Pickbench
Sample report

Which brands do AI models choose?

In each edition we ask a panel of named AI models, through their APIs, to choose between brands in real buying scenarios. We publish shares with intervals, the models, the conditions and the date. Rank and placement are never sold.

Open the AI Brand ChartFor brands

AI Brand Chart: latest editions

Leading brands of a few categories, with 95% confidence intervals. Tiers and rank intervals, not strict ranks: intervals on current boards are wide, so brands in one tier are hard to separate.

  • Smartphones

    5 AI models, 45 buying scenarios, as of October 16, 2026.

    beta: intervals not small-sample calibrated

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Brand names only: the models saw each option as a brand name, with no product facts, and answered from their own knowledge.

    The panel: 5 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (20%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (20%)
    • google/gemini-4-argon-20261013 @ gemini-4-argon-20261013 via OpenRouter (20%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (20%)
    • deepseek/deepseek-v4.1-flash-20260910 @ deepseek-v4.1-flash-20260910 via OpenRouter (20%)
    1. 1Samsungchoice share 51% (40%–63%), tier 1, rank 1 to 7
    2. 2Applechoice share 48% (35%–60%), tier 1, rank 1 to 8
    3. 3Googlechoice share 46% (33%–58%), tier 1, rank 1 to 8

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: Smartphones

  • Compact SUVs

    4 AI models, 86 buying scenarios, as of October 16, 2026.

    beta: intervals not small-sample calibrated

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Models chose on EPA specification facts only; purchase price, reliability and availability were not assessed.

    The panel: 4 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (25%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (25%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (25%)
    • deepseek/deepseek-v4.1-flash-20260910 @ deepseek-v4.1-flash-20260910 via OpenRouter (25%)
    1. 1Toyotachoice share 81% (72%–89%), tier 1, rank 1 to 2
    2. 2Mazdachoice share 67% (57%–77%), tier 1, rank 1 to 4
    3. 3Hondachoice share 37% (26%–48%), tier 2, rank 2 to 9

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: Compact SUVs

  • CRM platforms

    5 AI models, 48 buying scenarios, as of October 25, 2026.

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Brand names only: the models saw each option as a brand name, with no product facts, and answered from their own knowledge.

    The panel: 5 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (20%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (20%)
    • google/gemini-4-argon-20261013 @ gemini-4-argon-20261013 via OpenRouter (20%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (20%)
    • deepseek/deepseek-v4.1-flash-20260910 @ deepseek-v4.1-flash-20260910 via OpenRouter (20%)
    1. 1HubSpotchoice share 51% (38%–64%), tier 1, rank 1 to 5
    2. 2Closechoice share 29% (17%–43%), tier 1, rank 1 to 10
    3. 3Zoho CRMchoice share 27% (16%–41%), tier 1, rank 1 to 10

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: CRM platforms

  • Laptops

    5 AI models, 60 buying scenarios, as of October 25, 2026.

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Brand names only: the models saw each option as a brand name, with no product facts, and answered from their own knowledge.

    The panel: 5 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (20%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (20%)
    • google/gemini-4-argon-20261013 @ gemini-4-argon-20261013 via OpenRouter (20%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (20%)
    • deepseek/deepseek-v4.1-flash-20260910 @ deepseek-v4.1-flash-20260910 via OpenRouter (20%)
    1. 1Lenovochoice share 55% (45%–66%), tier 1, rank 1 to 3
    2. 2Dellchoice share 39% (28%–< 50%), tier 1, rank 1 to 7
    3. 3Applechoice share 38% (25%–51%), tier 1, rank 1 to 8

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: Laptops

  • Domestic airlines

    4 AI models, 45 buying scenarios, as of October 16, 2026.

    beta: intervals not small-sample calibrated

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Brand names only: the models saw each option as a brand name, with no product facts, and answered from their own knowledge.

    The panel: 4 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (25%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (25%)
    • google/gemini-4-argon-20261013 @ gemini-4-argon-20261013 via OpenRouter (25%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (25%)
    1. 1Southwest Airlineschoice share 51% (37%–65%), tier 1, rank 1 to 5
    2. 2Delta Air Lineschoice share 50% (36%–65%), tier 1, rank 1 to 5
    3. 3JetBluechoice share 37% (25%–< 50%), tier 1, rank 1 to 8

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: Domestic airlines

  • Hotel chains

    4 AI models, 41 buying scenarios, as of October 16, 2026.

    beta: intervals not small-sample calibrated

    Fast panel: the fast or small current models of each vendor, not the full flagship panel.

    Brand names only: the models saw each option as a brand name, with no product facts, and answered from their own knowledge.

    The panel: 4 AI models
    • anthropic/claude-haiku-5.5-20261007 @ claude-haiku-5.5-20261007 via OpenRouter (25%)
    • openai/gpt-6-luna-20260922 @ gpt-6-luna-20260922 via OpenRouter (25%)
    • google/gemini-4-argon-20261013 @ gemini-4-argon-20261013 via OpenRouter (25%)
    • mistralai/mistral-large-4-0-20261006 @ mistral-large-4-0-20261006 via OpenRouter (25%)
    1. 1Marriottchoice share 36% (24%–< 50%), tier 1, rank 1 to 7
    2. 2Hiltonchoice share 28% (17%–41%), tier 1, rank 1 to 8
    3. 3Hyattchoice share 26% (15%–39%), tier 1, rank 1 to 8

    Choice share in shortlists of 4; the line marks chance (25%).

    Full board: Hotel chains

All 91 categories

For brands and agencies: Pickbench Category Data

Pickbench Category Data is one pre-registered measurement per category and edition on named API models, available to any brand or agency in that category. Today we measure AI model choice share with 95% confidence intervals and tiers (never strict ranks) and model-stated reasons. Model Memory has published studies in 8 categories and fact banks for 4 car categories, live search with cited sources is in progress, change alerts are coming, and agent measurement is in development until retrieval concordance passes. Rank, inclusion and public rows are never sold, and paying customers are disclosed in the conflicts register.

  • Measured today: AI model choice share by category and named API model with 95% confidence intervals, plus model-stated reasons. Need and budget cuts are reported where a category-depth run reaches 30 tasks per cut.
  • Model Memory: published studies in 8 categories, plus fact banks for 4 car categories; live search with cited sources in progress (STUDY-V1a); change alerts coming.
  • Free public category boards, Brand Battles and fact corrections (Pick Bench opens soon); Category Report $490 one-off; Category Data $490/mo; Agency $499/mo for up to 3 client brands; custom research on request.
Request the sample reportCategory Data and pricing

Rankings, inclusion and placement are never sold.

How we measure

Each run follows an analysis plan registered before it starts. Models answer through their APIs with web search off, the buying scenario is the unit of analysis, and every board is signed so anyone can verify it. Results name the API model, the conditions and the date.

Read the methodology