Afternoon BriefAI Search & Discovery

6.77 Million AI Sessions Exposed the Discovery Engine Hierarchy and Most Brands Are Optimizing for the Wrong One

I'm Jaxon Parrott. Previsible just published 6.77 million sessions of AI discovery data. ChatGPT looks like a monopoly at 92.4% of standalone traffic. The fragmentation data tells a different story. Here is what the multi-engine numbers actually mean for where your brand shows up in AI answers.

Jaxon Parrott
Jaxon ParrottJul 8, 2026

Previsible just analyzed 6.77 million AI-driven sessions across 166 websites spanning 19 months. ChatGPT commands 92.4% of standalone LLM referral traffic. That number looks like a monopoly. It is not. The discovery market is fragmenting faster than anyone building single-engine strategies can absorb, and the fragments are where the conversion happens.

The Headline Number Is the Wrong Number

Every marketing deck in circulation this week will cite Previsible's 92.4% figure and conclude ChatGPT is the only engine that matters. I have run citation pipelines for nearly a decade. Here is what that number actually says when you read it with the context Previsible provides: ChatGPT's share is measured against standalone LLM platforms only. Google's AI Overviews and AI Mode represent more AI-influenced traffic than all standalone LLMs combined. Previsible excludes Google from the comparison because it operates on a fundamentally different measurement model.

That exclusion changes the conclusion. The brand that "optimizes for ChatGPT" and ignores Google AI is optimizing for a subset of a subset. BrightEdge data compiled by Axis Intelligence shows AI Overviews now appear in 73% of US informational queries, up from 47% at the start of the year. A 58% year-over-year increase. Google's own AI features reduced clicks to the top organic page by 58% for queries with AI Overviews, and full AI Mode produces a 93% zero-click rate.

The 92.4% is real inside its frame. The frame does not include the largest AI discovery surface on the internet. If you are building your AI visibility strategy around one engine's traffic share, you are building on a number that does not account for the surface sending more AI-driven traffic than every LLM assistant combined.

Claude Grew 64x and Nobody Adjusted Their Strategy

Previsible's data shows Claude went from 133 sessions in November 2024 to 8,528 in May 2026. A 64x increase. Claude overtook Perplexity in March 2026 for standalone LLM referral sessions and has stayed ahead since.

The growth correlates with Anthropic's expansion into coding tools, professional workflows, and enterprise adoption. Claude Code and Claude Cowork are driving technical buyer traffic that most brands do not have a visibility strategy for. Previsible puts it directly: "If your audience includes technical buyers, developers, or professional services, Claude visibility is becoming a real factor, and the window for early positioning is now."

StatCounter data compiled by Axis Intelligence shows Claude's referral share rose from 0.30% in April 2025 to 2.66% in April 2026. That looks small until you see the conversion data. Digital Bloom's analysis found Claude generates the highest conversion rate of any AI platform at 16.8%. For context, AI referral traffic overall converts at 7.1% via Similarweb data, already second only to paid search at 7.8% and 2.5 times higher than Google organic.

Claude sends fewer sessions. Those sessions convert at more than double the rate of any other AI engine. A brand invisible in Claude is invisible to its highest-converting discovery surface.

The Fragmentation Index Dropped 50% in Twelve Months

Axis Intelligence built an AI Search Fragmentation Index that synthesizes referral share concentration, monthly share shift, and conversion quality into a single score. The index dropped from 178.5 in April 2025 to 88.3 in April 2026. A 50.5% decline in market concentration in one year.

The structural story is clear from StatCounter's monthly data. ChatGPT's referral share fell from 84.21% in April 2025 to 76.85% in April 2026. Gemini rose from 2.31% to 9.00%. Perplexity held between 7% and 8%. Claude climbed from 0.30% to 2.66%. Every challenger took share from ChatGPT, and the conversion quality across all platforms rose from 3.1 times Google organic to 4.8 times over the same period.

Axis Intelligence's conclusion should end the single-engine debate: "Any single-platform AI search strategy built in 2025 is structurally obsolete in 2026."

The market is simultaneously splitting across more platforms and delivering higher-quality traffic per session. Fragmentation is not a problem. It is the operating reality.

Perplexity Peaked and Copilot Collapsed. The Lesson Is Not About Them.

Perplexity peaked at 17,507 monthly sessions in March 2025 and fell 61% to 6,788 by May 2026. Copilot peaked at 8,651 sessions in August 2025 and collapsed 96% to 339 by May 2026. Both are shifting toward retaining users inside their own ecosystems rather than referring them to external sites.

The lesson is not that these engines failed. The lesson is that the measurement surface changed underneath the strategy. Copilot's referral decline does not mean B2B buyers stopped using it. It means Copilot is keeping users inside the experience. PPL Studio's analysis of 120 brands found Copilot citation share inside enterprise workflows is the highest-correlated metric with B2B SaaS pipeline lift, at r = 0.71 across a 40-brand cohort. Referral traffic went to zero. Pipeline influence did not.

