B2B buyers trust AI search more than your sales team

Ninety four percent is not a figure that leaves much room for a "wait and see" strategy. Per Ahrefs, Forrester's data shows that share of business buyers now using AI search during purchase decisions, and rating it above vendor websites, product specialists, and the sales reps enterprises spend fortunes training. NielsenIQ's consumer number, 42%, looks almost reassuring by comparison. It is not. It is the early stage of the same curve.

For three decades, B2B marketing has operated on a durable assumption: that the sales rep, backed by a polished microsite and a few well-placed case studies, is the last mile of trust before a contract gets signed. Forrester's number says that last mile has moved. It now runs through a model synthesising answers from sources the buyer never sees named and the vendor never gets to argue with. A procurement officer at a bank evaluating risk software, a UN agency comparing logistics vendors, an industrial group shortlisting EPC contractors: increasingly, their first serious research step is a prompt, not a call.

The mechanism nobody budgeted for

Here is where Ahrefs' own framing, that AI search ROI is "messy," undersells the problem. The mess is not measurement noise. It is structural. A sales rep can be coached, corrected, and held to a script. An AI answer engine cannot. It draws on whatever combination of third-party analyst reports, review sites, forum threads, and indexed documentation the model's retrieval layer deems trustworthy at query time, and it updates that judgment on its own schedule, not the brand's. A CMO who spent years building rep enablement decks has, in effect, built assets for a channel that buyers are quietly deprioritising.

This matters most for sectors where the sales relationship has traditionally done the heaviest lifting: financial services, where relationship managers have long substituted for comparison shopping; multilaterals and policy institutions, where credibility has been a function of who you know inside the Secretariat or the Bank; industrial groups, where technical sales engineers are the primary conduit for specification detail. All three now face buyers who form a provisional view before any human conversation starts, built from whatever the model has already decided to cite.

The uncomfortable part is that this provisional view is sticky. Buyers anchor on first impressions. If a model's answer to "best enterprise risk management platforms" or "leading providers of disaster risk financing" names three vendors and omits a fourth, the fourth is not merely behind in the pitch; it is fighting an anchor that formed before anyone picked up the phone. Sales reps have always had to overcome price objections and competitor comparisons. Now they have to overcome an AI-generated shortlist the buyer formed independently, often without realising how incomplete or stale that shortlist might be.

What actually earns the citation

The useful implication is not "invest in AI search," which is vacuous advice dressed as strategy. It is that the inputs to AI search are not the inputs to traditional SEO, and B2B brands that treat them as interchangeable will keep losing citations to competitors who do not. Models weight structured, well-attributed, independently corroborated content: analyst write-ups, documentation with clear authorship, third-party comparisons, Wikipedia-adjacent reference material. A glossy microsite optimised for a human scrolling past a sales rep's shoulder is not optimised for a retrieval system scoring source credibility. The two disciplines look similar and are not.

For a financial institution, that means the analyst briefings and regulatory submissions sitting in a compliance folder are more valuable to AI visibility than another white paper behind a gate. For a multilateral, it means the technical reports already published for transparency reasons are doing silent work in model training and retrieval, whether anyone intended that or not. For an industrial group, it means specification sheets and standards-body citations outrank the glossy capabilities deck.

Forrester's 94% is a verdict on where trust has already relocated. The institutions that keep funding the old channel at the expense of the new one are not being cautious. They are being overtaken, one uncited prompt at a time.