B2B buyers entrust their preselection to two prescribers, the analyst and the chatbot, who judge on the same criteria. Most companies don’t fill them out.
According to The Answer Economy study published by G2 in April 2026 among 1,076 B2B software buyers, 51% of them now start their search in an AI chatbot rather than a search engine. They were 29% a year earlier. The same study places these assistants at the forefront of influences on the constitution of short lists, at 54%, ahead of review sites and ahead of the sites of the publishers themselves.
Reread that last point. What your own site says today carries less weight in the selection than what a machine says about you. And the machine is only the second prescriber in history. The first has been officiating for forty years.
Inquiry, the invisible channel that creates short lists
The general public knows analyst firms for their quadrants and waves. Practitioners know that the main thing is played out elsewhere, in a more discreet exercise: the inquiry, this half-hour interview where a subscribed customer asks for advice before choosing. Gartner claims more than 500,000 customer interactions per year, and an analyst covering a in-demand category handles several hundred alone.
We imagine the scene as a directory consultation. The reality is richer. The analyst does not come up with three names, he compares: what distinguishes the solutions from each other, what customers with a comparable profile have concretely obtained by deploying them, the reservations that accompany each option. This is the material for a recommendation.
This material comes from somewhere: briefings with publishers, their sites, their published customer cases. A company whose differentiator is not written anywhere, whose customer benefits lie dormant in a commercial deck, gives it nothing to redistribute. This is the case that we most often encounter in audits: the material exists, it has simply never been written. An analyst cannot make up your story for you. More often than not, he just doesn’t quote you.
The chatbot does the same job, in self-service
Ask the chatbot the question the client is asking their analyst, and look at the answer. Not a list of links: a comparative summary, with the options, their respective strengths and user feedback. Structurally, it’s an inquiry. Without appointment, without subscription, served dozens of times per day in each category.
The two prescribers do not read each other, the analyst reports being reserved for subscribers and out of reach of the engines. But they apply the same selection criteria. Neither of them measures the intrinsic quality of a product; they do not have the means. They recommend what they can understand and compare. A formulated differentiator and quantified benefits, linked to a named target, give them something to work on. A generic and interchangeable message gives them nothing.
An exclusion without witness or trace
No one warns those who fail. The analyst does not mention the publishers he did not mention, even less the AI. The deal never appears in the loser’s CRM, since for him it never existed.
The G2 study gives an idea of what is happening in this silence: 69% of buyers surveyed ultimately chose a supplier other than the one they had planned, based on the responses of a chatbot. One in three purchased from a publisher they had never heard of before the conversation. Established brands thus lose business that they will never know they played, in favor of lesser-known but better-known players.
The French deficit falls at the worst time
It remains to be seen how many companies pass this double test. It’s measurable, and we measured it. The Message-Market Fit Observatory, published in the second quarter of 2026, audited 369 French startups post-raise on their public data alone, that is to say what a prescriber sees before any first contact. The average message clarity score is 5.33 out of 10, and 75.9% of audited companies fall below the critical threshold.
Two results deserve attention. Raising funds does not improve the message: the correlation between amount raised and clarity is almost zero across the entire corpus. And the preparation of sites for reading by AI caps at 18% of its potential. Compare these figures with those of G2: when 69% of buyers change their minds based on a chatbot response and three out of four companies are unreadable for this chatbot, the calculation of what is lost is quickly done.
The cheapest variable of the growth plan
There is no such thing as an analyst relations project and an AI visibility project to be financed side by side. There is a clarity project, from which the other two benefit directly. What the analyst expects to recommend a company in an inquiry, namely a clear positioning, formulated differences, numerical and public proofs, is word for word what an AI expects to quote it in a response.
This work does not require product redesign or recruitment, much less an additional round of funding. In an environment where the two prescribers who count filter on the same criteria, the clarity of the message has become the cheapest growth variable to correct. It is also, all the data shows, the most neglected.
