Text-only posts vs image posts on LinkedIn: what 12,988 posts show

Across 12,988 LinkedIn posts we scored by engagement rate, image posts are more common in the bottom half than in the top decile: 40.3% of weak posts carry an image against 30.7% of strong ones. Text-only posts run the other way, at 19.3% of the top decile versus 16.6% of the bottom half. So the standard advice to add an image because images get more engagement does not survive contact with our data. The format that actually separates the groups is neither: native video appears in 26.3% of top-decile posts and 10.1% of the bottom half, an over-index of about 2.6x.

Do text-only posts or image posts perform better on LinkedIn?

Here is the full format mix, showing what share of each group each format accounts for. Read it as: of the 1,298 posts in the top decile, 30.7% carried an image; of the 6,494 posts in the bottom half, 40.3% did.

Three things stand out, and only one of them is about images.

• Video is the separator. A 2.6x over-index is a large effect by the standards of anything else we measured, including every hook pattern in the full study.
• Links are the strongest negative signal. Shared article and link posts make up nearly a third of the bottom half and around a fifth of the top decile, an over-index of roughly 1.5x in the wrong direction.
• The text versus image question is a small effect on both sides. Text-only moves 2.7 percentage points in one direction, images move about 9.6 in the other. Neither is the difference between a post that works and one that does not.

What does each format earn on its own?

Share of a group tells you about composition. It does not tell you what a typical post in each format earns. Here is the per-format view, cut across the whole cohort rather than by group. Every format below covers at least 13 separate accounts, so these are not one creator's habits.

The cohort-wide median engagement rate is 0.40 per 1,000 followers, so text-only and video both sit above it and image and link posts both sit below. The comment column is the one worth staring at. A median text post earns 21 comments; a median image post earns 10. Since comments are the scarcest engagement signal we measured, that is a meaningful gap even where the reaction counts are closer.

The document row deserves its caveat immediately: 60 posts across 13 accounts is a small sample, and the format was new during the period the dataset covers. The rate is striking and we would not bet a strategy on it. More on that in our write-up on document posts.

Why do image posts skew toward the bottom half?

This is where we have to separate what we measured from what we think it means.

What we measured: images appear more often in weak posts than strong ones. That is a fact about the cohort. What we think explains it: an image is the default thing you attach when a post has nothing else going for it. The stock photo of a handshake, the abstract gradient with a quote on it, the conference banner, the graphic pulled from a deck. None of those add information. They exist because the person publishing felt a post should have a picture. So the image is a marker of a low-effort post rather than a cause of its performance.

That interpretation is ours, and the data does not prove it. We can see that a post had an image. We cannot see what the image was, so we cannot separate the stock graphic from the screenshot that carried the whole argument. Anyone who tells you the mechanism with confidence, including us, is reasoning past the evidence.

Three competing explanations that would fit the same numbers:

• Displacement. Every image post is a post that is not a video post, and video is the format with the strong signal. If image is what people default to instead of the harder format, images look weak by comparison without doing anything wrong.
• Content type, not format. Certain kinds of post are image-shaped: event photos, award announcements, team pictures, conference recaps. Those tend to be corporate and low-tension regardless of medium. The image may be a proxy for the subject matter.
• Attention split.An image gives a scroller somewhere to look that is not your first line. Given LinkedIn's own engineering work suggests members make a skip-or-not decision within seconds, anything that competes with the opening line is a plausible cost.

All three are consistent with the numbers. We think the first explanation, low-effort dressing, does most of the work, because of what the top-performing image posts in our dataset actually look like. They are almost never graphics. They are photographs of something that really happened.

Is there other evidence that image posts are lower effort?

There is one piece, and it is the strongest support we have for the interpretation above. If image posts really are the default dressing on posts nobody worked hard on, you would expect them to carry the other markers of low effort too. They do.

Image posts carry a hashtag wall almost twice as often as text-only posts do. Hashtag stuffing is itself a bottom-half marker in our data, appearing about 1.6x as often in weak posts as in strong ones, and the hashtag numbers are blunt about why. So the two habits travel together: the same instinct that adds a stock graphic adds seven tags underneath it.

