The Quiet Channel: Why Claude Keeps Showing Up First in B2B Recommendations

Two numbers that shouldn't coexist, but do

Previsible's July analysis of 6.77 million LLM-driven sessions across 166 websites found ChatGPT commanding 92.4% of standalone AI referral traffic, a consolidation story, with the platform still gaining share. If referral sessions were the whole picture, optimizing for anything else would be a rounding-error hobby.

Now the recommendation picture from our own monitoring. Across the 180-prompt panels we run for B2B customers, real buying questions, three providers, live product surfaces, every month:

• econ solutions (industrial energy-management software): Claude drives 56% of AI recommendations, ChatGPT 22%, Google AI Overviews 22%.
• BFE Institut (energy-efficiency consulting): Claude drives 74%, Google AI Overviews 23%, ChatGPT 3%.

In both fields, the platform with a single-digit-to-low-double-digit share of referral headlines is the dominant source of active recommendations for the brand. That's not noise, it's stable across months and repeats across our fleet's B2B deployments.

The industry data quietly agrees

The referral studies themselves, read closely, point the same direction for B2B specifically:

• Goodie's Wave-2 longitudinal study of B2B brand sites found Claude at 18.5% of measurable B2B AI referrals by March-April 2026, with ChatGPT down to 62.6% on that panel, from near-90% a year earlier. Their conclusion: optimizing only for ChatGPT now covers a third less of the AI landscape than it did twelve months ago.
• Previsible's own data shows Claude grew 64× over the tracked period and overtook Perplexity in March 2026, with its strength concentrated in professional and technical audiences.
• Multiple analytics write-ups this spring note Claude referrals as low-volume but highest-value: professional users, long sessions, strong conversion behavior.

So the reconciliation is straightforward. ChatGPT owns consumer-scale volume. Claude is disproportionately where professional and technical work happens, engineering teams, consultants, analysts, the people inside a B2B buying committee running vendor comparisons during their workday. Referral traffic measures where the crowd is. Recommendation share in a B2B field measures where your buyers' questions are being answered. Those are different markets, and the gap between them is largest exactly in categories like industrial software and specialist consulting, the fields our fleet serves.

Three implications if you sell B2B

1. Provider mix is a per-field fact, not an industry constant. econ's 56/22/22 and BFE's 74/23/3 differ meaningfully despite adjacent industries and a shared corporate parent. Any strategy built on someone else's aggregate, including ours, is guesswork; you need your own split. This is why our monthly monitoring reports recommendations per provider, not as a blended score.

2. "Optimize for ChatGPT" is an incomplete brief. Not a wrong one, ChatGPT matters everywhere, and it's 22% of econ's recommendations. But a B2B brand that tunes exclusively for ChatGPT while Claude drives three-quarters of recommendations in its field is optimizing the minority channel. The good news: the fundamentals converge. Clean machine readability, dense verifiable facts, third-party corroboration, every provider's retrieval rewards them. The differences show up at the margins (crawler behavior, fetch budgets, source preferences), which is precisely what monitoring exists to catch.

3. Watch Claude's search infrastructure. Cloudflare Radar's crawler data this summer shows Anthropic's dedicated search crawler (Claude-SearchBot) growing to the largest dedicated AI-search crawler on the network, the crawling that grounds live answers and produces citations, as distinct from training crawls. A provider investing that heavily in retrieval infrastructure is a provider whose answers will be increasingly grounded in what it can fetch cleanly right now. Being the cheapest, clearest fetch in your category compounds there first.

What we watch next

Two honest caveats. First, our fleet skews DACH B2B, industrial, energy, specialist consulting; consumer categories will look different, and ChatGPT's recommendation share likely dominates there the way its traffic share suggests. Second, provider mixes move: BFE's ChatGPT share at 3% is strikingly low and worth watching rather than extrapolating.

But the core observation has held for months across every B2B deployment we monitor: the loudest platform and the most influential platform for your category are not necessarily the same platform. The only way to know yours is to measure, which is, not coincidentally, the "+1" that sits above everything else in our GEO framework.

We were in the room with Anthropic's technical team in Berlin last month talking about exactly this layer of the stack, notes here. And if you want your own provider split instead of the industry's averages: the free AI Visibility Audit breaks recommendations down by ChatGPT, Claude and Google AI Overviews from day one.

Sources: FAIND fleet monitoring (econ solutions, observed Jan-Apr 2026; BFE Institut, observed Feb-Jun 2026); Previsible AI Traffic Report, July 2026 (6.77M sessions, 166 sites); Goodie AI Search Traffic Report Wave 2, May 2026; Cloudflare Radar crawler data, Q2 2026.