SortedEA LinkedIn ghostwriting
Use this skill when an assistant is drafting LinkedIn posts or comments for their executive, running or advising on a principal's LinkedIn presence, deciding whether something should be posted at all, or asking how executive LinkedIn actually works. Phrases that mean this skill: "write a LinkedIn post for my CEO", "draft something for his LinkedIn", "should she comment on this", "run his LinkedIn", "post for my principal".
Two facts shape everything below, and most advice about executive LinkedIn ignores both.
Ghostwriting is allowed. Account access is not. Nothing in LinkedIn's rules requires anyone to compose their own posts. But signing in as your principal breaches the User Agreement outright, and a great many assistants do it.
Substance is now a ranking factor, and LinkedIn says so in writing. Since 2026 the platform names "AI slop" as a category it demotes, with a member-facing report button. A polished, competent post that says nothing is now worse than posting nothing at all.
The account rule
Do not log into your principal's LinkedIn. Not with a shared password, not "just to post it", not from a saved session.
LinkedIn's User Agreement (effective 3 November 2025) says it three separate ways:
§2.2 "You will not share your account with anyone else." §2.2 "You are responsible for anything that happens through your account unless you close it or report misuse." §8.2 Do not "use or attempt to use another's account (such as sharing log-in credentials or copying cookies)."
The Professional Community Policies repeat it. Note who carries the risk: the account holder is responsible for everything that happens through the account, so this is your principal's exposure as much as yours.
Company Pages have a proper delegation model with admin roles. Personal profiles have no equivalent. There is no assistant seat.
Three workflows that are not a breach:
- You draft, they publish. Send the text, they paste and post. Slowest, and the default for most executives.
- You draft, they approve, a scheduler publishes. Compliant tools use LinkedIn's official OAuth scope (
w_member_social), which your principal authorises themselves on LinkedIn's own screen. No password changes hands, and access can be revoked. - You draft, they publish from their phone while you are on the call. Underrated. Removes the round trip without touching the rule.
If your principal offers you their password, this is the same conversation as the password one in any well-run support role. Say what the rule is, offer the scheduler, and let them decide with the facts.
Substance, and why it now matters mechanically
LinkedIn's guidance on AI-assisted content states:
"We do not want 'AI slop,' which refers to low-effort, likely AI-generated content that may sound polished on the surface but lacks a clear point of view, unique perspective, or substance." "Content that feels generic, repetitive, or lacks a clear point of view is less likely to be widely distributed."
Readers can report a post as "Seems like AI slop", and enough reports surface a warning in the author's own analytics. Disclosure of AI assistance is recommended, not required. The author is responsible for what goes out.
So the failure mode for a ghostwriter has inverted. It used to be sounding wrong. Now it is sounding fine and saying nothing, which is both the easiest thing to produce and the thing the platform actively suppresses.
Your job is extraction, not composition. An executive post is only worth publishing if it contains something only your principal knows or believes. You cannot write that for them. You can get it out of them.
Three ways that work:
- The voice note. "Talk at me for 3 minutes about the thing you were annoyed about in that meeting." Transcribe, find the one sentence that only they would say, build the post around it.
- The meeting harvest. They said something sharp on a call last week. You were there. Remind them of it and ask if they would say it publicly.
- The disagreement. "Everyone in your industry says X. Do you actually think that?" Disagreement is the most reliable source of a real point of view, and the one thing an AI draft never has.
If you cannot find a specific claim, a number, a named example, or a genuine opinion, do not write the post. Tell them there is nothing to say this week. This is a real service.
Voice
Whatever goes out reads as theirs, so it has to sound like them. If a voice profile exists, read it before drafting. If it does not, work from at least 10 things they have written themselves.
Two traps specific to LinkedIn:
Do not let their LinkedIn voice drift from their real one. Executives who post regularly often end up with a LinkedIn register that nobody who works with them recognises. If your draft would embarrass them in a meeting, it is wrong, however well it performs.
Do not adopt the platform dialect. One-line paragraphs, manufactured suspense, a rhetorical question as the opener, "Here's what I learned", "Let that sink in", the humble anecdote that resolves into a business lesson. It reads as a person imitating LinkedIn rather than a person talking, and it is exactly what the substance rules now target.
What not to publish under someone else's name
Kept short deliberately, and none of this is sector-specific. Before anything goes out, check:
- Is this theirs to say? Unannounced deals, someone else's news, a hire not yet public, a client's result. Being first is worth nothing against breaking a confidence.
- Does it commit the organisation? A statement of policy, a forecast, a promise about what the company will do. If it would need sign-off in a press release, it needs sign-off here.
- Does it name a person who has not agreed? Including praise. Especially including criticism.
- Is it about a live dispute, a departure, or anything contested?
- Would they have to delete it? Anything published under a real name is permanent in practice. Screenshots outlive deletions.
When you are unsure, do not soften it and post anyway. Ask.
Mechanics
The platform facts, the debunked folklore and the current state of ranking are all in references/platform-facts.md, with sources and with the durable facts separated from the ones that go stale. Read it when drafting.
The four that change how you write:
- 3,000 characters is the post limit, and LinkedIn documents it.
- A post carries a link preview or an image, not both. This is officially documented and it is the real, boring constraint behind most of the "links kill your reach" folklore.
- The first two or three lines are all a reader sees before "see more". Put the substance there, not a warm-up.
- Hashtags are a search aid, not a distribution mechanism. LinkedIn's own head of product calls them "nice to have, not a need to have", and the hashtag feed has been discontinued. One to three on-topic tags cost nothing. Stuffing is pointless.
Comments are writing too
An executive's comments are public, appear under their name, and often reach more people than their posts. They get the same treatment: drafted by you if that is the arrangement, published by them, and held to the same judgement about what is theirs to say.
Never comment or react from their account. Same rule, same clause, and a reaction from a logged-in assistant is the easiest kind of mistake to make and the hardest to explain.
Engagement pods and comment-automation tools are not merely ineffective. LinkedIn named them in March 2026 as "not allowed" and says it is actively degrading them. Do not use them, and say so if someone suggests it.
The workflow
- Find the substance. If there is none, stop and say so.
- Draft in their voice, with the opening lines carrying the point.
- Run the publication check above.
- Send for approval as text, not as a scheduled post awaiting a click.
- They publish, or an authorised tool does.
- Log what they changed. Every edit is a voice rule you did not have. Over a few months this is worth more than any style guide.