AMPM STANDARD
Transparencies
排名可以被質疑,所以方法必須公開。這一頁說明 AMPM100 怎麼選、分數怎麼算、證據怎麼驗、錢和排名如何隔離。
1. AMPM100 怎麼選
AMPM 不收錄所有 AI 工具,只持續監測最值得關注的 100 席。工具要成為正式成員(Confirmed Active),必須通過以下定義:
未通過者一律顯示 Under Review 與「尚待驗證」,不用假分數充數。
2. AMPM Score 怎麼算
AMPM Score 是問問貓對每個工具的綜合評分(0–100)。以下公開計算方法,但不公開每個工具的逐項內部數字(避免被操弄)。
- 划不划算:25%
- 可靠度:20%
- 好不好上手:15%
- 實用度:30%
- 成長:10%
Methodology Version:ampm-score-v2.0.0
3. 證據怎麼驗
每個公開的價格與事實都要能回溯官方來源與抓取時間。關鍵事實(價格、免費額度、方案變動)要求兩個獨立技術管道各自取得一致結果(two-pass)才算驗證完成;只有單一來源時如實標示信心等級,資料不足顯示「尚待驗證」,絕不補 0、不猜、不用模型記憶當事實。
4. 多久重查
Every day at 09:00/09:30 (Taipei time) automatically audit all monitoring sources (zero‑AI‑cost change detection, two independent pipelines), and at 10:00 rebuild the primary database—site‑wide data share the same data timestamp. Detected changes must be independently reproduced the next day before being recorded as official updates, to prevent promotional flashes and A/B testing from generating false events.
目前資料時點:2026-09-22T03:10:28+08:00|監測工具數:110
5. 錯誤怎麼更正
發現錯誤會修正資料本身並同步全部引用處,更正紀錄公開在 /audit/。
6. Commercial firewall
AMPM has affiliate links and ads, but they are isolated in the final monetization layer: Affiliates, Sponsors, and advertisements do not affect the AMPM Score, Category Rank, placement, ENTER/DROP, Battle outcomes, or Ask responses. Rank first, monetize later. All affiliate links must be labeled and routed through /go/ redirects.
7. 資料檔版本
coverage:2026-08-19T11:54:17Z
eligibility:2026-09-21T19:10:27Z
radar:2026-09-21T16:52:01Z
score:2026-09-21T19:10:28Z
taxonomy:2026-08-19T08:00:00Z
tools:2026-09-21T21:50:02+08:00
