60% of B2B Research Never Touches Your Website. Here's What to Do Next.

You're managing a buying committee with machines sitting on it.

Something has shifted in how B2B buyers actually do their research, and most of us felt it before we had the numbers to back it up. Zero-click search, getting your answer without ever clicking through to a site, has been creeping up for years. Forrester's latest Buyers' Journey Survey, nearly 18,000 business buyers worldwide, found 94% of them used AI somewhere in their most recent purchase, and when asked to rank what actually helped them decide, they put AI ahead of vendor websites, ahead of sales reps, ahead of product experts. Not close behind. Ahead.

We heard the same thing, completely unprompted, at our own AI panels this spring.

At our London event, one attendee put a number on it that stuck with us:

B2B buyers are already researching products inside LLMs at a rate in the high 80% range, and the "zero-click journey" isn't an edge case anymore, it's close to the default.

‍

The pain points are real, and they compound

Three problems are colliding at once. Search traffic is declining as AI Overviews intercept the query before a click happens, and most teams don't have a proven replacement channel. Last-click attribution - still the backbone of most B2B reporting - can't see influence that happens inside a chat interface with no click, no UTM, and no landing page visit. And until now, marketing leaders haven't had a direct way to reach buyers inside that environment at all.

This hits B2B harder than consumer categories, because B2B buying already involves multiple stakeholders, long research cycles, and dense, comparison-heavy content - exactly the conditions where a buying committee would rather ask an AI to synthesize ten datasheets than read them all individually. At our London panel, one marketing leader cited a Forrester figure that reframes the whole problem: the average B2B buying committee now runs about 16 people, nine internal and seven external - and several of those external touchpoints are AI tools and LLMs, not other humans.

‍

There's Now a Lever

OpenAI has opened ChatGPT to advertisers, and ProspectBase now offers managed access to it.

For the first time, there's a paid media surface sitting directly inside the exact moment your buyers are researching, comparing, and building their shortlist - not adjacent to it, not upstream of it, inside it.

Here's what that looks like in practice: when a buyer asks ChatGPT something like "best CRM for small businesses," a sponsored result can appear directly beneath the model's answer, clearly labeled Sponsored, with a short headline, a one-line description, and a link back to you. It's a small, contextual unit - closer in spirit to a well-targeted search or display ad than to a disruptive placement - but it sits somewhere no search campaign or display network has ever reached: inside the answer itself.

This is the moment marketing leaders have been waiting for since zero-click numbers started climbing with no lever to pull. Now there's a lever, and it's live today.

What This Is (and Isn't) Yet

Targeting right now is contextual, and is based on the content of the buyer's prompt - not audience or demographic targeting. Your ad shows up because the conversation is relevant to what you sell, not because of who's on the other end. Reporting is early too: you'll see impressions and clicks, not the deeper funnel diagnostics you're used to from paid search or LinkedIn.

None of that makes it less worth testing. It makes it something to test deliberately, rather than something to fund like a proven channel.

‍

How to Run Your First Test

Treat this as a structured experiment, not a channel migration. Pick one high-intent research moment your buyers already have - a comparison query, a "best X for Y" question, a category-defining prompt - and put a small, well-scoped test in front of it. Measure whether your brand shows up.

The bigger shift underneath the format is attribution itself. The meaningful metric is no longer last-click - it's whether your brand shows up at all inside an AI-generated answer. At our Silicon Valley panel, one marketing leader described attribution as an ongoing turf war among stakeholders, everyone claiming credit for the same result, no matter how much AI sits in the stack. Most B2B teams don't have a system for tracking visibility inside AI answers yet. Starting now, even informally, puts you ahead of competitors still reporting exclusively on website sessions.

‍

Getting started: The next 12 months

This is an early-stage environment, and it will change fast. Treat it as an experimentation budget line, not a bet-the-quarter reallocation. Run a small, well-scoped test this quarter, track brand visibility inside AI answers alongside your traditional metrics,and treat every learning as an input into a bigger 2027 media plan. The teamstesting now will have real performance data by the time this channel matures.The teams that wait will be reacting to a channel their competitors alreadyunderstand.

It's also worth watching what comes after this. At our London panel, Tiffany Jackson of Dell Technologies pointed to a shift already underway on the B2B side: agent-to-agent buying, where an AI agent acting on the buyer's behalf transacts directly with a vendor, no human clicking "buy" at all. Her view was that B2B is moving toward this faster than consumer categories, since business buyers are more comfortable handing that step to an agent than consumers are. Not every panelist shared her comfort level with it, but the direction was clear enough that it belongs on your radar as the next stage of the same shift you're testing for today.

‍

Get In Before the Channel Matures

The open web is no longer thedefault destination, and now there's a way to test what comes next. If you're weighing how this fits into a broader demand strategy, or want a second opinion on where to place your first test, get in touch with ProspectBase.

‍