LinkedIn Recommendation Examples: Five That Work, and Why

A LinkedIn recommendation carries weight when it names a specific situation, a specific result, and usually something the person had to overcome. The LinkedIn recommendation examples below — manager, peer, upward, client and a six-week collaborator — show that pattern in full text, alongside the current mechanics of asking for one, writing one, and fixing one that lands wrong.

What a LinkedIn recommendation actually is, and who can write one

A recommendation is a short written commendation from someone you have worked with, displayed in a dedicated section of your profile once you accept it. It is not an endorsement — endorsements are the one-click skill confirmations. Recommendations are prose, attributed to a named person with a visible job title, which is where their credibility comes from.

LinkedIn's own help documentation sets out the boundaries, and they are narrower than most people assume:

• Only 1st-degree connections can write one, and LinkedIn frames it as coming from someone you work or have worked with. You cannot collect recommendations from strangers.
• One recommendation per position, per member. A former manager who worked with you at two different companies can recommend you twice — once against each role. The same manager cannot write you two recommendations for the same job.
• There is no cap on the total number you can receive or give.
• Requests are made in your primary profile language, which matters if you maintain a profile in more than one. To request in another language you have to switch your primary language first.
• Disconnecting withdraws them. LinkedIn states that if you remove a connection, any recommendations between you and that person are withdrawn — and they are not restored if the connection is later re-established.

Where they appear is worth knowing precisely, because it changes who your recommendations are actually working on. Accepted recommendations sit in the Recommendations section of your profile under a Received tab, and LinkedIn says they are visible to 1st, 2nd and 3rd-degree connections while those people are signed in. Someone viewing your public profile signed out sees the number of people who have recommended you and a maximum of two of the recommendations themselves.

Recommendations you write for other people appear on your own profile under a Given tab, but LinkedIn is explicit that they do not show on your public profile. That asymmetry is useful: writing generously costs you nothing in profile clutter, and it is visible to anyone signed in who goes looking.

On length, one honest note. Third-party guides widely quote a 3,000-character limit for the recommendation text box, but LinkedIn does not publish that figure on its own recommendations help pages, so treat it as unconfirmed. It is also academic. Nothing in this article argues for a long recommendation, and the examples below run 120 to 220 words each — well inside any plausible limit and, more importantly, inside a reader's attention.

What separates a recommendation that carries weight from generic praise

The failure mode is almost universal and easy to name: the writer praises a person without describing anything that happened. "A pleasure to work with, always went above and beyond, would recommend to any team" contains no information. A reader cannot distinguish it from a recommendation for anyone else, so it converts to zero.

The recommendations that do work share three structural features. Not stylistic ones — structural. You can check for them mechanically before you post.

A situation with edges

A named project, a named quarter, a specific constraint. "During the 2024 checkout rebuild" tells a reader that something concrete is about to be described. "On numerous projects" tells them nothing is. The situation does not have to be dramatic — it has to be locatable in time and scope.

A result someone else could verify

A number is the cleanest form: a close cycle that went from five weeks to six days, a backlog median that dropped from 40 hours to six. But an artifact works too. "We shipped one pattern and it's still in the system" is verifiable in a way that "improved consistency" is not. If you genuinely have no number, name the thing that exists now because of them.

Something the person had to overcome

This is the counterintuitive one, and it is the single strongest credibility signal available to you. A recommendation that admits a starting weakness — presentation skills, a directness that landed badly at first, a gap the person closed — reads as written by someone who actually observed the person rather than someone returning a favour. It also implies the rest of the praise is calibrated. Handle it carefully: you are describing a weakness that was overcome, in the past tense, with the resolution attached. Never a live one.

There is a fourth feature that is optional but powerful: a placement statement. One sentence saying what kind of role or team this person is right for. It is the sentence a hiring manager is trying to construct in their own head while reading, and handing it to them is a favour to everybody.

Five LinkedIn recommendation examples, annotated

These are written to be read as models rather than copied. The names and numbers are invented for illustration; the shapes are not. Each one comes from a different vantage point, because a reader weighs a recommendation partly by what the writer was in a position to observe.

1. A manager writing for a direct report

Priya joined my team as a mid-level analyst and left it running our forecasting function. The thing I'd point to is the Q3 2024 inventory model. We were carrying around $4M in slow-moving stock because the old model treated every SKU the same. Priya rebuilt it with a seasonality layer and a manual override for promotional periods, and we cut that number by about a third over two quarters. I'll be honest about the starting point: when she joined, presenting was not her strength. Her first two exec readouts were dense and she lost the room. She asked me for feedback after the second one, went and got coaching from our comms lead, and by year end she was the person I sent into the CFO's review without a rehearsal. That progression is the reason I'd hire her again. Priya is at her best in a role where someone has to own a messy data problem end to end and then explain it to people who don't want the detail.

