Is the most popular tool the best one?
Popularity counts visits, not results. The most visited tool is often the best known or the cheapest to try, not the one that does your job best; our test-based rankings answer that question.
AI tools that lost the most traffic (September 2026)
Our reading
Adot, a Korean AI assistant, fell most (−40.5%), followed by ByteDance's TRAE (−37.5%) and Alibaba Cloud Model Studio (−34.4%). Kimi (−25.9%) and Tencent Yuanbao (−18.5%) also lost traffic.
| Rank | Product | Monthly visits | vs last month |
|---|---|---|---|
| 1 | AI search and notes assistant | 6.78M−40.5% | −40.5% |
| 2 | AI coding IDE and agent | 3.09M−37.5% | −37.5% |
| 3 | LLM API and AI app platform | 2.43M−34.4% | −34.4% |
| 4 | AI agent and office assistant platform | 2.34M−26.5% | −26.5% |
| 5 | AI app builder (vibe coding) | 5.58M−25.9% | −25.9% |
| 6 | AI assistant chatbot | 38.6M−25.9% | −25.9% |
| 7 | AI video generator | 2.3M−25.0% | −25.0% |
| 8 | AI long-form writing and reading | 5.14M−25.0% | −25.0% |
| 9 | LLM API gateway with free credits | 2.96M−24.7% | −24.7% |
| 10 | NotebookLMOur review AI research and notes assistant | 9.9M−24.7% | −24.7% |
| 11 | Qwen model API and MaaS platform | 2.11M−23.0% | −23.0% |
| 12 | AI travel planner | 14M−22.1% | −22.1% |
| 13 | Open-source AI coding agent | 13.6M−21.5% | −21.5% |
| 14 | AI video roleplay | 2.35M−21.0% | −21.0% |
| 15 | AI image upscaler | 2.67M−19.5% | −19.5% |
| 16 | AI character chat | 5.86M−19.5% | −19.5% |
| 17 | Open AI model community | 2.97M−19.1% | −19.1% |
| 18 | TuroNot an AI product peer-to-peer car rental marketplace | 9.13M−19.1% | −19.1% |
| 19 | free online AI image and file tools | 4.15M−18.8% | −18.8% |
| 20 | AI chat assistant | 11M−18.5% | −18.5% |
Where the numbers come from
Monthly website visits as published by AI产品榜 (AICPB) on 2026-10-12. GenGrades does not measure traffic itself; we check each product, add its official site and logo, and link our review where we have one.
What a visit counts
Visits to the product’s website in the month, from desktop and mobile browsers. Apps, desktop programs and command-line tools are not counted, so products used mostly that way, such as Claude Code or Codex, rank lower here than their real use, or not at all.
Traffic measures attention, not quality. A tool can be popular and still be the wrong choice for your task; our rankings by test score answer that question.
Common questions
Answer first, then why, an example, the limits and what to do next.
How is this ranking measured?
By monthly visits to each product’s website, as published by AI产品榜 (AICPB). Apps and command-line tools are not counted.