What makes a social post get shared
The scorer is built on published research. Here is what it shows, how strong each piece is, and what it doesn't show.
Concrete wording: the strongest evidence
Hu, Pilgrim, Zhao and Hills (2025, Quarterly Journal of Experimental Psychology) measured the words in millions of posts on Twitter and Reddit against ratings of how concrete each word is: "keyboard" is concrete, "privacy" is abstract. Posts with more concrete wording were shared more on both platforms. In a third study they rewrote real posts to be more or less concrete, keeping everything else the same, and the concrete versions were shared more. That makes concreteness the one effect with causal evidence, so it counts most in the score.
Arousal: real stakes help, hype doesn't
Words that carry energy (arousal) predicted more sharing in all three studies. But in ad copy, the most "arousing" words are hype: "excited", "amazing", "innovation". We checked: rated by arousal alone, ad copy scored as high as real top Reddit posts. So the scorer counts arousal only from words that aren't hype, like "deadline", "interview" or "crashed".
Positivity and jokes
Positive wording predicted more sharing on Twitter but less on Reddit. Jokey wording predicted less sharing on both. The scorer marks hype positivity down on Reddit and flags default humour.
Curiosity needs a closable gap
From Loewenstein's information-gap theory (1994): a first line that opens a specific question the post answers pulls readers in; a vague tease doesn't. This is theory rather than a tested effect on posts, so it counts less.
What the research does not show
- That wording alone makes a post go viral. Who posts, when, and the image or video matter a great deal; the studies couldn't measure follower counts.
- That one rule fits every community. The concreteness effect reversed in 6 of the 29 subreddits studied, which is why the scorer also checks each community's own rules and the shape of its top posts.
How we checked the scorer
We scored the titles of real top posts from the year against generic ad-style titles: real winners scored clearly higher (3.2 against 2.1 out of 5), on subreddits the scorer was tuned on and on 475 titles from 29 others. We also keep a set of labelled good and bad drafts that every change must still sort correctly. Inside a single subreddit's top 25, scores don't predict which post got more upvotes, so use the score to filter out weak drafts, not to rank good ones.
Sources
- Hu, Pilgrim, Zhao & Hills (2025). Concrete language enhances sharing of social media posts. QJEP. doi:10.1177/17470218251392831
- Brysbaert, Warriner & Kuperman (2014). Concreteness ratings for 40 thousand generally known English word lemmas. Behavior Research Methods.
- Warriner, Kuperman & Brysbaert (2013). Norms of valence, arousal, and dominance for 13,915 English lemmas. Behavior Research Methods.
- Berger & Milkman (2012). What makes online content viral? Journal of Marketing Research.
- Loewenstein (1994). The psychology of curiosity. Psychological Bulletin.