Every article is scored across 20 influence tactic categories
using a combination of ML models and NLP analysis. Scores range
from 0 (no indicators) to 100 (strong indicators).
The leaderboard aggregates these scores by source over a 7-day
rolling window. Sources need at least 10 analyzed articles to qualify.
Scoring method changed on
.
A category that cannot be judged from the text, or whose scoring failed, is now reported
as not determinable instead of being given a score, and the overall score is computed only
from the categories that were scored. A rounding error that raised some category scores
by one point was corrected on the same date.
News scores come from quick analysis, which leaves out timing,
phrase repetition and rapid behaviour shifts because they need dates or other sources.
Scores from before this date were computed the old way, so compare them with care.
0–25 LOW — Minimal manipulation indicators
26–50 MODERATE — Notable patterns present
51–75 HIGH — Significant manipulation indicators
76–100 SEVERE — Strong manipulation indicators
"You decide what this means."
Learn more about the ITP methodology →