How to write social media captions that actually convert

What does the research say makes a caption engaging?

Two findings come up again and again in the academic literature. The first is emotion. In a study of every New York Times article published over a three-month period, Berger and Milkman found that content evoking high-arousal emotions such as awe, amusement, anger or anxiety was shared significantly more than content evoking low-arousal states such as sadness [1]. The lesson for caption writers is not to be dramatic for its own sake, but to lead with the part of the story that genuinely stirs something.

The second is personality. Analysing 106,316 Facebook posts from 782 companies, Lee, Hosanagar and Nair found that content expressing brand personality, such as humour and emotion, was associated with higher likes, comments and shares, while directly informative content such as price and deal mentions was associated with lower engagement when it appeared on its own [2]. The same study found the combination worked: informative content performed better when paired with personality. A caption that only announces is easy to scroll past. A caption that sounds like a person announcing something is not.

One caution from psychology: emotional language is also the default we all reach for. Rocklage, Rucker and Nordgren showed that the mere intent to persuade automatically shifts people toward more emotional wording, even when the audience is rational and unlikely to respond to it [3]. Emotion in a caption should be a deliberate choice matched to the audience, not a reflex.

Should you ask questions in your captions?

Yes, if your goal is comments. In one of the most cited studies of brand post popularity, de Vries, Gensler and Leeflang analysed 355 posts from 11 international brands across six product categories and found that posing a question was significantly and positively related to the number of comments a post received [4]. Questions invite an answer, and comments are exactly the kind of active engagement platform algorithms tend to reward.

The same study carries a nuance most caption advice skips: questions were negatively related to the number of likes [4]. A question turns a post into a conversation rather than a statement people can simply approve of. So the honest guidance is to pick the outcome you want. If you want discussion, ask something specific and easy to answer. If you want broad approval and shares, make a confident statement and save the question for another post.

Does caption length actually matter?

Less than most guides claim. In the de Vries study, message length was included as a control variable and had no significant effect on either likes or comments once the content of the post was accounted for [4]. The brand posts in that sample averaged around 28 words, which is short, but the data suggests it was what the posts said, not how long they were, that drove results.

That does not mean length is irrelevant in practice. Each platform truncates captions at different points, so the first line has to do the emotional work identified by Berger and Milkman before the fold [1]. A useful rule: write as long as the substance requires, put the hook in the first sentence, and cut anything that is neither emotion, information nor an invitation to act.

Should your caption say "you"?

The evidence says yes. Cruz, Leonhardt and Pezzuti ran a field study of real brand posts on Facebook plus two controlled experiments and found that second-person pronouns, words like "you" and "your", enhanced consumer involvement and brand attitude, because readers process the message in relation to themselves [5]. "You can book a spot this week" tends to outperform "Spots are available this week" for a reason that is measurable, not stylistic.

How confidently you say it matters too. Pezzuti, Leonhardt and Warren analysed brand messages on Facebook and Twitter and found that posts using certainty words such as "always", "everything" and "forever" attracted more likes, comments and shares, a result they replicated in two controlled experiments [6]. Hedged captions read as forgettable. Within the bounds of honesty, commit to your claim.

Where should the call to action go, and what kind?

Match the call to action to the objective. Lee, Hosanagar and Nair found that directly informative content, including deals and promotions, performed worst for likes and comments but was effective at driving click-throughs, the behaviour that actually leads toward purchase [2]. So a link-and-offer caption is not a failure when its engagement is low; it is doing a different job. Judge it on clicks and conversions, not likes.

For engagement-led posts, de Vries and colleagues found that contests were positively related to likes, while a generic "call to act" instruction showed no significant effect on likes at all [4]. Telling people to engage does little. Giving them a concrete, low-effort reason, a question to answer, a choice to vote on, something specific to tag a friend about, is what the data supports.

How do you turn this into a repeatable caption structure?

The research above compresses into a four-part structure you can apply to almost any post.

The per-channel part matters because the same message truncates, renders and performs differently on each platform. Enki Socials' drafting assistant applies these evidence-based principles for you and adapts the result to each channel, with a live preview of exactly how the caption will appear before you schedule it. The structure itself:

• Open with the genuine emotional core of the message, high-arousal where honest, in the first sentence [1].
• Deliver the substance in a voice that sounds like a person, addressed to "you", stated with confidence [2, 5, 6].
• Close with a single action matched to the goal, a specific question if you want comments, a clear link if you want clicks [2, 4].
• Keep only the words that serve one of those three jobs, because length itself will not save a caption [4].

Sources

1. 1.Berger, J. & Milkman, K.L. (2012). "What Makes Online Content Viral?" Journal of Marketing Research, 49(2), 192-205. https://doi.org/10.1509/jmr.10.0353
2. 2.Lee, D., Hosanagar, K. & Nair, H.S. (2018). "Advertising Content and Consumer Engagement on Social Media: Evidence from Facebook." Management Science, 64(11), 5105-5131. https://doi.org/10.1287/mnsc.2017.2902
3. 3.Rocklage, M.D., Rucker, D.D. & Nordgren, L.F. (2018). "Persuasion, Emotion, and Language: The Intent to Persuade Transforms Language via Emotionality." Psychological Science, 29(5). https://doi.org/10.1177/0956797617744797
4. 4.de Vries, L., Gensler, S. & Leeflang, P.S.H. (2012). "Popularity of Brand Posts on Brand Fan Pages: An Investigation of the Effects of Social Media Marketing." Journal of Interactive Marketing, 26(2), 83-91. https://doi.org/10.1016/j.intmar.2012.01.003
5. 5.Cruz, R.E., Leonhardt, J.M. & Pezzuti, T. (2017). "Second Person Pronouns Enhance Consumer Involvement and Brand Attitude." Journal of Interactive Marketing, 39, 104-116. https://doi.org/10.1016/j.intmar.2017.05.001
6. 6.Pezzuti, T., Leonhardt, J.M. & Warren, C. (2021). "Certainty in Language Increases Consumer Engagement on Social Media." Journal of Interactive Marketing, 53, 32-46. https://doi.org/10.1016/j.intmar.2020.06.005