ActuallyUses
Menu

Trust layer

How Actually Uses decides what is real

Every number on ActuallyUses is computed from creator-mention data with receipts: source link, timestamp, short quote, and a sponsored-or-organic label. No editorial opinion, no paid placement in rankings.

receipt
“I plan every video in Notion.”
source
YouTube
time
05:12
label
organic
01

Public source

YouTube, podcasts, X, and newsletters enter the pipeline as source items.

02

Tool mention

Extraction captures the tool, usage type, timestamp or position, and a short quote.

03

Trust label

Rule and model layers classify organic, sponsored, or unclear before scoring.

04

Weighted signal

Recent, high-confidence, organic usage signals move rankings the most.

Sponsorship detection

Rules always win before a claim reaches the page.

The deterministic layer checks #ad / #sponsored / #werbung / #anzeige hashtags, sponsor-segment phrases, disclosure language, and affiliate-coded links. A model layer handles subtler cases, but explicit rules override it.

Organic 1.0xUnclear 0.5xSponsored 0.25x
Authenticity80% organic
organic 72 unclear 10 sponsored 18how we detect this →

Trending score

Four weights, one signal.

w_timexw_reachxw_authxw_usage
  • 14-day half-life makes recent mentions matter most.
  • Reach is log-scaled and capped around one million followers.
  • Organic counts 4x more than paid.
  • Uses and recommends beat passing mentions; criticism can subtract.

Money firewall

Affiliate revenue has no scoring path.

The code that computes rankings has no access to affiliate tables. Money never changes organic rankings or authenticity labels.

Corrections and quality

Every public data point can be challenged.

  • Quotes are public only up to 120 characters and always source-linked.
  • Corrections are versioned and audited.
  • Creators can claim pages and hide false mentions.
  • Prompt/model changes must pass precision >=95% and recall >=80% gates.

Questions or corrections? About & contact