95% of B2B Deals Go to a Vendor Already on the Shortlist. Falling Website Traffic Might Not Be Your Real Problem.

6sense’s new diagnostic makes a claim worth checking against your own funnel data before panicking about traffic: 95% of B2B deals are won by a vendor already on the buyer’s initial shortlist, and generative AI hasn’t shortened buying cycles, sped up first contact, or reduced the number of stakeholders required to align on a purchase. What’s actually changed, per 6sense, is where out-of-market buyers do their casual research — shifting from vendor websites toward LLM summaries — while active, in-market buying-group behavior stays largely intact. The real risk isn’t falling demand, it’s losing shortlist inclusion and brand visibility while a prospect is still out-of-market and invisible to pipeline reporting. The piece offers three diagnostic reads: traffic down but demand stable calls for account-level engagement tracking and LLM/analyst visibility investment; traffic and opportunity creation both down calls for more always-on brand investment; traffic stable but conversion weak calls for prioritizing buying-signal detection over lead capture.

That reframe lines up with Gong Labs’ buyer-side research from a few weeks earlier. Forrester’s 2026 survey of 18,000 global buyers found AI usage in purchase research climbed from 89% to 94% year over year, with generative AI now the single most-used research source — ahead of vendor websites and sales reps. But Gong’s read adds the stabilizing half of 6sense’s argument: 69% of buyers still consult a rep specifically to validate AI-generated findings, and buyers are 32 percentage points more likely to feel confident in a decision once a rep is involved, partly because 51% report encountering misleading information from AI during their own research.

Read together, the two data sets point the same direction: AI has moved where research happens, not who ultimately gets shortlisted or who validates the shortlist. Chasing traffic numbers that don’t move the shortlist is optimizing the wrong metric.