How to Humanize a LinkedIn Recommendation

How to Humanize a LinkedIn Recommendation

AI LinkedIn recommendations praise everything and prove nothing. Lock one project, your role as the writer, and a checkable detail. Then a short voice pass.

•
4 min read
linkedin recommendationworkplacerecommendationsai writingcareer

A LinkedIn recommendation is a named person saying one true thing about working with someone. ChatGPT writes a different genre: “results-driven leader” stacked on “cross-functional excellence,” with no project a stranger could picture.

You are the author of record. Humanize only after the incident is yours.

Hubs: LinkedIn posts and newsletters, recommendation letter (longer, often a manager letter), speaker bio.

What readers skip

PatternWhy it fails
Adjective stacksCould describe anyone
“It was a pleasure” with no workNo signal
Metrics you did not seeYour credibility, not theirs
Four projects in one blurbNothing sticks
Their voice, posted as yoursObvious ghost draft

Lock one project

FieldYour entry
Your relationshipManager, peer, client. Say which
One projectName a launch, migration, or deal you were on
What they specifically didA decision or behavior, not a trait
Result you can stand behindOnly if you know it
What you will not sayPrivate feedback, medical, gossip

Prompt: “Write 80–120 words in first person from this card. One project. No buzzword list. End with whether I would work with them again, if that is true.”

Before and after

AI sludge:

Alex is a visionary, results-oriented professional who consistently leverages synergies to drive impactful outcomes. Any organization would be lucky to have them.

After one project:

I managed Alex on the 2025 billing migration. They caught a tax-rounding bug in staging and wrote the rollback note the on-call team actually used. I would hire them again for a messy data project.

The second version is short enough to post and specific enough to trust.

If they ask you to “just sign” their draft

People sometimes write their own recommendation and ask you to post it. You can still help, with a narrower job:

  1. Circle one sentence that matches something you saw. That sentence can stay.
  2. Delete claims you cannot confirm, even if they sound modest.
  3. Add your relationship in the first line so a stranger knows the vantage point.
  4. Humanize only after those cuts. Polishing a draft you do not believe makes it more convincing, which is worse.

If nothing in their draft matches your memory, send a short note: you are glad to write one from a project you name, and you are not posting this version. That is a normal boundary, not a tone problem.

Peer praise on a profile is different from a formal letter with a signature block. For the letter, keep incidents and your title exact, as in recommendation letter. For a networking note that is not a recommendation, see networking message.

Workflow

  1. Pick one project. If you cannot name it, do not draft yet.
  2. Write the card in your own words first, even as bullets.
  3. Optional model draft from the card only.
  4. One Human Writes pass so it sounds like your email voice, not a press release.
  5. Read it as if a recruiter will ask you about the story. Cut any line you would stumble on.

This is not a performance review. For review language, see performance reviews. For a formal letter a school or employer requested, see recommendation letter.

After you post

A recommendation is not finished when LinkedIn accepts the text. A week later you should still be willing to say the same thing on a call. If a recruiter asks “what did they do on that migration?”, your answer should match the paragraph without a second story appearing. That is the test of a single-project draft.

Do not go back and “strengthen” it with a model a month later. Added adjectives are how a specific note turns generic. If you remember another project, write a second recommendation only if the product allows it and you still mean it. Two vague ones are worse than one concrete one.

Keep a private copy of the card. If they ask for a similar blurb for a talk or a job form, you can reuse the same incident without inventing a new one. Match the channel length. A conference bio is a different artifact: conference bio.

Bottom line

One project, your role, a detail you saw. Then polish. A LinkedIn recommendation works when a stranger could retell the incident. Human Writes helps the wording after that incident is real.