material. Any brand-level AI citation monitoring that excludes Grok is already working from an incomplete picture.

Q: Should I prioritize AEO over SEO in 2026?

Neither replaces the other, and treating them as an either/or choice is the most common strategic mistake following this data. Total search volume is growing in absolute terms, and 85% of buyers still verify AI answers through subsequent searches, meaning SEO remains central to the validation stage of the buyer journey. The content signals that improve AI citation rates overlap closely with strong SEO fundamentals: answer-first paragraphs, clear H2 structure, FAQ schema markup, cited statistics, and direct definitional content all serve both goals. One content system built to perform across both discovery environments is more effective and more efficient than two separate programs.

Q: How does AI search affect B2B vendor research and buying decisions specifically?

According to the Eight Oh Two / Search Engine Land 2026 AI and Search Behavior Study, 37% of B2B buyers now start product research with AI instead of Google, 47% say AI influences which brands they trust, and 63% plan to use AI more for finding products and services in 2026. Demandbase data puts 25% of B2B buyers already using generative AI for vendor research, typically during the awareness and shortlisting phase, before decision criteria are even set. The revenue implication is asymmetric: missing a Google click costs you a potential visitor; being absent from AI-powered discovery can remove you from a shortlist before you knew the opportunity existed.

Further Reading

• Answer Engine Optimization (AEO) Guide: How to Rank in AI Search

• ChatGPT vs Perplexity vs Gemini: AEO Platform Comparison

• GEO vs SEO: What's the Difference and Why It Matters in 2026

• How to Rank Higher in AI Search: A Practical Guide

• Beyond Last-Click Attribution: AI Marketing Revenue Correlation

Data sources: Graphite.io / Ethan Smith, "AI Is Much Bigger Than You Think" (March 2026, primary dataset); Harvard/NBER Working Paper 34255, "How People Use ChatGPT"; Eight Oh Two / Search Engine Land, "2026 AI and Search Behavior Study" (January 7, 2026); Datos/SparkToro Q4 2025 State of Search Report; Exposure Ninja, 2026 AI Search Statistics; Gartner Newsroom, February 2024.