Video is the interesting exception. It carries plenty of hashtags too, but it asks a question in the hook far more often than any other format, at 23.5% against roughly 14% everywhere else. A question is an invitation to comment, which fits the comment counts video earns. That is a correlation and we cannot say which way it runs, but it does suggest video posts are made with more intent about what the reader should do next.

None of this proves the low-effort theory. It is consistent with it, which is a weaker and more honest claim.

Why do other LinkedIn studies say the opposite?

They do, and reconciling it is more useful than picking a side. Socialinsider's LinkedIn benchmarks, built from 1.3 million posts across 16,645 business pages between January 2024 and December 2025, report image posts at a 5.30% engagement rate against 4.50% for text. That is the opposite ordering to ours.

Four differences do most of the reconciling. Company pages publish different content than individual creators do, and a brand page's image post is often a designed asset rather than a stock filler. Engagement per impression and engagement per follower answer different questions: the first asks how good a post is once it has been shown, the second asks how much of your audience it reached and moved. Our method compares extremes and theirs compares averages, which weights outliers differently. And the date ranges do not overlap.

Notice what survives both methods, though. Video beats image in both. Link posts come last in both. Documents score highest in both. The disagreement is confined to exactly the comparison this article is about, which is a decent sign that the text versus image effect is small enough to flip under changes of method. Treat it as a tiebreak, not a strategy.

When does an image genuinely help a LinkedIn post?

The rule we use: add an image when the image is evidence, not when the post feels bare. Concretely, that is a short list.

• A screenshot of the thing you are describing. The dashboard, the message, the code, the ugly spreadsheet. Nothing you write replaces it, and readers trust a screenshot more than a description of one.
• A chart of your own data. If you have numbers nobody else has, the chart is the post and the text is the caption. Keep it to one chart with one point.
• A photo of a real moment. The first day, the room, the whiteboard, the thing you built. Every strong image post in our exemplar set is this. Nobody has ever stopped scrolling for a stock photo of a handshake.
• Before and after. The single most efficient use of an image on LinkedIn, because the comparison does argumentative work that would take three paragraphs.
• A diagram of something genuinely hard to describe. A process with branches, an architecture, a timeline. If you had to draw it on a whiteboard to explain it out loud, draw it here.
• Your face, occasionally. Not as a rule, but a real photo of you attached to a personal post is consistent with the first-person pattern that over-indexes about 2x in our top decile.

When does an image hurt?

• Stock photography. It is instantly recognisable, it says nothing, and it tells the reader the post was assembled rather than written.
• Quote cards. Text rendered as an image, which is unsearchable, unselectable, often unreadable on a phone, and duplicates what your post already says.
• The graphic pulled out of a slide deck. Designed for a room with context and a presenter. In a feed it is dense, mislabelled and unreadable at thumb size.
• An image on a post whose hook needs the space. If your opening line is doing something delicate, do not put a picture next to it competing for the same second of attention.
• An image chosen to hit a posting quota.If the honest answer to “why is there a picture here” is “posts do better with pictures”, our data says that premise is wrong anyway.

If you do use an image, size it for mobile first. LinkedIn's own documentation for link preview images specifies a minimum of 1200 by 627 pixels, a recommended 1.91:1 ratio and a 5MB maximum. That spec is for images pulled from your website into a link preview rather than for images you upload directly, but the underlying constraint is the same: a wide, light image that survives being rendered small.

Why do text-only posts hold up?

Because the format removes everything that is not the writing, and the writing is what the evidence keeps pointing at. LinkedIn's engineering team has published on using dwell time to rank the feed, noting that members make a skip-or-not decision within a small number of seconds and that dwell time is a more reliable signal than clicks, which can be rare and noisy. In a text-only post, the only thing that can earn those seconds is your first line.

A text post also gets the full width of the fold. The visible hook in our data runs to a median of 206 characters before “see more” truncates it, and that number is identical between top and bottom performers, so the space is not the variable. What you put in it is. That is covered in how to start a LinkedIn post.

What about video and documents?

