Models
The frontier models behind AI search, and how rarely they agree on who to recommend. Every brand ChatGPT, Claude, Gemini, Perplexity and more name for the same question, compared head to head.
The models we track
Eight frontier engines, ranked by how often each one returns a brand answer. Each shows the model it lines up with most.
Claude23%
Claude27%
Claude35%
Claude26%
Claude35%DeepSeek35%
AI Overviews17%ChatGPT20%Who agrees with whom
Every pair of models, by how often they name the same brand for a question. Greener means they agree more.
ChatGPTClaudeGeminiPerplexityDeepSeekGrokMeta AIAI OverviewsClaude+DeepSeek35%Perplexity+Meta AI10%Claude sits closest to the pack; Perplexity is the biggest outlier. Win there and you reach an audience the others miss.
How often do they line up
The spread of agreement across every comparison. Most questions land in the middle; near-unanimity is rare.
Agreement is the share of a question's brand picks the two models share. The long middle is the real story: the models mostly half-overlap, so where you rank depends on which engine a buyer asks.
Where they split
Cross-model agreement by the kind of question asked. Tight, comparative questions converge; open ones scatter.
See it in the wild
Real questions from the study — where the models split on who to recommend, and the rare ones where they all agree.
“best project management software for a fast-growing remote team”
13%agreeChatGPTAsanaClaudeClickUpGeminiMonday.comPerplexityLinearDeepSeekTrelloGrokNotionMeta AIWrikeAI OverviewsJira“which CRM should a small B2B sales team use”
13%agreeChatGPTHubSpotClaudePipedriveGeminiZoho CRMPerplexityFreshsalesDeepSeekFolkGrokSalesforceMeta AICloseAI OverviewsCopper“compare accounting tools for freelancers and sole traders”
13%agreeChatGPTQuickBooksClaudeFreshBooksGeminiXeroPerplexitySageDeepSeekBonsaiGrokWaveMeta AIZoho BooksAI OverviewsFreeAgent“good email marketing platform for an online store”
13%agreeChatGPTKlaviyoClaudeMailchimpGeminiBrevoPerplexityConvertKitDeepSeekActiveCampaignGrokOmnisendMeta AIDripAI OverviewsMailerLite“best password manager for a small business”
13%agreeChatGPT1PasswordClaudeBitwardenGeminiDashlanePerplexityRoboFormDeepSeekKeeperGrokNordPassMeta AILastPassAI OverviewsProton Pass“what is a good website builder for a portfolio site”
13%agreeChatGPTSquarespaceClaudeWebflowGeminiWixPerplexityFormatDeepSeekWordPressGrokFramerMeta AICarrdAI OverviewsPixpaBuilt by giving the same prompts to every model and comparing the brands each one returns. 599K high-quality comparisons across 44K reports, Aug 12, 2025 to Mar 11, 2026. Agreement is the overlap of brand picks between two models; appearance is how often a model returns a usable brand answer.