Blog
B2B suppliers and AI: the shortlist is made before the first contact
In 95% of cases B2B buyers choose a vendor that was already on their Day One shortlist (6sense, 2025). What it means for a manufacturer when that list is put together by an AI, and how to measure it.
In 95% of cases B2B buyers end up choosing a vendor that was already on their Day One shortlist. The figure comes from 6sense's 2025 Buyer Experience Report (November 2025, nearly 4,000 B2B buyers in North America, Europe and Asia-Pacific). The same study finds that first contact with a seller happens when 61% of the buying journey is already done, 58% in Europe, and that in over 80% of cases it is the buyer who reaches out, not the other way round.
Who makes the list is therefore decided before the company knows it is being evaluated. And a growing part of that early research goes through an AI assistant.
How the list is formed today
For years the shortlist came from a Google search, word of mouth, a trade fair. Today a question typed into ChatGPT, Perplexity, Gemini or Copilot joins these channels: "who are the main filter press manufacturers in Italy?", "who makes die-cast aluminium components for automotive?", "which Valpolicella wineries export to the United States?".
According to G2's Buyer Behavior Report (May 2025), 79% of software buyers say AI search has changed how they research vendors. For industrial goods there are fewer public figures, but the mechanism is the same: the generative answer suggests three to eight names, with a line of reasoning for each. A company that does not appear is not ruled out: it is never considered.
Where the AI gets the names
For category questions the models do not start from the company's website. They start from sources that describe the market:
- Lists and comparisons from trade press, industry associations, directories and B2B portals.
- News: an acquisition, a new plant, an award, a trade fair with figures.
- Structured profiles: Wikidata, Wikipedia where it exists, Google Business Profile, LinkedIn.
- The company website, but only if it says explicitly what it makes, for whom and where, with verifiable facts: product ranges, certifications, markets served, production capacity.
A company with an excellent product and a website that speaks in images and generic claims may not be recognised as an answer to a precise question. It is the most frequent case among mid-sized manufacturers: strong in their market, almost invisible in the sources the AI reads.
What is worth measuring
The first step is knowing whether the company appears, for which questions and next to whom. Ten questions written the way a buyer would write them, by product, application and region, asked to ChatGPT with web search and to Perplexity, are enough. For each answer three things are noted: whether the company is named, in what position, which sources are cited. The method is described in the DIY test.
If the company does not appear, the cited sources show where to act: often they are few, and half of them are outside the company's direct control but within reach. The baseline interventions cover structured profiles, website data and directories; the rest is content and relationship work over time.
Why it matters to sales
The practical consequence concerns the sales team more than marketing. If first contact comes after more than half of the journey, and almost always at the buyer's initiative, the share of qualified enquiries depends on how many shortlists the company reaches without knowing it. A supplier suggested by an AI is contacted by people who have already compared it with others, with precise questions and shorter timelines.
This is why for manufacturers the GEO Assessment starts from buyers' questions, not from questions about the brand.