Why it works. The manager has authority over outcomes, so the number belongs to them to report. The weakness admission is doing the heaviest lifting here — it converts the whole thing from a reference letter into an observation. And the closing sentence tells a recruiter exactly which requisition to match her against.

2. A peer writing for a peer

Marcus and I were both senior designers at Northwood for three years, on different pods, which means I saw his work the way a colleague does rather than the way a manager does — mostly in critique, mostly under deadline. The specific thing I'd tell anyone hiring him: during the 2024 checkout rebuild, our two pods had built conflicting patterns for the same address form and it was going to ship broken. Marcus didn't escalate it. He booked ninety minutes, brought both versions, and walked us through which parts of each were load-bearing. We shipped one pattern and it's still in the system today. He's also the person on that team whose critiques I actually wanted — he'd tell you a layout wasn't working and then stay to work out why, which is rarer than it should be. I'd join a pod with him again tomorrow.

Why it works. Peer recommendations fail when they try to sound like manager recommendations, claiming visibility into results the writer never had. This one opens by naming its own vantage point, which makes everything after it more credible, not less. The example chosen is a conflict resolved rather than a compliment paid — peers see conflict, and that is exactly the material a hiring manager cannot get from a reference call with a former boss.

3. A direct report writing upward

I reported to Dana for two years at Halvorsen, through a reorg that took our team from eleven people to six. I want to describe what she did in that period, because it's the part that doesn't show up on a résumé. She told us the headcount number the day she knew it, before she had a full plan, which was uncomfortable and correct. She then spent about six weeks personally working referrals for the five people leaving; three of them landed before their last day. The six of us who stayed had our scopes rewritten, in writing, within a fortnight, so nobody spent a quarter guessing. On the work itself: she took our support backlog from a 40-hour median first response to under six, mostly by killing three reporting workstreams nobody read. I've had seven managers. She's the one whose calendar I'd still get on.

Why it works. Upward recommendations are underused and disproportionately valuable, because reports observe something nobody else does: how a leader behaves when the news is bad. This one does not claim to assess Dana's strategy — it reports behaviour, with dates and counts, and lets the reader draw the conclusion. The final line earns its place because the seven-managers comparison gives it a denominator.

4. A client writing for a contractor

We hired Tomas as a fractional CFO for what we thought was a six-week cleanup and kept him for fourteen months. The situation: we'd raised a seed, had 30 people, and were closing our books about five weeks late every month, which meant the numbers in our board deck were always stale. Tomas rebuilt the close process, moved us off a spreadsheet consolidation, and got us to a six-day close by month four. He also found that we'd been misclassifying roughly $180k of contractor spend, which mattered a great deal once we reached diligence. Worth saying plainly: he is not a cheap hire, and he is direct in a way that surprised two of our founders early on. If you want someone to agree with your model, he's the wrong person. If you want someone who will tell you the model breaks in month nine, he's exactly right. We've since referred him to two other companies in our portfolio.

Why it works. A buyer's recommendation answers a different question than a colleague's: was this worth the money? The engagement length, the deliverable and the diligence detail all answer it. The caveat about cost and directness is the same move as the weakness admission in the first example, relocated — it prices the person honestly instead of overselling them, and the referral line at the end is the strongest possible proof that the caveat is not a complaint.

5. A short one for a brief collaboration

I worked with Ana for six weeks — she ran the accessibility audit on our public site while I owned remediation. Two things I noticed. She scoped the audit so we could act on it: every finding was tagged by effort and legal exposure, not just severity, which is why we shipped 80% of the fixes instead of arguing about them. And when we pushed back on two findings, she conceded one and held the other with a citation. That's the behaviour you want from an auditor. Six weeks isn't long enough for me to speak to how she is over a year. It was long enough to know I'd bring her back.

Why it works. Short collaborations produce the weakest recommendations precisely because people try to hide how short they were, and the vagueness required to do that drains the text. This one leads with the duration and the exact division of labour, then makes one narrow claim well. The penultimate sentence — refusing to claim more than the writer can see — is what makes the final sentence land. A 110-word recommendation that is honest about its own limits beats a 400-word one that pretends to a friendship it never had.

