ChatGPT's AI Referral Lead Fragments as Claude and Gemini Surge

What happened

ChatGPT's share of AI chatbot referral traffic has fallen to its lowest level on record, according to StatCounter Global Stats, whose recent press release was titled bluntly: "ChatGPT falls to all-time low as AI chatbot referral market continues to fragment." The story is no longer about one assistant sending most of the AI-referred traffic on the web, it is about a market splitting into several meaningful sources at once.

The exact figures depend on which dataset you read, and the spread is instructive. StatCounter's worldwide measurement put ChatGPT's share of AI-assistant referrals at roughly 76.85% in a recent month, down from about 84.21% a year earlier. Similarweb data cited by PPC Land tells a steeper story for general web traffic: ChatGPT's share of worldwide generative-AI website traffic fell from 76.4% to 52.7%, while Claude tripled from 1.6% to 8.9% over the same window, an OpenAI decline of about 23.7 percentage points in a single year, with erosion accelerating in the back half. And when you narrow to B2B referrals specifically, the fragmentation is starkest: averaged across March–April 2026, ChatGPT sat near 62.6%, Claude reached about 18.5%, Gemini about 10.6%, and Perplexity roughly 7.3%.

The shifts coincided with a wave of user churn that commentators labeled the "#QuitGPT" movement. Reporting estimated that over 2.5 million people deleted ChatGPT in favor of alternatives, with app uninstalls spiking sharply on individual days. Whatever the precise catalyst, the direction is consistent across independent measurement firms: the AI-referral pie is being carved up, and Claude in particular has moved from an also-ran to the number-two source of AI-referred traffic in the B2B datasets.

Why it matters for practitioners

For a bootstrapped, product-led founder who has started to see real signups arrive from AI assistants, this is the moment the strategy has to mature. Optimizing your content and product to be cited by a single dominant assistant made sense when one platform sent the overwhelming majority of AI traffic. It is now a concentration risk. The discipline of getting recommended by AI assistants, answer-engine optimization, has to become explicitly multi-platform, the same way SEO tooling long ago had to account for more than one search engine.

1. Single-platform AEO is now a losing bet. If Claude, Gemini, and Perplexity together account for something like a third or more of B2B AI referrals, tuning only for ChatGPT leaves a growing share of your addressable AI-referred audience on the table. Worse, each assistant sources and weights citations differently, Perplexity leans on fresh, linkable sources; Claude and Gemini synthesize differently again, so a one-engine playbook is not just incomplete, it is structurally fragile.

2. You cannot optimize what you cannot see. The prerequisite for any of this is measurement. Most default analytics setups still bucket AI-assistant visits into generic "referral" or "direct," which makes it impossible to know whether ChatGPT, Claude, or Gemini is actually driving your signups. Until you can attribute traffic to specific assistants, you are optimizing blind. This is exactly the gap that tools like Plausible closed by shipping a dedicated AI Assistants channel that surfaces ChatGPT, Claude, Gemini, and Perplexity as first-class sources rather than lumping them together.

3. Fragmentation favors the nimble. A splintered landscape is bad news for anyone who bet everything on one channel and good news for founders willing to spread a few small bets. The content that gets cited across assistants tends to share traits: it is data-rich, factually specific, frequently updated, and structured for extraction. That is squarely within reach of a small team, and it is a place where a focused bootstrapped product can out-execute a larger competitor that treats AI referral as an afterthought.

Key details

• StatCounter: ChatGPT's worldwide AI-assistant referral share hit a record low (~76.85% in a recent month, down from ~84.21% a year earlier)
• Similarweb (via PPC Land): ChatGPT's share of worldwide gen-AI website traffic fell from 76.4% to 52.7%; Claude tripled from 1.6% to 8.9%
• OpenAI decline: ~23.7 percentage points lost in a single year, with erosion accelerating in the second half
• B2B referrals (Mar–Apr 2026 avg): ChatGPT ~62.6%, Claude ~18.5%, Gemini ~10.6%, Perplexity ~7.3%
• Claude's rise: from a low-single-digit also-ran to the #2 B2B AI referral source
• #QuitGPT movement: an estimated 2.5M+ users reported switching away from ChatGPT, with sharp single-day uninstall spikes
• Datasets diverge: worldwide vs. B2B vs. general web traffic all show fragmentation, at different magnitudes

Market implications

The most useful way to read the conflicting numbers is not to fixate on any single figure but to notice that every independent measurement points the same direction: concentration is falling. For founders, that reframes AI referral from "a ChatGPT channel" to "a portfolio of channels," and portfolios demand different management than a single bet. The practical response is to treat AEO as a multi-engine discipline within your broader SEO and content strategy, producing the structured, verifiable content that assistants of every stripe tend to synthesize, and checking how each one actually cites you rather than assuming.

Measurement is where most teams will win or lose. The founders who adapt fastest will be the ones who can already see, in their analytics, which assistant sent which visitor and what that visitor did next. That visibility turns a vague sense that "we're getting some AI traffic" into a concrete allocation decision: double down where conversion is strong, ignore where it is noise. The tooling to do this is now available in mainstream and privacy-first analytics alike, and companies like Plausible have made a point of surfacing AI-assistant channels by default, partly because their own AI-referred traffic has grown fast enough to matter.

The broader takeaway is that the AI-referral channel is following the same maturation curve as every channel before it. It began concentrated, it is fragmenting, and it will eventually require the same portfolio discipline that search and social already demand. Bootstrapped founders who internalize that now, by measuring per-assistant and optimizing for more than one engine, will be positioned to ride the channel as it grows, rather than waking up dependent on a single platform whose dominance is visibly slipping.

Related resources

• SEO Tools Landscape, Why answer-engine optimization now has to span multiple assistants
• Analytics Landscape Analysis, Tracking AI-assistant referral traffic as a distinct, per-source channel
• Plausible Case Study, A privacy analytics tool that surfaces AI-assistant channels by default