The First Two Slots: How AI Answers Compress the B2B Shortlist

Search results pages were generous. Ten organic positions, ads above and below, a second page for the persistent — a vendor ranked seventh still existed, still collected clicks, still entered evaluations late and sometimes won them. The answer layer has no seventh position. Asked to recommend an enterprise tool, a model typically names one or two vendors and moves on (Chris Donnelly, Searchable). Everyone else is not ranked lower; they are absent.

This is the fourth report in our research series. The earlier reports mapped how engines diverge, what feeds their answers and what happens when those answers are wrong. This one measures the doorway itself: how many brands fit through it, and how early it closes.

Methodology

A synthesis, openly sourced

Like the rest of the series, this is a research synthesis, not a Brantial panel measurement. Every figure comes from published third-party research and 2026 conference presentations, cited inline; interpretation is labelled as such. The Turkish B2B shortlist measurement — which vendors Turkish-language prompts actually surface, sector by sector — is planned as a follow-up on Brantial’s prompt panel.

The frame

Four questions, in the order a deal experiences them: when does the shortlist form, how many slots does it hold, how sticky is it once formed, and what earns a place on it.

Finding 1: The shortlist forms before the first sales conversation

The numbers stack from every direction. 94% of B2B buyers consult an LLM before speaking to a sales representative (Paolo Provincali, LinkedIn — Zero Click NY 2026). Half of B2B buying journeys now open in chat, and among software buyers specifically, 51% start with ChatGPT or Claude (Tim Sanders, G2; G2 data via Andy Hoare). At Gelato, Henrik Müller-Hansen puts the structural version plainly: around 80% of the decision is made before any vendor contact, in a buying population where 71% of decision-makers were born after 1990 and treat conversational research as the default.

The evaluation that remains is shorter, too — cycles have compressed by roughly 40% since AI entered the journey (LinkedIn). By the time a rep hears about the deal, the question is rarely “who should we consider?” It is “which of these two should we pick?”

What this means for a vendor: pipeline instruments measure the wrong end of the funnel. The decisive contest happens in a conversation your CRM never sees, weeks before the first form fill.

Finding 2: One or two slots — and no metric for the excluded

The compression itself is the story. Where a results page distributed attention across ten positions, an answer concentrates it on one or two names (Donnelly). Provincali’s phrasing at Zero Click NY 2026 identifies the accounting problem this creates: there is no metric for “never evaluated.” A vendor excluded from the answer generates no impression, no bounce, no lost-deal record — nothing to alarm a dashboard. The deal simply happens elsewhere.

Donnelly’s blunt version: if you are third, you are not evaluated at all. The shortlist is not a ranking with a long tail; it is a binary with a rounding error.

What this means for a vendor: the first measurement worth having is presence-in-answer across the prompts that describe your category — a number most B2B teams have never seen for their own brand, because nothing in the traditional stack produces it.

Finding 3: The day-one list is sticky — but not closed to unknowns

Two published figures hold the tension of this finding. First: eight out of ten buyers ultimately purchase from their day-one shortlist (Tim Sanders, G2 — Zero Click NY 2026). The list formed in that opening conversation survives the committee, the demos and the procurement review. Combined with Finding 1, this is the whole argument for treating answer presence as pipeline work rather than brand work.

Second, the counterweight: roughly a third of buyers purchase from a vendor they had never heard of before the process began (Kristin Fracchia, Gamma — Zero Click NY 2026). The slots are sticky, not hereditary. A model assembling a shortlist from retrieval has no loyalty to incumbents — it recommends whatever its sources support today, which is precisely how unknown vendors appear on day-one lists they could never have reached through brand awareness alone.

And this is not a committee-free process being described: enterprise purchases still involve 11 to 20 stakeholders and around 27 touchpoints before sales contact (Searchable customer data). The AI answer does not replace the committee; it hands the committee its starting file.

What this means for a vendor: incumbency is worth less and eligibility is worth more than at any point in the search era. The same mechanism that makes the shortlist hard to crack for laggards makes it unusually open to prepared challengers.

Finding 4: The slots are becoming contested inventory

The scarcity has been noticed. ChatGPT now runs self-serve advertising, and the defensive pattern arrived immediately: Salesforce shows paid placements on roughly 40% of the prompts where it already appears organically (Jasman Singh, Profound — Zero Click NY 2026) — paying to hold a slot it already holds, so a rival cannot buy the adjacency. Early CPMs run around four times Meta’s, and about 30% of users already see ads in answers.

Read together with the organic data, the economics are straightforward: as paid pressure on the answer layer rises, the value of organic slot ownership rises with it. The published operational note cuts the same way — answer-layer work shows measurable movement within the first 60 days, against the 6-to-9-month horizon of classical SEO (Donnelly). Early positions are cheaper than defended ones.

What this means in Türkiye

Three implications transfer, each measurable:

Turkish B2B categories are largely unclaimed. In most Turkish-language vendor prompts, no domestic brand has done deliberate answer-layer work — meaning day-one shortlists are being assembled from thin, often outdated sources. The first mover in a category does not join a contest; it defines the default answer the next 8-of-10 buyers will act on.

The export shortlist is a separate battle. A Turkish vendor selling abroad is shortlisted through English-language prompts and English-language sources — a corpus its local marketing never touches. Presence must be measured in both languages, because Finding 2’s “no metric for never evaluated” applies twice.

Slot scarcity concentrates the source work. With one or two slots per answer and a thinner Turkish source ecosystem, the handful of review platforms, comparison sites and community threads that feed a category’s answers carry disproportionate slot-weight — a tractable list to audit, and a short one.

The takeaway

The B2B funnel has acquired a gate it never had: a one-or-two-name answer, formed before sales knows the buyer exists, sticky enough that eight in ten deals end where it pointed. Nothing in the traditional measurement stack reports whether you are inside it.

The sequence follows from the findings: enumerate the prompts that describe your category, measure presence per engine, audit the sources feeding the current winners, and do the work while the slots in your category are still cheap. The brands that treat the answer as the funnel’s front door will be the ones the committee is comparing — and the rest will never learn which deals they weren’t in.

Sources

• Paolo Provincali, LinkedIn — B2B buying behaviour and the “never evaluated” problem, Zero Click NY 2026
• Tim Sanders, G2 — chat-first journeys and day-one shortlist stickiness, Zero Click NY 2026
• Kristin Fracchia, Gamma — unknown-vendor purchases and organisational response, Zero Click NY 2026
• Jasman Singh, Profound — answer-engine advertising and defensive placement, Zero Click NY 2026
• Chris Donnelly, Searchable — shortlist compression and answer-layer operations, 2026
• Andy Hoare — B2B adoption data (Forrester and G2 figures), 2026
• Henrik Müller-Hansen, Gelato — pre-contact decision share and buyer demographics, 2026
• Brantial research series (2026): Divergence, Source Ecosystem, Brand Misinformation

Figures are as presented in the cited talks and publications and primarily describe English-language, US-market behaviour. Brantial’s follow-up study will measure the Turkish B2B shortlist — sector by sector, engine by engine — on our own prompt panel.