How B2B Buyers Actually Buy in 2026: The Evidence

The short version

What HASN'T changed

I've made the point before that the B2B buying process has gone dark. Let me correct myself: it's always been dark to an extent. What's happening now is that it's getting darker.

Not being able to fully see what B2B buyers are doing is nothing new. We may have envisaged a linear process and expected one click to lead to another that we could always track, but that was never the case - and we knew it was never the case. In my past lives, we either tried to track based on click paths, or we built models that made assumptions about how buyers were navigating between sources and stages. It was never an exact science, even if we pretended it was.

Despite the popularity of progressive funnel models, buyers went backwards and forwards, and sideways. In reality, it was a maze of different paths - depending on the context, the size of the buying committee, the complexity of the product or service, and so forth.

We've always had the adage that a buyer is 60%, 70%, or 80% through the journey before they talk to a salesperson - CEB was publishing 57% back in 2012.

None of this is new - it's been the case for some time, and we've known it.

So why am I saying the process has changed? Because there's now a machine inside. The signals from it are even harder to track. And the whole process is getting faster and more condensed.

The five stages

Here's the journey as the data describes it:

1. The Day One List - the field of vendors in the buyer's head before the purchase even starts, built over years from experience, peers, and reputation.
2. The Silent Research Phase - the scoping and gathering you never see: defining the problem, setting requirements, reading, asking around.
3. The Comparison Moment - AI takes the field and the research and compresses both into a working shortlist. List in, shortlist out.
4. The Trust Loop - the buyer checks whether what AI said is actually true: your website, reviews, peers.
5. The Validation Close - the buyer confirms it's safe to act with a human - the seller's new job.

The first three decide who's in contention. The last two decide who wins.

The Day One List

The field is set before the purchase starts. 85% of B2B buyers purchase from their "Day One List" - Bain's term for the vendors they have in mind before any research began - and 97% have prior experience with at least one vendor they're considering (6sense, 2025). Buying groups now average 11 people (6sense, 2025), and every one of them arrives with names already in mind.

The Silent Research Phase

Then the scoping starts, and you don't see any of it. 64% of business buyers are now Millennial or Gen Z (Forrester, 2025), and they self-serve by default: 67% prefer a rep-free experience, 70% prefer fully digital self-service (Gartner, 2026). By the time a vendor hears from the buyer, 61% of the journey is already complete (6sense, 2025). Everything gathered here - requirements, peer input, broad search - becomes the raw material they hand to AI next.

The Comparison Moment

The Comparison Moment is where AI takes the buyer's field and their research and compresses both into the working shortlist.

In the buying data, I see two scenarios.

The first is the buyer who goes to AI with a preconceived predetermined list - built from past experience, peer recommendations, research done elsewhere. They're not going in cold. This is one of the biggest fallacies in the commentary right now: the implication that AI is doing everything. This buyer has already done some of the filtering through what they are inputting into the AI tools. They're then having a conversation with AI, which is a very different mechanic to traditional search. They're using it almost as a sounding board - going back and forth, asking repeated questions. It's not one keyword in, one results page out. It's iterative. It develops, and they're using that interaction to form the decision: first the short list, then the prioritisation within it, and ultimately a preference. These buyers show up clearly in the research: 94% of buyers who use LLMs use them to summarise and organise research they've already gathered (6sense, 2025).

The second scenario is the buyer in a new market - no history, no context, no pre-established list. They go to AI cold, and AI gives them the first pass on who the players are, forming their understanding of the market before the shortlisting starts. The published research finds the same two modes and one more thing besides: most buyers toggle between both in a single session (Omniscient Digital/Wynter, 2025).

In both scenarios, AI is condensing the process - reducing the time and effort the buyer spends filtering, analysing, and building the shortlist and preference they decide on. The condensing shows up in hard numbers. Average buying cycles shortened from 11.3 to 10.1 months in a single year, and first vendor contact moved from 69% of the way through the journey to 61% (6sense, 2025). Buyers reach the conversation faster because AI has compressed everything before it. And it redraws the lists as it compresses them: seven in ten software buyers say AI surfaced information that changed which vendor they chose (G2, 2026).

Compression is the right way of viewing this. What would previously have been done by a human working through different information sources and supplier sites, using their own judgement to build a shortlist and a preference, is now literally being done in minutes - in a condensed iterative process between the buyer and the AI. List in, shortlist out.

