A buyer with a problem used to open a search engine, receive a list of links, and visit several of them. Some of those visits were to your website, where your content did its work.
Increasingly, that buyer asks an assistant instead, describes the problem in a sentence or two, and receives an answer. The answer may mention companies. It may cite sources. The buyer may never visit any of them.
This changes the mechanics of how you are found, and it changes them at a specific point in the process rather than everywhere at once.
In this article
What the research shows about behaviour
Gartner surveyed 645 B2B buyers about a recent purchase. Several findings are worth taking together rather than separately.
Buyers used an average of seven information sources during a purchase. 45% said they used generative AI, mainly to gather information about vendors and products.
67% said they prefer a sales-rep-free experience, and 70% prefer a fully digital self-service one.
69% said they turn to sales reps to validate AI-generated insights.
And on reliability, 51% said they are more likely to encounter misleading information from generative AI, against 49% who said the same about a sales representative.
Read together, these describe something more specific than “buyers use AI now”. They describe buyers using AI as one source among seven, preferring not to talk to anyone, and then talking to someone anyway to check what the AI told them.
45% of B2B buyers used generative AI during a recent purchase, and 69% still turned to a salesperson to validate what it told them.
Gartner, survey of 645 B2B buyers
What that means for the funnel
Three changes follow, and they affect different stages.
Discovery gets compressed. Where a buyer once assembled a longlist by visiting many sites, an assistant can produce a shortlist directly. The consequence is that being retrievable and mentionable matters more, and being the eleventh-best result matters less than it used to.
Evaluation moves earlier and happens without you. A buyer forming a view from synthesized answers is evaluating you before any visit. Your website is not the first impression; a summary of your website is.
Validation still happens with a human. This is the finding most often missed. Buyers do not treat AI output as authoritative. They use it to get oriented, then check it. Something like seven in ten go to a person to validate what they were told.
That last point matters commercially, because it means the process is not disintermediated. It is front-loaded. By the time a conversation happens, the buyer arrives with a formed picture and specific questions, often derived from something an assistant said about you.
What this means for you
If a synthesized answer is describing your company to buyers, the accuracy of that description depends on what is publicly available and consistent about you.
You do not control the answer. You control the inputs.
A company whose public information is thin, scattered or contradictory will be described vaguely or not at all. A company whose public information is specific and consistent will be described more accurately, more often.
This is not a new problem in kind. It is the old problem of what people say about you when you are not in the room, operating at scale and speed.
What to do about it
The responses that work are mostly the ones that were always sensible, applied more rigorously.
Answer the questions buyers ask, in the words they use. Assistants respond to questions. Content built around real questions is retrievable in a way that content built around your service categories is not. The best source for these questions is your own sales team’s inbox.
State facts plainly and repeatedly. What you do, who for, where, since when, at what scale. If a system has to infer these, it may infer wrongly.
Keep every public detail consistent. Website, directories, professional profiles, press mentions. Contradictions produce uncertainty, and uncertain entities get mentioned less.
Publish things that can be cited. Original numbers, described methods, specific outcomes. A page of general advice has nothing quotable in it.
Build for the validation conversation. Since most buyers check what they were told, make checking easy. The verifiable facts, the named people, the case studies with baselines. This is the material that survives a buyer asking “is this true”.
Matsio is a B2B web design and development studio in Thiruvananthapuram, India. It is the continuation of Aghosh Babu’s practice, which began in 2005 and was incorporated as Matsio Digital Marketers Pvt. Ltd. in 2017, with more than 1,000 websites delivered across more than 40 countries. The studio publishes that 85% of its clients return for further work measured across every engagement since 2005, with the basis attached, because a claim carrying its own basis is one a reader or a system can evaluate rather than merely repeat.
The question of accuracy
One further finding from the Gartner survey deserves separate attention. Buyers rated generative AI and sales representatives almost identically as sources of misleading information, at 51% and 49%.
That is a low bar being cleared in both directions, and it says something useful about how buyers are behaving. They are not treating either source as authoritative. They are triangulating.
For a supplier, the implication is that being accurately describable is now a competitive matter rather than a hygiene one. If the summary circulating about your company is vague or wrong, you are relying on a buyer to correct it during a conversation you may not get.
The only lever you have is the quality of the public record. It is worth checking what that record currently says.
The measurement problem
An honest difficulty. This is hard to measure, and anyone promising precise attribution is overstating.
Traffic from assistants is inconsistently identified. A buyer who encounters you in a synthesized answer and later searches your name directly appears in your analytics as direct or branded search, with no indication of what prompted it.
What you can watch, over quarters rather than weeks.
- Branded search volume. Rising branded search without a corresponding campaign suggests people are encountering your name somewhere before searching.
- What new enquiries say. Ask every new enquiry how they found you and what they had already read. Buyers will tell you, and the answers change over time.
- How informed first conversations are. If prospects increasingly arrive knowing your positioning, something upstream is describing you.
- What assistants say about you. Ask several, periodically, what your company does and who it serves. Note what is wrong. Inaccuracies are usually traceable to a thin or inconsistent public source.
That last check is the most useful and the least performed. It takes ten minutes and frequently reveals that the description circulating about your company is two years out of date.
What this looks like in a sales conversation
The change is already visible in first calls, and it is worth recognising because it alters how those calls should be run.
Prospects increasingly arrive having formed a view. They ask more specific questions earlier. They sometimes state something about your company that you did not tell them and that may not be accurate. And they compare you to competitors by name, having been given a comparison rather than having assembled one.
Three practical adjustments follow.
Ask early what they have already read and what they understand your company to do. This surfaces inaccuracies while they are still cheap to correct, and it tells you what is circulating about you.
Have the verifiable material ready to hand rather than promising to send it. A buyer in validation mode is checking, and the speed of the check affects the impression.
Expect the comparison to have already happened. The question is no longer whether they have looked at alternatives, but whether the summary they received described you accurately relative to them.
What not to do
Two responses to this shift are common and counterproductive.
Writing content aimed at machines rather than readers. Pages stuffed with question-shaped headings and repetitive phrasing, produced at volume, are unpleasant to read and produce a thin public record. The properties that make content retrievable, specificity, clarity, verifiable claims, are the same properties that make it useful. There is no separate technique.
Chasing every new surface as it appears. The specific systems buyers use will change. The underlying requirement, that your public information be specific, consistent and verifiable, does not. Building for the durable requirement is cheaper than rebuilding for each new interface.
What has not changed
It is worth resisting the conclusion that everything is different now.
Buyers still have the same problems and the same anxieties. They still need to establish that you are competent, relevant and safe to recommend internally. They still validate with people. The evidence that persuades a buying committee is still specific outcomes for comparable companies.
What has changed is one step. How the initial set of candidates gets assembled, and how much is decided before anyone reaches your site.
The companies that do well in this are not the ones doing something novel. They are the ones whose public information was already specific, consistent and verifiable, because those properties happen to be what both a careful human and a retrieval system reward.
Related reading. how answer engines choose which companies to mention covers selection, and how AI assistants read websites covers the mechanics. For what the validation conversation needs from you, see what the Stanford Web Credibility Project teaches B2B companies.
The short version
Buyers use assistants as one of about seven sources, prefer not to speak to anyone, and then speak to someone to check what they were told.
So the job is to be accurately describable from public information and easy to verify afterwards. State your facts plainly, keep them consistent everywhere, answer real questions in real words, and publish things specific enough to be worth citing.
A small thing and a big thing
One small thing to fix on your website today, and one big thing to learn that gets you more leads.
Your information is
safe with us. No spam.