79% of global B2B buyers now use AI tools to research solutions, and 51% begin vendor research there. A survey this year found that 96% of B2B companies are effectively invisible in AI-driven discovery, appearing only in late-stage queries if at all.
And the figure that reframes the rest: AI search traffic converts at around 14.2%, against 2.8% for Google organic. A little over five times better.
Why this does not contradict what I wrote last month
Search economics
AI discovery economics
High volume, low intent
Very low volume, very high intent
Rewarded publishing coverage
Rewards being correct on a few key questions
Visitors arrive unqualified
Visitors arrive pre-shortlisted
Converts at roughly 2.8%
Converts at roughly 14.2%
I argued in August that AI assistants send almost no traffic and that chasing them was mostly wasted effort. The volume point stands. The conversion figure means the conclusion needs refining.
A channel that is one percent of the size but five times the conversion rate is not a channel to ignore. It is a channel to treat completely differently from search, because the economics are inverted: search was high volume and low intent, and this is the opposite.
The practical difference is where you spend. Search rewarded publishing volume, because you needed coverage to catch a diffuse stream of low-intent visitors. This rewards being unambiguously correct on a small number of high-stakes questions, because the few people arriving have already been told you are a candidate and are close to deciding.
Why the intent is so high
“A channel that is one percent of the size at five times the conversion rate is not one to ignore. It is one to treat completely differently from search.”
Because the filtering already happened. Somebody asking an assistant which tools handle a specific problem for a specific industry has described their situation in detail, received a short list, and arrived at you pre-qualified by a system that had no commercial reason to recommend you.
That last part is the whole mechanism, and it is why 85% of buyers reportedly view AI-cited vendors more favourably. A recommendation from something with no stake reads as closer to a colleague's referral than to an advert, and it carries the trust that comes with that.
A channel that is one percent of the size at five times the conversion rate is not one to ignore.
It is also why buyers still verify. 94% fact-check AI responses at least some of the time, and demos, trials and reviews remain the things that close. The assistant does not make the decision. It decides who is in the running, which has always been the hardest part of B2B marketing to influence.
Turn what you know into what you own.
Vibepreneur builds structured ventures from professional expertise, with positioning, launch assets, and growth systems included.
Join the WaitlistThe 96% is the opportunity
Almost every company in this position is absent from the stage of the journey where candidates get chosen, and present only at the end, when the buyer searches a name they already have.
That ordering is backwards and it is fixable. Being named at the shortlist stage requires existing, in clear language, as an answer to the question a buyer actually asks. Not your category. Their problem, in their words, with enough specificity that a system summarising the field has something concrete to attribute to you.
This is also where the citation rate matters. ChatGPT reportedly names brands in around 30% of answers but cites sources in only about 20%, which means being mentioned and being linked are different events with different causes. You want the mention. The link is a bonus.
The committee problem underneath
One further number worth planning around: B2B buying committees have grown from roughly five people to sixteen, and 74% report internal conflict during the process.
Sixteen people means your product is being described, second-hand and imperfectly, by somebody internal to somebody else internal. The material that wins is the material that survives being relayed: a one-sentence description, a concrete number, a named comparison. If your positioning requires a conversation to make sense, it will not make it through sixteen people, whatever the assistant said about you.
See where the AI search losses land for the volume argument, Google says there is no such thing as GEO for what is worth doing, and B2B buyers research with AI first for the behaviour shift. The marketer track has the rest.