Native video is the format with the largest gap in our data, appearing 2.6x as often in the top decile as in the bottom half, with a median of 181 reactions and 27 comments per post across 1,613 posts from 46 accounts. Video posts also ask a question in the hook far more often than average, which may be part of why they draw conversation.

The honest caveat on video is production cost. A video that takes ninety minutes to make has to beat a text post by a lot to be worth it on an hourly basis, and most people cannot sustain the cadence. If you can hold a phone and talk for sixty seconds without a script, the arithmetic works. If every video needs editing, it does not.

Documents score the highest median engagement rate of any format in our data at 2.42 per 1,000 followers, but on 60 posts from 13 accounts. That is enough to be worth noting and nowhere near enough to be worth restructuring around. The format was also newer during the period the dataset covers, and new formats tend to enjoy a temporary distribution advantage that fades. We go through the trade-offs in the carousel write-up.

Should you vary formats or pick one?

Pick a default and vary from it deliberately. Format variety for its own sake is a rule people invent to feel like they have a system, and it costs more than it returns because every format switch adds a production decision to a habit that is already fragile.

Buffer's analysis of its own post data found that posting two to five times a week improves both impressions and engagement. At the top of that range you are publishing something like 20 posts a month, and the only way most people sustain that is if the default costs twenty minutes rather than two hours. A sensible shape: text-only as the default, an image when the post has something to show, one video a week if you can do it without editing, and a document post when you genuinely have a sequence worth stepping through. We go through the cadence trade-off in how often to post on LinkedIn.

Format also matters less than presence. Whichever format you publish in, the twenty minutes you spend replying to comments afterwards does more for the post than swapping an image for a video would, and it is the part people skip. That argument, and the folklore around it, is in the LinkedIn golden hour. If you are early enough that none of your posts get many comments to reply to, the format question is premature anyway and the answer is in getting your first 1,000 followers.

How should you actually decide, post by post?

1. Write the post first, with no format in mind. Almost every bad format decision is made before the writing, when someone decides today is a carousel day and then goes looking for something to say.
2. Ask whether anything in it is easier to show than to say. If yes, that thing is your image. If no, publish text only and stop worrying about it.
3. Check whether the image would survive being seen at thumb size. If it needs zooming, it is a document post or it is nothing.
4. Keep links out of the body.Link posts are the worst-performing format in both our data and Socialinsider's. Put the link in the first comment if you need one, and read the mixed evidence on link penalties before treating that as settled.
5. Do not add an image to rescue a weak post. This is the specific behaviour our data argues against. A weak post with a picture is a weak post.

How do you test text versus image on your own account?

Your audience is not our cohort, and this is an easy thing to measure directly.

1. Alternate formats over 20 posts, ten each. Fewer than ten per arm and one outlier decides the result.
2. Hold the subject matter constant. Do not put the personal stories in one arm and the industry commentary in the other, or you are testing topic rather than format.
3. Compare medians, not averages. One post that travels will drag a mean anywhere.
4. Record comments separately from reactions. Our per-format gap is twice as wide on comments as on reactions, so an average that blends them hides the effect.
5. Then check the effect size against everything else. If your format difference is smaller than the difference between your best and worst hooks, you are optimising the wrong variable.

The short answer on text-only posts versus image posts on LinkedIn

Text-only posts and image posts on LinkedIn are closer than the advice suggests, and what gap exists runs against the conventional wisdom: images make up 40.3% of the weakest posts in our cohort and 30.7% of the strongest, while text-only runs 16.6% against 19.3%. Our reading is that a generic image is what a low-effort post gets dressed in, which makes it a symptom rather than a cause. Add an image when it carries information the text cannot. Otherwise publish the text, spend the saved time on the first two lines, and if you want a format that genuinely moves the number in our data, the answer is native video.

Caveats, restated because this article leans hard on one dataset. The cohort is 12,988 English posts from 65 creators, all with at least 1,000 followers and at least 25 reactions per post, so it skews toward established creators and tells you little about small accounts. Engagement counts are lifetime-cumulative at scrape time and post ages vary. We can see that a post carried an image but not what the image was, which is the central limitation on everything above. All of it is correlational: these are the formats winning posts happen to use, not proof that a format will make a post win.