How to ask so the recommendation actually gets written

Most requests die of vagueness. "Would you mind writing me a LinkedIn recommendation?" hands the recipient a blank page and a research task: they now have to remember what you worked on, decide what to praise, and find a number. That work is why the request sits unanswered for three weeks and then quietly expires.

The mechanics first, because the request flow constrains what you can send. LinkedIn gives two paths. From your own profile, scroll to the Recommendations section, click the plus icon and choose Ask for a recommendation. From a connection's profile, open the More menu and choose Request a recommendation. Either way you must specify your relationship and your position at the time, and you get an optional message box. A few constraints worth knowing:

• You can only request one recommendation from a given connection at a time, though there is no cap on the total you request or receive.
• Once a request is sent, the relationship type can't be changed. LinkedIn's instruction is to ask the connection to delete the request so you can send a new one.
• The message box is optional — and an empty or generic one is the version that sits unanswered. LinkedIn publishes no data on response rates, but a request that hands someone a blank page makes them do all the remembering. Replace it.

What to put in that box: a specific ask with suggested talking points. Not a draft to be signed — that is transparently ghostwritten and reads that way — but the raw material that lets someone write in ten minutes instead of forty.

1. 1Name the window and the work. "You managed me on the pricing migration, roughly March to October 2024." This alone saves the writer five minutes of memory work.
2. 2Offer two or three angles and let them pick. "If it's useful, the things I'd find most helpful covered are the migration itself, the vendor negotiation, or how I handled the handover to the ops team — whichever you actually saw most of."
3. 3Supply the numbers you know they'd have to look up. "For reference, we came in 11 days ahead and about $40k under." You are removing a lookup, not dictating the content.
4. 4Say what it's for. "I'm moving toward product roles, so the cross-functional side is the part that's relevant." Context lets a writer aim.
5. 5Give a genuine exit. "If it's not a comfortable yes, please just say so — it won't be awkward." People decline gracefully far more often than they write reluctantly, and a reluctant recommendation is worse than none.

One more piece of etiquette that pays: ask while the work is warm. A request that arrives eighteen months after you left forces the writer to reconstruct a period they have stopped thinking about. Two to eight weeks after a project lands, or in the fortnight after you leave a role, is when the detail is still retrievable.

How to write one when someone asks you

You give a recommendation from the introduction section at the top of a 1st-degree connection's profile, via the More button and Recommend. LinkedIn asks you to set the relationship and the person's position at the time, then gives you the message field. LinkedIn states you can edit or delete recommendations you have sent at any time afterwards.

The drafting method that works fastest is to write the middle first. Skip the opening line entirely, and start with the single most specific thing you remember them doing. Once that paragraph exists, the framing sentence at the top writes itself in about thirty seconds, because you now know what you are framing.

Then run three checks before you submit:

• The swap test. Replace the person's name with a colleague's. If the recommendation is still true, it is generic — go back and add something only this person did.
• The first-line test. LinkedIn collapses longer recommendations behind a "see more" link, so whatever is in your opening sentence is what most people will read. If your first line is "It was a pleasure working with Sam," you have spent your most valuable real estate on nothing.
• The honesty test. Is there anything in here you would not say out loud to the person's face, or anything you have inflated? Inflation is detectable, and it costs you as the named writer more than it costs them.

If you are asked and you cannot write a good one — because you did not see enough of their work, or because you would have to strain — say so directly and offer the alternative you can honestly give. "I only overlapped with you on the one sprint, so anything I wrote would be thin. Happy to take a reference call instead" is a better outcome for everybody than 90 words of nothing.

How many is enough, how recency reads, and the reciprocity trap

There is no published threshold and anyone quoting you a precise multiplier is inventing it. What is observable is how a reader behaves: they scan the writers' names and titles first, read one or two in full, and stop. That behaviour tells you what to optimise for, and it is not volume.

A workable target is three to six visible recommendations, spread across relationship types. Three well-chosen ones — one manager, one peer, one client or report — outperform twelve, because twelve invites the suspicion that they were farmed. Coverage across vantage points is what a reader is unconsciously checking: if every recommendation you have is from a peer, the absence of a manager is itself a signal.

Recency works on a soft gradient rather than a cliff. A recommendation attached to a role you held in 2019 does not become false, but it stops describing the professional you are now, and a profile whose newest recommendation is six years old reads as a profile that stopped being maintained. If you are actively job searching, having at least one from the last 18 months matters more than the total count. That is the same maintenance logic behind our guide to how often to update your LinkedIn photo: a stale profile element is not just neutral, it dates the whole page.

