How to write about DEI on LinkedIn, from 462 ranked posts

To write about DEI on LinkedIn effectively, write from something that happened to you or inside your organisation, show a person rather than a statistic, and decide how you will handle the comments before you publish. Across 462 diversity posts from 48 creators, the topic's own strongest decile used native video 30.4% of the time against 7.8% for its weakest, opened in first person 17.4% against 11.3%, and carried fewer numbers, not more: 21.7% against 33.8%. It is also the topic with the widest spread between success and failure in our whole study.

What do 462 diversity posts show?

We isolated the diversity topic inside our 12,988-post cohort: 462 posts from 48 creators, covering inclusion, representation, accessibility, and equity at work. Then we ranked those posts against each other and compared the topic's own top decile (46 posts, 11 authors) with its own bottom half (231 posts, 29 authors). This is a within-topic comparison: good diversity content against weak diversity content, not diversity content against the rest of LinkedIn.

The performance spread is the widest of any topic in this series. The median top-decile diversity post drew 441 reactions and 99 comments. The median bottom-half post drew 74 reactions and 4 comments. That 441 figure is the highest reaction median of any group we broke out, higher than leadership, productivity, or mental health.

Read those two numbers together and you have the shape of the category. Diversity content does not perform moderately. It either travels a long way or it disappears, and the difference is almost entirely about whether there is a person and an experience in it. The topic's overall median, 106 reactions and 9 comments at 0.44 per 1,000 followers, sits just above the cohort median and tells you almost nothing useful, because very few posts actually land there.

What does the backlash risk actually look like?

It is worth being specific rather than vague, because vagueness is what makes people either avoid the topic or walk into it unprepared. Our data cannot measure sentiment, so this section is observation rather than measurement. In practice the risk shows up in four distinct forms, and they need different responses.

• Good-faith disagreement. Someone who thinks a specific policy is ineffective, argued reasonably. This is the most common form and it is not a problem. It is the comment section working. Answer it properly and your post gets better.
• Whataboutism. Comments that redirect to a group your post did not mention. Usually not worth engaging beyond a short, non-defensive line, because the exchange has no end state.
• Bad-faith provocation. Written to make you respond at length in public. The correct move is not a rebuttal. It is to hide the comment and move on.
• Targeted abuse. Sometimes coordinated, sometimes brigading from off-platform. This is a moderation and reporting problem, not a discussion, and the faster you treat it as one the less it costs you.

The asymmetry worth knowing: the writers most likely to take the fourth kind are the ones writing from personal experience of the thing they are describing, which is also the writing the data says performs best. That is an unfair trade and pretending otherwise helps nobody. What you can control is preparation.

One of the strongest posts in our diversity set is literally about this. It is a video about being on the receiving end of an attack in a comment thread, and it drew more comments than almost anything else in the topic.

How do you handle a comment section that turns?

Set the rules before you publish, not during. Once a thread is moving, every decision you make will be made under pressure and in public, which is the worst possible condition for judgment.

1. Decide what you will answer. A useful default: reply to anyone making an argument, ignore anyone making an accusation. Write that down before you post.
2. Hide, do not argue. LinkedIn lets you hide comments on your own post. It is quiet, it does not notify anyone, and it removes the audience the commenter came for. Use it early and without guilt.
3. Never reply twice to the same hostile thread. The second reply is what converts a comment into a fight, and a fight is what gets screenshotted.
4. Answer the quiet ones first. The people asking genuine questions get drowned out by the loud ones. Prioritising them changes what the thread looks like to everyone who arrives later.
5. Have someone to call. If the post is about your own identity or experience, tell one colleague you are publishing it. Reading a hostile thread alone is the part that actually wears people down.
6. Know when to close the laptop. Nothing in your engagement metrics is worth a bad evening. The post keeps working whether or not you are watching it.

The general mechanics of running a comment section, including why the first hour matters, are in the comments strategy guide.

Why do awareness-day posts underperform?

Because they are usually a link and a graphic, and links are the failure format in this topic: 36.8% of the weakest diversity posts against 10.9% of the strongest. The awareness-day post is the purest expression of the pattern. It is triggered by a date rather than by something that happened, it says a thing everyone already agrees with, and it asks for nothing.

