Title Framing Style
How the video title structures its hook to draw viewers in. This is classified by an LLM from the title text alone and captures the rhetorical strategy — whether the title accuses, questions, reacts, or tells a personal story.
YouTube's algorithm surfaces videos that earn clicks from impressions. Different framing styles trigger different psychological responses — a question creates a knowledge gap the viewer wants to close, while an accusation activates tribal identity and outrage. The framing you choose determines who clicks and how fast.
Each value below shows two numbers when available: Regular/Shorts is the effect on a typical video, and Top 10% is the effect on videos in the top 10% of views (from a quantile regression). Use the Top 10% column to see factors that disproportionately help videos that go big.
List Number
Regular +61% Shorts +72% Top 10% +75% SIGNIFICANT p = 0.014 · vs. averageThe title uses a number to promise a structured, scannable breakdown. Signals the viewer will get organized, digestible information.
Real examples from the dataset
Reaction
Regular +41% Shorts +27% Top 10% +74% SIGNIFICANT p = 0.001 · vs. averageThe title frames the video as a response to something that happened. Leverages the viewer's existing awareness of a news event and promises the politician's take on it.
Real examples from the dataset
Label
Regular +45% Shorts -16% Top 10% +73% not significant p = 0.359 · vs. averageThe title assigns a characterizing label to a person, policy, or situation — branding it as something specific to shape how viewers think about it.
Real examples from the dataset
Confrontation
Regular +18% Top 10% +46% not significant p = 0.245 · vs. averageThe title frames the video as a direct clash between two people or sides. Implies dramatic conflict the viewer can watch unfold.
Real examples from the dataset
Quote
Shorts +26% Top 10% +0% not significant p = 0.476 · vs. averageThe title features a direct quote — usually a punchy soundbite from the politician or an opponent. Gives the viewer a taste of the drama inside.
Real examples from the dataset
Announcement
Shorts -20% not significant p = 1.000 · vs. averageThe title presents news or information in a neutral, declarative way. Common on official channels but tends to feel like a press release rather than compelling YouTube content.
Real examples from the dataset
Question
Regular -17% Shorts +17% Top 10% -9% not significant p = 1.000 · vs. averageThe title poses a question the viewer wants answered. Creates an information gap that compels the click — the viewer literally cannot know the answer without watching.
Real examples from the dataset
Other
+0% views Top 10% +120% not significant p = 1.000 · vs. average
Accusation
Regular -10% Shorts +2% Top 10% -15% not significant p = 0.870 · vs. averageThe title directly accuses someone of wrongdoing, corruption, or dishonesty. Creates an immediate sense of stakes and conflict.
Real examples from the dataset
Breaking News
Regular -19% Shorts +14% Top 10% -27% not significant p = 0.464 · vs. averageThe title signals time-sensitive, urgent information that just happened. Borrows the conventions of cable news chyrons to create urgency.
Real examples from the dataset
Personal Story
Regular -20% Shorts +10% Top 10% -20% not significant p = 0.395 · vs. averageThe title frames the video around a personal experience or narrative. Creates intimacy and makes the politician feel relatable rather than institutional.
Real examples from the dataset
Declaration
Regular -20% Top 10% -12% not significant p = 0.089 · vs. averageThe title makes a bold, definitive statement — the speaker is taking a stand or stating a position with conviction rather than hedging.
Real examples from the dataset
Call To Action
Regular -25% Shorts -10% Top 10% -40% not significant p = 0.137 · vs. averageThe title directly asks the viewer to do something — sign a petition, call their representative, or take a specific action. Frames the video as a mobilization moment.
Real examples from the dataset
Other
Shorts -53% Top 10% -58% SIGNIFICANT p = 0.034 · vs. average