More B2B buyers now build their vendor shortlist in an AI chat, buy from it most of the time, and rarely click through to the brands it names. That was the clearest message from Profound's Zero Click conferences this year. Our own research points the same way.
Profound sells AI visibility software, and this year it ran Zero Click in Singapore, Sydney, London, New York and San Francisco. We went through all 52 sessions, more than 14 hours of talks, in which its researchers shared the stage with marketing leaders from Nestlé, LinkedIn, Stripe, G2, Estée Lauder, Bank of Ireland and Virgin Australia. A vendor's own event comes with a vendor's view, and several of the figures below are Profound's. Three findings held up against what we see in our own data.
The statistic that stayed with us came from Jennifer Zou, a researcher at Profound, presenting its study of how people shop with AI:
"We're aware that hardly anyone clicks on the links in ChatGPT conversations. So if you're looking for that UTM source equals ChatGPT tag, you're looking in the wrong place. Our data shows that people are 9 times more likely to actually navigate on their own to a named or cited brand's website… Google is precisely where people are going immediately after a shopping conversation, an entire order of magnitude more than any other domain."
A buyer asks ChatGPT for a recommendation, sees your name, closes the chat and searches for you on Google. Your reports credit the visit to branded search or direct traffic. The assistant that did the persuading gets nothing. The influence is still there, filed under the wrong source.
Profound's figure comes from a panel of consumer shoppers, which means the B2B numbers will differ, although speakers from software and financial services described the same pattern in their own data. Bank of Ireland's organic traffic fell 35% in the first quarter of 2026 while conversion on the same product rose 40%. LinkedIn sees visitors who arrive from AI convert at almost twice the rate of its other traffic.
Few marketing teams can see this yet. In the AI Visibility Index survey run by Agentcy, the platform we built, only 35% of 104 senior B2B marketing leaders formally tracked AI referrals, and 10% could consistently connect AI to revenue. Almost half believed AI already influenced between 6% and 20% of their pipeline.
Paolo Provinciali, who leads growth performance at LinkedIn, put the second finding most directly:
"When the buyer decides to buy and go to market, our research shows that they go to market already with a shortlist in mind. And 90% of the time they end up buying from that shortlist. Guess who gave them the shortlist? … If your brand is not part of what AI says, you're not even part of the competition. And there is no metric for never considered, but that's something that everyone in your company, your CEO, your CMO, your CFO should really care about."
G2 gave the same picture from its buyer surveys. Eight in ten software buyers chose their vendor from their day-one shortlist, and more than half now build that list in an AI chat. Tim Sanders of G2 described work that used to take a buyer 7 to 12 hours of searching being done "one shot with a single prompt".
The prize is a place on the list. A typical AI answer names three or four brands, and Profound found that only about one answer in six picks a single winner.
Our data shows where those lists come from. When a buyer's question names a company, the assistants take about 29% of their sources from that company's own site, and when the question asks which product suits a need, the share falls to about 6%, so the category questions that form shortlists are answered mostly from other people's pages. That comes from more than 1.3 million citations we analysed for how to rank in ChatGPT.
We have seen it from the inside, too. In our UK B2B Tech PR AI Visibility Index, ChatGPT cited our comparison page in dozens of answers each week, used it as a guide to the market and shortlisted other agencies from it. Being cited and being recommended are separate results.
The third finding is the one PR teams should act on first. Tim Sanders again, citing Profound's research:
"Not all earned media is AI visible… Less than 3% of all earned media citations are going to tier 1 publications like AP, Bloomberg, Wall Street Journal. Why? Because they're crazy, and they block AI, and the blocks are working. It's the tier 2 publications that get 97% of the citations. So when a PR firm wants to charge you $100 grand to get you a Bloomberg piece, perhaps if you're an analytics firm, getting VentureBeat or CIO Dive has a multiple more value."
Our own analysis reached the same conclusion from a different dataset. On buying questions about 116 B2B technology brands, three in four media citations went to trade and specialist titles, against about one in eight for the national press. TechTarget, CIO, ITPro and Computer Weekly came up most often, and the full list is in which B2B tech trade publications AI cites most.
National coverage still earns its place. Trade titles pick up the big stories, and those stories shape how a category gets described. Prestige alone no longer justifies the budget, though, when you can see which outlets the assistants actually read.
All three findings point the same way. Buyers form their preferences inside AI conversations that brands can't see. The shortlist they act on is one a brand either makes or misses, and the assistants build it largely from what other people say. A speaker from the Australian health insurer ahm put it in one line: "If no one else says it, AI might not believe you."
That puts earned media at the centre of AI visibility. Too much of the conference treated PR as a way to get a link into an article an assistant already cites. Profound found that a typical page's share of citations halves within 11 days, which would argue for constant placements, while our data shows trade coverage lasting much longer, with the typical trade article about nine months old by the time an assistant cites it. Both can be true. News fades quickly, while the trade features and buyer's guides that explain a category keep being read.
For a B2B tech brand, the work is a steady programme of specific, credible coverage in the trade titles the assistants read, with spokespeople journalists come back to and an account of the company that other people repeat. Measure it by whether you appear on the shortlist for the questions your buyers ask, which is a different number from share of voice.
To see which pages the assistants read and recommend from in your own category, ask us for a free AI visibility snapshot.