The tell is the ratio. One awareness-day post in our top exemplar set drew 240 reactions and 8 comments. Compare that with the 297-comment example above. Reactions without conversation is the signature of content people approve of and do not engage with, and it is the most common outcome in corporate DEI communication.

A post in our top set makes the criticism better than we can.

If you are going to post on a designated day, the version that works points at something concrete your organisation did in the other 364. A number from your own hiring, a policy you changed, a person whose promotion you can describe. Anything except a graphic.

The link problem is worth separating from the awareness-day problem, because it is broader. Sharing a well-reported article about representation is a reasonable thing to do and a weak thing to post, for the same reason it is weak in every other topic: the post has no argument in it, so there is nothing for a reader to agree or disagree with. If you want to share the piece, say what you think first and let the article corroborate you. The external links analysis covers the general trade-off, including the studies that disagree about whether the platform suppresses outbound links at all.

There is one more thing to say about permanence. A DEI post is more likely than most to be read years later, out of context, by someone deciding whether to work with you. That is not a reason to avoid the topic. It is a reason to write the version you would still stand behind when the surrounding conversation has moved, which in practice means describing what happened rather than declaring what everyone should think.

Why does video dominate this topic?

Native video accounts for 30.4% of the strongest diversity posts and 7.8% of the weakest, close to a four-fold difference within the same topic. Across the whole cohort the split is 26.3% against 10.1%, so the effect is sharper here.

The reasonable explanation, and it is a hypothesis rather than a measurement, is that this is a subject where the reader is weighing who is speaking as much as what is said. A camera answers that instantly. It also does something a text post cannot: it makes the writer a specific person rather than a position, which changes how disagreement in the comments behaves.

If video is not available to you, the next best thing carries the same properties in text: first person, one specific situation, no borrowed statistics. Format specifics are in the video posts breakdown.

Why do statistics hurt in this topic?

A digit appears in the visible hook of 21.7% of the strongest diversity posts and 33.8% of the weakest. That is a reversal, and it is one of only a few topics where we see it. Across the whole cohort numbers are neutral, splitting 32.1% to 32.8%.

The likely reason is that representation statistics are the genre's wallpaper. Every reader has seen the percentage-of-boards figure, the pay-gap figure, and the pipeline figure. Reposting one is not an argument, it is a signal that you have read the same article as everybody else. And in a topic where readers arrive with strong priors, a borrowed number changes no minds; it just tells people which side you are on, which they could already tell from the first line.

Numbers you gathered are different, and they stay rare. One of the posts in our top set works precisely because the figures belong to the author.

What kinds of DEI post are there?

Sorting the category helps, because most people write only one of these and conclude the topic does not work.

The first-hand post is the strongest performer and the one with real personal cost attached. That trade-off is the honest centre of this whole topic, and anyone advising you to “just be vulnerable” without naming it is not being straight with you.

More hooks from this topic with the engagement attached are on the diversity examples page.

Should you post about DEI if you are not from an underrepresented group?

Yes, and the useful version looks different from what most people attempt. The failure mode is the post that borrows someone else's experience to demonstrate your own decency. It reads exactly as what it is, and it is the reason a lot of well-meant content in this category lands badly.

What works instead is the material you actually have. If you run a team, you have hiring data, promotion patterns, meeting dynamics, and decisions you made. Those are first-hand observations about a system you operate, and almost nobody publishes them.

• The process you changed. Structured interviews, salary bands, who writes the job description. What it cost, what it did, whether it worked.
• The thing you got wrong.A promotion round you would run differently, a hire you talked yourself out of. This is the highest-credibility post available to a manager and it does not require anyone else's story.
• Amplification with specifics. Naming a person and describing exactly what their work did, rather than a general endorsement. Specific credit is a favour. General praise is a gesture.
• The question you do not know the answer to.Posted as a genuine question, not as a lead-in to your own view. This topic's strongest posts ask far less than average, so a real question stands out.

What to skip: speaking on behalf of a group you are not in, and posting a statement of support with nothing behind it. Both are safe and both do nothing, which is why the feed is full of them.

There is a version of the risk calculation that goes the other way here too. A manager who publishes their own hiring numbers is exposed to scrutiny about those numbers, which is uncomfortable and is also the point. If nothing in the post could be held against you, the post probably did not contain anything.