The compression would matter less if the answer stayed stable as the buyer got more specific. It doesn't. When I benchmarked the UK business laptop market in July - twelve real buyer questions across five tracked AI engines, 140 brand mentions, 205 cited URLs - all five brands appeared when the question was broad: Lenovo, Dell, HP, Asus and Acer. Across the eight questions that named an attribute or a use case, Acer wasn't detected at all, another held complete coverage on the general questions and half of it on the specific ones, and a third disappeared from remote work entirely. The field the buyer sees at the category question is not the field they see at the buying question. And it's the second one the shortlist is built from.

That's not randomness - different questions pull from different sources, and a brand that's missing from the sources a question reaches for is missing from its answer. Acer didn't drop out of eight answers by bad luck.

The Trust Loop

The Trust Loop is the buyer asking one question: is what the AI just told me actually true?

Buyers use AI for filtering and structuring and to help them form a preference. But they don't take the answer on trust - they check it. 80% of buyers check AI's output against the vendor's website, 60% cross-reference third-party reviews, 55% ask peers - and only 22% act on what AI told them without checking anywhere (Omniscient Digital/Wynter, 2025). When AI names a vendor the buyer doesn't know, half go straight to that vendor's website to see whether it's real. Buyers consult an average of seven information sources per purchase - AI is never the only one (Gartner, 2026).

This is what most so-called AI visibility experts miss, because they focus on citations and stop there. Buyers coming out of an AI conversation are more informed, educated, and preconditioned than buyers who historically arrived through search. The conversion data backs the observation: AI-referred visitors convert at 4.4 times the rate of traditional organic traffic (Semrush, 2025) - and Adobe's 2026 retail data shows the same direction, with AI traffic now converting 42% better than non-AI channels, a complete reversal from a year earlier (Adobe, 2026). Which means it's imperative that your website and other touchpoints corroborate how AI is presenting, describing, and recommending you. Unfortunately, in my observations, that's usually not the case. I see different language, different ways of describing products and services, misalignment between offers.

Something else jumped out of that laptop market analysis, and it surprised me: where the answers were coming from. It was almost never the brands' own websites. Across the 205 cited URLs, the four reported brand domains accounted for just eleven citations between them - everything else came from third parties: tech media like TechRadar, RTINGS and PCMag, and YouTube. One third-party article was cited 25 times, on its own more than all four brand domains combined. Which sounds like the website no longer matters. It's the opposite. The vendor website is a weak input to AI's answer - but it's the decisive checkpoint after the answer, the first place four in five buyers go to check what AI told them. Two different jobs. Most vendors are optimising their site for a third job: ranking in search, which increasingly serves neither.

The peer element matters just as much. In my previous experience, especially at companies like Spiceworks, I learned how heavily B2B buyers rely on the perspectives and lived experiences of their peers. The trust data hasn't moved:  buyers trust peer recommendations (85%) and third-party reviews (78%) far more than anything a vendor publishes (Omniscient/Wynter, 2025). What they're looking for is perspectives that support and ratify what they learned through AI. So brands need to understand how they are being presented through AI, then build for the check that follows: experiences that detect buyers arriving from AI, mirror what the AI presented, and earn the peer endorsement that closes the loop.

The Validation Close

The Trust Loop checks the facts. The Validation Close is something different: the buyer confirming the decision itself. The question changes from "is this true?" to "am I safe to act on it?" - and that's the question buyers still take to humans.

That quote is absolutely correct - and look at what it's actually saying. The buyer used AI. They got an output. What they won't do is carry the risk of acting on it alone. The data on where trust sits is striking: 51% of buyers say they're more likely to get misleading information from Gen AI, and 49% say the same about a sales rep (Gartner, 2026). Buyers trust neither on its own - which is exactly why 69% turn to a sales rep specifically to validate AI-generated insights (Gartner, 2026), and why a quarter won't let it near the final call (Omniscient/Wynter, 2025). The buyer isn't asking the rep for information - they already have it. They're asking the rep to share the risk. If the decision goes wrong, "the AI recommended it" protects nobody; "the vendor confirmed it" does. And the rep helps carry the decision through the rest of the buying group - which is where Gartner finds reps still matter most.

There's a real danger at the moment that companies and brands dismiss the role AI is playing - especially amid the pushback around things like 'AI slop'. That would be a massive mistake, because the close is where the machine's work gets converted or wasted.