The reciprocity trap

The most common way to devalue a recommendation is to trade for it. Two people writing for each other on the same afternoon produces two recommendations that a reader discounts to roughly zero — and because recommendations you have given are visible under the Given tab on your profile, the swap is not hidden. Anyone who cares enough to check can see it in about four seconds.

This is not an argument against ever recommending someone who has recommended you. Real working relationships are genuinely mutual, and mutual recommendations between people who worked closely for three years are perfectly credible. What kills it is the timing and the symmetry: same day, same length, same structure, each praising the other in the same terms. If you want to reciprocate, do it, and then wait a few weeks and write something that stands on its own detail. The tell is not mutuality — it is matched-set writing that could have been generated from a single template.

The related trap is asking too broadly. A request blast to fifteen connections produces four thin recommendations from people who felt obliged, which is a worse profile than three strong ones. Ask the small number of people who actually have something to say, and accept that some will decline.

It is worth thinking about the whole credibility layer together. Recommendations are what other people say about you; your headline and About section are what you say about yourself. They need to agree without echoing — our guide to writing a personal brand statement covers the self-authored half, and the two work best when a recommendation supplies concrete evidence for a claim your own copy makes in the abstract. If you are job searching, it also pays to think about how the recommendations sit next to your availability signals; we cover the trade-offs in Open to Work: green frame or recruiters only.

What to do when a recommendation lands wrong

Sometimes what arrives is inaccurate, oddly worded, or praises the wrong thing entirely. LinkedIn's controls here are specific, and knowing them prevents an awkward conversation. The critical fact: you cannot edit the text of a recommendation you receive. Only the person who wrote it can change the words.

• Accept it. Pending recommendations reach you by LinkedIn message and email; choosing Add to profile publishes it, and LinkedIn notes that an update goes out to your network when you accept.
• Ask for a revision. From the pending recommendation's More menu you can request a revision through LinkedIn messaging. The writer edits, and LinkedIn is explicit that a revised recommendation has to be reviewed by you again before it appears on your profile.
• Dismiss it. Also under More. Note the one-way door: LinkedIn states that once you have dismissed a recommendation you received, it can't be retrieved, and the sender has to delete it before you can request a new one. Ask for a revision before you dismiss.
• Hide it after the fact. Open the Recommendations section of your profile, click Edit, and on the Received tab toggle the recommendation on or off. Hiding is reversible; dismissing is not.

On the giving side you have separate controls. Recommendations you have written appear under the Given tab with a Visibility button offering All LinkedIn members, 1st-degree connections only, or Only you — and LinkedIn is clear that choosing to hide or unhide it there does not affect how it displays on the recipient's profile. You can also edit or delete anything you have given at any time. Deleting one removes it from their profile, and LinkedIn notes the recipient is not notified when you do — which is a reason to say something rather than let it quietly vanish.

The practical sequencing for a recommendation you are unhappy with: leave it pending, message the person warmly, name the one specific thing you would love added or corrected, and use Ask for a revision. Most people are relieved to be told — they wrote it in ten minutes and would rather it be right.

The profile a recommendation sends people to

A recommendation is a referral mechanism. Someone reads a colleague's name attached to a specific, credible paragraph about you, and the next thing they do is look up. They land on your photo, your headline and your banner — and a profile that looks half-finished undercuts the paragraph that sent them there. This is the only place in this article where the visual layer matters, and it matters as a floor, not a ceiling.

We keep the photo guidance in dedicated articles rather than repeating it here. Start with our complete guide to LinkedIn profile pictures for what to aim at, and read the 12 LinkedIn photo mistakes that cost interviews for the specific failures to rule out — the first is the positive case, the second is the diagnostic checklist. If your banner is still the default blue gradient, what to actually put in your LinkedIn cover photo covers the one strip of the profile most people leave empty.

None of that substitutes for the recommendation itself. A polished photo attached to a profile with no written credibility is a well-lit blank. The point is only that the two are read within a few seconds of each other, and the weaker one sets the impression.

Then go back to the part that actually does the work. Pick three people who saw you do something specific, write each of them a real request with the dates and the numbers attached, and give one unprompted recommendation this week to someone who deserves it. That last one is not altruism: writing a good recommendation is the fastest way to learn what a good recommendation looks like.

The short version

If you take four things from this: recommendations only work when they describe something that happened; an admission of a weakness that was overcome is the strongest credibility signal available; a specific ask with talking points gets written in ten minutes while a blank request never gets written at all; and an obvious swap devalues both sides. Everything else is mechanics, and the mechanics are three clicks.