TL;DR
The through-line this cycle is proof over promise: buyers now use AI everywhere in their research, but they trust neither the AI nor the vendors, and they're rewarding whoever can verify claims independently. Chatbots have quietly become the #1 shortlist gate, incumbents hold a 5-to-1 trust edge, and the federal government is finalizing a hard compliance floor for LLM procurement that will shape enterprise contracts more broadly.
The buyer's split brain: AI everywhere, trust nowhere
Enterprise buyers have adopted AI as their primary research instrument while structurally distrusting what it tells them — and sellers are being repositioned as the verification layer, not the information source.
"B2B buyers are more comfortable using digital channels and GenAI to navigate the purchase process on their own, but that does not eliminate the role of the seller… Buyers still turn to sales reps to validate AI-generated insights." — Gartner Survey Finds 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights
Gartner's 645-buyer survey found buyers averaging seven information sources per purchase, with 45% using GenAI — yet 69% turn to reps to check AI output, and perception of misleading information is at near-parity between GenAI (51%) and sales reps (49%) [buyers-use-ai-dont-trust-it]. The distrust of AI output itself is corroborated by G2, which found 64% of buyers frequently encounter inaccurate chatbot recommendations [buyers-use-ai-dont-trust-it].
This matters because the rep's job has inverted: the win condition is no longer being the first source a buyer hears from, but being the one whose claims survive the buyer's verification gauntlet. Vendors whose messaging can't withstand an AI cross-check plus a review-site sweep are filtered out before they ever know they were in a deal.
What to watch: Whether "arming reps as validators" becomes a formal sales-motion category, and whether the misleading-info parity (51% AI / 49% reps) shifts as buyers gain experience.
The answer economy: chatbots assemble the shortlist before you're in the deal
AI chatbots have crossed from novelty to the single most influential gatekeeper in B2B software shortlisting — and review citations are becoming the trust signal inside the AI answer.
"...Buyers trust AI chatbot recommendations enough to change their purchasing decisions, but citations from software review sites are the signal that makes them confident enough to act." — New G2 Research: Half of B2B Software Buyers Now Start Their Research With AI Chatbots
G2's 1,076-buyer survey found 51% now start research with a chatbot more often than Google — up from 29% seven months earlier — and 69% of buyers changed vendors based on chatbot guidance [review-platforms-ai-citation-substrate]. Review-site citations were the most confidence-inspiring element of an AI answer for 45% of buyers, and review platforms are the only source besides chatbots that gains influence deeper in the funnel (40% at discovery → 47% at retention) [review-platforms-ai-citation-substrate].
The strategic consequence: being legible to the models and cited with third-party validation is now top-of-funnel infrastructure, not marketing hygiene. Notably, 33% of buyers purchased from a vendor they'd never heard of — the best opening AI-natives have ever had against incumbent trust advantages, if the citation layer backs them [review-platforms-ai-citation-substrate].
What to watch: The next edition of G2's annual series (expected ~Q1–Q2 2027) — whether chatbot-first research share keeps climbing and whether citation confidence hardens into a de facto ranking requirement [watch-g2-answer-economy-ai-search-annual-report-update].
The trust collapse: buyers doubt vendors, not the technology — and incumbents win 5-to-1
The skepticism in enterprise AI buying isn't about AI capability; it's about vendor honesty, and it's driving consolidation toward established platforms.
"The buyers that we spoke to for this research, they are not skeptical of the technology. They're skeptical of the vendors." — The Trust Collapse: What Enterprise Buyers Think About AI (INFUSE webcast)
INFUSE's mid-year survey of 310 enterprise buyers found 73% optimistic about AI at work, yet 61% evaluating, planning, or already cutting platforms because of AI — with established vendors preferred 5-to-1 [enterprise-ai-trust-collapse]. What buyers actually evaluate: 47% want proof of integration, 32% want evidence of a similar client implementation, 23% explicitly reward transparency about what the AI does and doesn't do — and only ~4% cite third-party validation as decisive [enterprise-ai-trust-collapse].
The efficiency-over-transformation split is the sharpest edge here: 62% buy for operational efficiency while just 12% buy for innovation or competitive positioning — meaning, as INFUSE puts it, most vendor messaging "is organized around the 12%" [enterprise-ai-trust-collapse]. Last cycle's ROI-chasm framing (deep deployers seeing 6x the returns of pilot-stuck buyers) extends cleanly: the winning pitch is deployment depth and change management, not transformation narratives [zip-state-of-ai-in-spend].
What to watch: Whether the 61% actively cutting or evaluating vendors translates into early renewal renegotiations becoming standard practice across enterprise software categories.
Washington writes the compliance floor
The GSA's proposed LLM safeguarding clause moved through its comment period and is now heading toward a final rule — and its architecture will likely bleed into commercial procurement.
The comment window for GSAR 552.239-7001 closed August 3, 2026 with 77 comments; the clause would ban training on government data, require automated "eyes-off" handling that keeps humans out of data paths, mandate 72-hour incident notification, and flow obligations down to Developers, Operators, Integrators, and Service Providers regardless of contract position [gsa-american-ai-clause-gsar-552-239-7001]. One structural gap stands out: the clause governs data but says nothing about agent authority — what an autonomous system may do with tools once it has the data [gsa-american-ai-clause-gsar-552-239-7001].
For founders, the "eyes-off" requirement is a product requirement, not a policy one — it favors vendors who already run zero-retention, no-human-review infrastructure [gsa-american-ai-clause-gsar-552-239-7001].
What to watch: The final rule's treatment of the contested "unbiased AI" provisions, expected around October 2026.
What surprised us
- IT is the biggest shadow-AI offender. 57% of respondents have used unapproved AI tools — but among IT respondents it's 64%, the highest of any function [zip-state-of-ai-in-spend]. The enforcers are the violators.
- Scale is an ROI handicap. Organizations with 50,000+ employees are the most likely to deploy AI widely (33%) and the least likely to see clear ROI (11%) [zip-state-of-ai-in-spend]. Big budgets don't buy depth.
- "Never heard of them" is no longer disqualifying. A third of buyers bought from an unfamiliar vendor on chatbot guidance [review-platforms-ai-citation-substrate] — a genuine crack in the 5-to-1 incumbent trust advantage [enterprise-ai-trust-collapse].
- The buyer distrust is symmetric. Buyers see misleading information from GenAI (51%) and sales reps (49%) at nearly identical rates [buyers-use-ai-dont-trust-it]. Sellers aren't losing to AI; they're being held to AI's standard.
Open threads worth a vote
- [watch] G2 Answer Economy / AI Search annual report update — steer which stats we track when the next edition lands.