And the conversion odds are set earlier than most sales leaders think. The vendor a buying group favours before first contact wins 77% of the time (6sense, 2025). And seven in ten software buyers say AI surfaced information that changed which vendor they chose (G2, 2026). The favourite isn't crowned in the sales conversation. It's crowned during the comparison phase - the phase AI now runs.

For sales-led organisations, teams need to understand this process and be prepared for conversations that validate what the buyer has already learned and decided - rather than playing the traditional role of a persuader. Gartner's own long-range read points the same way: by 2030, they predict 75% of buyers will prefer buying experiences that prioritise human interaction - the pendulum swings back precisely at the close, where validation lives.

The Selection Layer

What's important now is understanding how we interject and influence each part of this process - whether that's interfacing with humans or interfacing with machines. That's the work I've been mapping as The Selection Layer, and the journey and the layers line up directly: getting into the answer, being framed right, and being ranked first (Inclusion, Positioning, Preference) is how you win the first three stages; an arrival experience that confirms what AI just said (Selection) is how you win the last two.

The tactical implications follow from the whole journey, not just the close. Take lead and demand generation: when to engage buyers, around what, and in what context all need rethinking when the comparison is happening inside an AI conversation you can't see. And I believe the role of brand is now more important than ever - entity recognition, credibility, and trust are primary signals for AI tools, so brand marketers need to understand what branding means in this new buying process, from both a human and a machine perspective, and how the two work together.

If I had to choose a place to start, it would be the interface between traffic coming from AI tools and that first engagement - on the vendor website or through the sales organisation. Make sure there's consistency in how things are described, how they're positioned, and the offers in place, so they align with what's being presented through the AI platforms.

Before you get there, you have to understand what's going on right now - especially in the Comparison Moment, which is now mediated by AI. That's where the AI Shortlist Scan gives you a starting point: whether you're present in the shortlists AI is mediating, how you're presented within them, how you're compared to your competitors, and how AI is deriving the preference that gets communicated to your potential buyers.

From there, the work is engineering the journey end-to-end - maximising what buyers bring to the front, ensuring you're represented and preferred within the AI-mediated comparison, and connecting that to what your teams, website and channels communicate so the journey continues. Ultimately, you build pipeline and you sell what you're promoting.

The simplest way I can put it is the way I'd say it to a friend who runs a UK B2B scale-up: Decoding Growth helps B2B brands get selected - not just mentioned - when AI tools such as ChatGPT, Gemini and Perplexity build the shortlists their buyers act on. The buying process has changed. The brands that understand how it now works - human, machine, and the interplay between them - are the ones that will be in the answer, on the shortlist, and across the line. Now is the time to look.

Next steps

Want to know where your brand stands? The AI Shortlist Scan is a decision-grade analysis of how AI is currently representing your brand across all four layers of The Selection Layer. £1,995, delivered in seven days, money-back guarantee.

Get The Selection Layer in your inbox - fortnightly thinking on how B2B brands get included, positioned and chosen in AI-mediated buying. Subscribe here.

A note on method

Every statistic in this piece is drawn from named, published research: Bain & Company (2025); 6sense's 2025 B2B Buyer Experience Report (c. 4,000 global buyers); Gartner's 2026 buyer surveys (645 B2B buyers, fielded Aug–Sept 2025); Forrester's Buyers' Journey Survey (2025) and State of Business Buying 2026; G2's The Answer Economy (April 2026; 1,076 B2B software buyers across North America, EMEA and APAC, surveyed March 2026); Omniscient Digital and Wynter's "From Prompt to Purchase" (Oct 2025, 100 B2B SaaS decision makers); and conversion data from SEMrush (2025) and Adobe (2026). G2 figures describe software buyers specifically and are labelled as such.

Some widely circulated numbers were deliberately left out. A frequently quoted claim that 90% of buyers click through AI citations to fact-check conflicts with observed-behaviour research (Pew, 2025), showing citation clicks near 1% across everyday search - the two measure different behaviours, and neither clearly supports the claims usually hung on them. "AI usage" statistics ranging from 7% to 89% measure different things; this piece anchors on usage during active vendor research and comparison. And where AI chatbots were briefly reported as the single biggest influence on B2B shortlists, later data from the same source (G2's 2026 Buyer Behavior Report, July 2026) shows review sites and chatbots effectively level - so no such ranking claim is made here.

The client observations are mine. The UK business laptop benchmark was run through Otterly.AI: twelve buyer prompts, five tracked AI engines, fourteen days to 31 July 2026. They are presented as observations, not as a study.