Claude Is Now the #2 Source of AI Referral Traffic for B2B Brands: What That Means for Your Brand
In Goodie's May 2026 report on a B2B brand panel, Claude reached 18.5% of average measurable B2B AI referral share for March–April 2026, behind ChatGPT at 62.6%. That made Claude the second-largest source in the study's reported comparison. The practical lesson is straightforward: watching only ChatGPT leaves blind spots. Brands should track which assistants mention them, recommend them and send attributable visitors, while keeping referral reporting separate from visibility data. Claude's rise is an opportunity to diversify measurement, not a reason to abandon ChatGPT.
What the report actually measured
Goodie published its second AI referral report on May 21, 2026. Every referral-share figure in this article comes from that report, which studied a panel of B2B brands: it compared a 2025 B2B brand panel with a later anonymized B2B brand panel and drew on GA4 referrer information plus supporting external signals. The report is linked in full under Sources below. Its headline comparison showed ChatGPT dropping from about 89% in the earlier wave to 62.6% in the March–April 2026 average, while Claude rose from 1.4% to 18.5%. It is one B2B referral study, not a measure of overall AI usage or market share, so read the figures as a dated snapshot of that panel.
The report warns that the headline engine percentages are cross-source brand averages rather than slices of one single denominator. Don't add the figures as if they describe the exact traffic distribution of every business.
Referral share is not usage share
An assistant can be popular but send little traffic because users finish their work inside the chat. Another can attract fewer users but generate more clicks for research-heavy tasks. These are different behaviors. Referral share measures identifiable visits, not the proportion of all AI answers mentioning a brand.
Industry mix matters too. A B2B software publisher, a local services company and an ecommerce retailer may see very different assistants in their acquisition reports. Use your own data to decide where to invest.
Why tracking multiple engines is now a basic requirement
Suppose ChatGPT recommendations improve but Claude consistently names a competitor. A single-engine dashboard would call that progress while overlooking a meaningful portion of potential buyer discovery. Track a consistent prompt set across assistants and break out results by category and purchase stage.
AskLift's multi-engine monitoring approach is useful here because visibility, citation and recommendation patterns need not move together.
What to do over the next month
First, audit GA4 traffic sources and landing pages. Second, identify the prompts your customers ask across ChatGPT, Claude, Gemini, Perplexity and any other relevant product. Third, compare which sources are cited and which brands are recommended. Fourth, improve your best pages with concrete answers and credible supporting evidence.
Don't build five near-identical versions of one guide for five assistants. Start with a useful canonical resource, check technical access and respond to actual gaps in the outputs.
Keep the claim accurate when sharing it
When you quote the 18.5% figure, put Goodie's publication date, time window and methodology beside it, so a reader can see exactly what the number describes.
AI search is changing quickly, so check for a comparable newer study before relying on it.
Check the figures against your own data
Put your own acquisition reports beside any published figure, and describe what the number covers whenever you reuse it.
The more useful headline is that dependence on one engine can hide changing behavior. A dashboard focused only on ChatGPT could miss leads arriving from Claude, Perplexity, Gemini or Google experiences. Yet measuring five sources does not require declaring an overall winner. Start with the platforms that matter to your buyers and expand the coverage as the evidence warrants.
Referral traffic is the visible tip
Many AI answers satisfy a question without sending a click. A prospect might see a brand repeatedly, search for it later, and convert through branded organic search or direct navigation. Standard referrer reports will not label that influence as an AI visit. Conversely, a click from an AI assistant is not automatically qualified or incremental. It could be a researcher, student, existing customer or prospective buyer. That is why visibility tests, session analytics and conversion data answer different questions. A mature report puts them side by side without pretending they are the same measurement.
Build an engine-by-engine decision table
For each assistant, track a fixed set of prompts, observable brand mentions, explicit recommendations, linked citations when available, referred sessions and assisted or direct conversions where measurement is possible. Add a note about the user experience: does the tool frequently link sources, offer shopping-style recommendations, or answer without outbound links? Do not blend those differences away. A platform that cites your research but rarely recommends your service may call for better product positioning. A platform that recommends the brand but sends almost no traffic may be generating awareness rather than direct-response visits.
Review monthly trends rather than chasing daily changes. As interfaces and default models evolve, source linking and recommendation formats can shift even when your website does not. Record major product updates and analytics tagging changes next to your trendlines. That context is often more valuable than a spurious decimal point in a market-share chart.
A realistic editorial response
Create useful pages for the problems people bring to each system, not separate pages that repeat “Claude” or “ChatGPT” for its own sake. A strong comparison article, pricing explanation, implementation guide or glossary can help across multiple search and answer experiences. Where a platform cites original research, make your methodology accessible. Where users ask buying questions, give them accurate fit criteria and limitations. Then measure whether your own audience actually responds. A third-party industry study should open a question for your business, not substitute for your business's answer.
Frequently asked questions
Is Claude the second most-used AI assistant?
No. The study measures identifiable B2B referral traffic in one brand panel.
Should I optimize separate content for Claude?
Usually start with one technically accessible, well-supported source page and measure differences before creating specialized content.
Can GA4 identify all Claude visitors?
No. Missing or stripped referrers can leave some visits unattributed.
Sources
Related reading
Mentioned vs. Recommended vs. Cited: The 3 AI Visibility Metrics That Matter
A mention tells you an AI assistant knows your brand exists. A recommendation tells you the assistant actively puts your brand forward for a particular need.
Read article
Is AI Traffic Hiding in Your Analytics? How to Measure Visits from ChatGPT, Claude and Perplexity
Some visits from AI assistants can be identified in GA4 by referral source or its AI Assistant channel grouping, but not every visit retains a usable referrer.
Read article
Why ChatGPT Recommends Your Competitor (and How to Change It)
ChatGPT may recommend a competitor because that company appears to fit the user's request better, has clearer public evidence, is discussed by credible outside sources, or is simply easier to identify and compare.
Read article