The September 2026 Index Report
78 categories changed hands · WebFX climbed 23 places in web design companies · 130 categories left the AI models split. Every line below is computed from the frozen September 2026 edition and its predecessor. Nothing here is written by hand, and nothing is for sale.
New #1s · 12
Biggest risers
Biggest fallers
Notable entrants · new to the top 5
Where the AI models disagree
AI inference platforms divides the AI models: 7 different top answers across the 8 with answers, led by AWS SageMaker (Claude and DeepSeek).
AI sales tools divides the AI models: 7 different top answers across the 8 with answers, led by Clay (Gemini and Google AI Mode).
AI evaluation tools divides the AI models: 6 different top answers across the 8 with answers, led by Braintrust (Claude, DeepSeek and Copilot).
No agreement in home gym equipment: Peloton Cross Training Bike+ leads for ChatGPT, Bowflex Xtreme 2 SE for Claude, and 4 other picks split the rest.
No agreement in marketing podcasts: Marketing School leads for ChatGPT and Grok, HubSpot for DeepSeek, and 4 other picks split the rest.
No agreement in webcams: Logitech MX Brio Ultra HD 4K Webcam leads for ChatGPT and Copilot, Insta360 Link 2C for Claude, and 4 other picks split the rest.
AI agents divides the AI models: 5 different top answers across the 8 with answers, led by OpenAI (4 of the 8 AI models with answers).
AI search APIs divides the AI models: 5 different top answers across the 8 with answers, led by Tavily (4 of the 8 AI models with answers).
Closest races
Regional divergence
45 of 48 categories crown a different #1 locally
46 of 48 categories crown a different #1 locally
46 of 48 categories crown a different #1 locally
20 of 48 categories crown a different #1 locally
Every claim above links to the frozen September 2026 record: the full ranking, the per-model splits, and every verbatim answer as captured. Rankings are computed from AI responses alone and are not for sale. Browse the September 2026 edition → Read the methodology →
Integrity: run-level root hashes for this edition
Global sweep 822f520398b77b22f14ac88f82061f955288dd1441a39c51b4cf4ef807e793eb
Regional sweep 6c4838d7903cd2d288121681f46a96e5806bb332bf9d1f8d3e2db8aa1d763172
Each hash is a SHA-256 over every category’s frozen answer corpus. Verification spec ships in every record’s JSON download.