SandsDX

Research

The AI Shortlist Report

AI assistants now hand buyers a shortlist before a vendor gets a word in. The AI Shortlist Report measures those shortlists. Each wave asks every frontier model the same buyer questions, word for word, runs each shortlist question five times per model, and publishes the method with the findings. Corrections are printed in the log below.

  1. Wave 3,

    The AI Shortlist Report, Wave 3: The AI Gives Different Answers to a 50-Person Company, a Buyer With a Problem, and a Customer at Renewal

    Telling the AI the buyer's company size changes its rankings four times more than rewording the question, and at renewal it recommends renewing outright at most 5 times in 35.

  2. Wave 2,

    The AI Shortlist Report, Wave 2: The New Claude Agrees With the New GPT More Than With the Old Claude

    The new Claude's shortlists overlap the new GPT's at 83%, more than they overlap the previous Claude's at 76%, and 12 of 25 categories have no agreed leader.

  3. Wave 1,

    The AI Shortlist Report: What One Model Release Did to 21 B2B SaaS Categories

    One model release, with no market event in between, changed 13 of 21 category shortlists and 13% of all top-five places.

Wave 4 reads the same instrument against the Wave 3 baseline, on the schedule set in a protocol published before Wave 3 collection began. Read the pre-registered protocol.

Essays

  1. The Repositioning Lag: Why Your Pipeline Still Sells the Old Story

    A median of 21 months after repositioning, 87% of 47 audited B2B SaaS companies still had a major public surface telling the old story.

  2. AI-first GTM: what it actually means

    Go-to-market run as one connected system, where AI agents do the repeatable work and humans keep judgment and final approval.

  3. Signal-based selling: the system I built at a Series A security company

    The system that monitored 400+ named accounts in Clay and scored each one hot, warm, or cold every week at a Series A security company.

  4. GTM Diagnostic: How to Identify What's Broken Before You Invest

    An assessment of the revenue engine across six areas that caps the resulting priorities at three to five.

  5. Your AI agents need a company brain: how to build one

    A versioned set of Markdown files that every AI agent reads before it writes.

  6. GEO/SEO Executive Guide for B2B SaaS Revenue Leaders

    Google AI Overviews appear on 13% of queries and cut organic click-through by 61%. The guide covers how SEO and GEO fit together.

How the research works

Every wave sends its buyer questions through the API with no system prompt and no web access, except for one search-grounded control. The exact prompts are printed in each report, so anyone can rerun the instrument.

Every shortlist question runs five times per model, and results are scored across the runs. In our calibration study, five runs recovered the same top five as twenty in every cell tested. In the noisiest category, a single query recovered only 67% of the consensus shortlist.

When a finding fails a stricter retest, the correction is published in the report and in the log below.

Corrections log

July 7, 2026

Wave 1 reported three category leaders losing the number one slot. Wave 2's retest showed those single-slot changes are sensitive to how many vendors the question asks for. The membership churn, naming gap, and maturity gradient all replicated. Both reports now state this, and crown changes are held to a multi-phrasing standard from Wave 3 forward.

The raw data

Every wave's run data, collection harness, metric definitions, and company-mapping tables are kept in a private archive. Researchers, journalists, and operators who want to verify a number can request access.

What it means for your company

The reports show where the market moved. A conversation covers where your company stands in it and what to change before the next model release.

Schedule a Conversation