Has the environment for this topic changed?

The posts in our corpus were collected before the current climate, and the conversation has moved since. Some organisations have grown quieter, some have not, and the reception a post gets now is not identical to the reception the same post would have got when the data was gathered. That is a real limit on how far you should read our numbers.

What has not changed is the structural finding, and it happens to be the finding that travels across climates. First-hand beats borrowed. A specific process beats a general principle. A number from your own organisation beats a number from a report. Those are not properties of a moment. They are properties of writing that contains information, and they survive whatever the mood is.

If anything, a quieter climate raises the value of the operational post. When statements of support become cheap or contested, the manager who publishes what their own hiring data actually shows is doing something nobody can dismiss as positioning. The general case for writing under your own name rather than an institutional one is in the personal branding guide.

How do you know if a DEI post worked?

Not by the reaction count, which is the least informative number in this category. The awareness-day example above drew 240 reactions and 8 comments; the video about being attacked in a comment thread drew 783 and 297. Reactions measure approval. Comments measure whether anyone had something to say.

Three better signals. Comments from people describing their own experience, which means the post reached past the people who already agreed. Messages you did not expect, which is where most of the real response to this material goes. And whether anyone copied the thing you described: if you published a process change and two people ask how you implemented it, the post did more than a viral one.

The signal to worry about is a thread that is entirely agreement. It usually means the post stated a position rather than contributed a fact, and the fix is the same as everywhere else in this article: more specifics, fewer principles. What to put at the end of a post to get the right kind of reply is covered in how to end a LinkedIn post.

Should this come from a person or a company page?

A person, with the company amplifying. Every property that separates the strong posts from the weak ones in this topic is personal: first person at 17.4% against 11.3%, video at 30.4% against 7.8%, direct address at 28.3% against 16.9%. A company page cannot do any of those convincingly, which is why corporate DEI content converges on the statement-of-values post that collects reactions and no conversation.

If you run communications for an organisation, the highest-return move is not writing better statements. It is supporting named people to write their own, and then resisting the urge to edit the voice out of them. That includes accepting that the resulting posts will be less tidy, will occasionally attract argument, and will be much more widely read.

It also means protecting them. If an employee posts under their own name about their own experience and the comments turn, the organisation's response in the first hour is the entire message, to that person and to everyone watching.

How do you write a DEI post yourself?

1. Start from an incident, not a principle.Something that happened, with a date and a room. Only 2.2% of this topic's strongest hooks open with a question word; the strong ones open in a scene.
2. Cut the borrowed statistic. If a number belongs in the post, make it one you can source from your own organisation.
3. Say what you did, not what should be done. Prescription invites argument about the prescription. Description invites people to compare it with their own experience.
4. Name the reader. Second person runs 28.3% at the top against 16.9% at the bottom, and it is unusually low across the topic as a whole, which makes it available.
5. Decide your comment rules first. What you will answer, what you will hide, when you will stop looking.
6. Keep the hook inside the fold. Around 206 characters, with the specific detail early; the hooks study covers what survives the cut on mobile.
7. Cap the hashtags.Four-plus stacks appear on 62.3% of this topic's weakest posts against 47.8% of its strongest. The hashtag data has the wider picture.

The craft of putting a real scene at the front of a post, which is what most of the strong examples above are doing, is covered in the storytelling guide.

What this analysis cannot tell you

These are correlations, not causes. We can say that the strongest diversity posts in our cohort use video nearly four times as often as the weakest. We cannot prove the format caused it, and this topic's top decile rests on 46 posts from 11 authors, so individual habits move the percentages. We also cannot measure sentiment, tone, or what happened in the comments, so everything in this article about backlash is observation rather than measurement, and we have tried to mark it as such. The cohort skews toward established creators with at least 1,000 followers across 65 accounts. Engagement counts are lifetime-cumulative at scrape time and post ages vary. Every text finding describes the visible hook, not the full body. Method and caveats in full are in the 34,000-post engagement study.

Writing about a different subject?

Each subject behaves differently, and we compared every topic against its own top decile rather than against the whole cohort: burnout and mental health, productivity, remote work, hiring.