This is why measuring AI visibility by referral traffic alone is measuring the wrong thing. The brand that appears in a Copilot answer inside a buyer's Microsoft 365 workflow never generates a referral click. It generates a shortlist position that converts at the decision layer, not the discovery layer. If you are still measuring AI success by sessions in GA4, you are watching the scoreboard from the wrong game.

35% of Buyers Start With AI. 13.6% Start With Search Engines.

Similarweb's Market Research Panel found that 35% of US consumers now start product discovery with AI tools. 13.6% start with search engines. That inversion happened in January 2026 and it is not reversing.

Gartner's forecast of a 25% decline in traditional search volume by 2026 is tracking close to actual numbers. Google referral traffic to publishers dropped 33% globally in 2025, with the US down 38%. AI referral sessions grew 1,200% over the same period.

I built AuthorityTech and coined Machine Relations precisely because this multi-engine reality was visible two years before Previsible's data confirmed it. The discipline of earning AI engine citations through trusted third-party sources is not an optimization tactic for one engine. It is the architecture that compounds across all of them. Placements in publications that Google AI, ChatGPT, Claude, Perplexity, and Gemini all trust are the only placements that survive a market where every quarter redistributes traffic share.

What Multi-Engine Citation Architecture Costs You to Ignore

PPL Studio's 120-brand cohort data provides the mid-2026 citation share thresholds that should replace every metric you are currently tracking. 1% to 3% citation share means your brand surfaces occasionally but is not part of the default shortlist. 5% to 8% means consistent presence in roughly 1 of every 15 relevant answers. 10% to 15% means shortlist position, where citation traffic converts at 2.3 to 3.6 times the rate of paid social cold traffic for DTC brands. Above 20% means category leader status, where the engine treats you as the canonical answer.

Brands below 5% citation share on their top commercial queries are losing 30% to 55% of informational traffic year over year without a corresponding paid offset. The cost of ignoring multi-engine citation is not hypothetical. It is measurable. It is compounding. And it is accelerating every quarter the market fragments further.

Previsible's own recommendation mirrors what I have been building at AuthorityTech since 2024: "create citation-worthy evidence, build authority across trusted third-party sources, make websites accessible to AI systems to read and extract, optimize for answer journeys, and measure business impact." That is Machine Relations described by a research firm analyzing 6.77 million sessions without ever using the term. When independent data arrives at the same architecture, the discipline is not a framework someone invented. It is the operating reality the market discovered on its own.

The question is no longer whether your brand needs multi-engine citation architecture. The question is whether you build it before your competitors do, or after.

FAQ

What is the current AI search engine market share in 2026?

StatCounter data through April 2026 shows ChatGPT at 76.85% of AI chatbot referral share, followed by Gemini at 9.00%, Perplexity at 7.73%, Copilot at 3.76%, and Claude at 2.66%. Previsible's analysis of 6.77 million sessions gives ChatGPT a higher figure of 92.4% because their methodology measures standalone LLM referral traffic specifically. Both studies exclude Google AI Overviews and AI Mode, which represent more AI-driven traffic than all standalone LLMs combined.

How fast is Claude growing as an AI discovery engine?

Claude grew 64x in referral sessions between November 2024 and May 2026, overtaking Perplexity in March 2026. StatCounter shows its referral share climbing from 0.30% to 2.66% over 12 months. Digital Bloom analysis found Claude generates the highest conversion rate of any AI platform at 16.8%, compared to the overall AI referral conversion rate of 7.1%.

What is the AI Search Fragmentation Index and why does it matter?

The AI Search Fragmentation Index is a cross-source metric developed by Axis Intelligence that combines market concentration, momentum shift, and conversion quality. It dropped from 178.5 to 88.3 between April 2025 and April 2026, a 50.5% decline in market concentration. The fragmentation is occurring alongside rising conversion quality, meaning single-engine strategies face both shrinking share and missed conversion on the engines they ignore.

How does Machine Relations address multi-engine AI visibility?

Machine Relations is the discipline of earning AI engine citations through trusted third-party sources. I coined the term in 2024 after documenting that earned media placements in high-authority publications are the citations AI engines retrieve across ChatGPT, Google AI, Claude, Perplexity, and Gemini simultaneously. AuthorityTech built its model around citation architecture that compounds across all discovery surfaces, rather than optimizing for one engine at a time.

What percentage of buyers start product discovery with AI tools in 2026?

Similarweb's Market Research Panel found that 35% of US consumers start product discovery with AI tools as of January 2026, compared to 13.6% who start with search engines. This inversion means brands without multi-engine AI visibility are invisible to more than a third of their potential buyers before the purchase journey even begins.

Additional source context

Continue Exploring

The most relevant next read from elsewhere on AuthorityTech.

BlogJun 30
AI Citation: What It Actually Means When an AI Engine Names Your BrandExecution playbook
BlogJul 2
AI Citation Accuracy in 2026: What 2.2 Million Citations Reveal About How Engines Select SourcesExecution playbook
Glossary
Source CredibilityRelated glossary concept
ResultsFeb 17
GC Finder Press Release Distribution: AI Agent Platform for Government Contract DiscoveryLive proof