# The Profile Recommendation Set, From the Ask to a Placed Testimonial

*You will pick the right recommenders, send asks that get a yes, and end with three to five diverse recommendations a recruiter reads as real proof.*

- Canonical URL: https://www.refolk.ai/candidates/guides/profile-recommendation-set
- Pillar: Positioning and materials
- Format: Playbook
- Published: 2026-09-02
- Last reviewed: 2026-09-02
- Reading time: 15 min

You need a handful of strong, credible recommendations on your profile, and you do not know who to ask, what to say, or how many. This guide is for a job seeker assembling that set from scratch or rebuilding a stale one. It gives you a repeatable procedure with targets, a sequence, and copy-paste asks, so you finish with three to five diverse recommendations a recruiter reads as real proof rather than filler.

Every other part of your profile is written by you: the headline, the summary, the keywords, the skills list. The recommendations section is the only element a third party writes, which is exactly why it converts. As one practitioner puts it, it is really the only element of a profile the owner cannot manipulate. In a market where AI-generated profiles are flooding the platform, a set of specific, relationship-anchored testimonials clears a trust bar that a polished self-written section cannot.

## Why the recommendations section carries weight nothing else can

The recommendations section is the only self-proof-immune part of your profile, and that is the whole source of its value. Everything you write about yourself can be doubted; a named person vouching in their own words cannot be self-generated. That is the mechanism behind its signal.

Two data points frame the stakes. According to a LinkedIn survey, profiles with recommendations receive 14 times more profile views than those without. Separately, a controlled field experiment sending 24,570 fictitious resumes found that applications backed by a comprehensive LinkedIn profile drew a 13.5% callback rate versus 7.9% without one, a 71% lift. Neither figure isolates recommendations from the rest of a complete profile, so treat them as evidence that the section matters, not as a promise tied to a specific count.

**14x - More profile views for profiles with recommendations than without**

Attributed to a LinkedIn survey; it compares having recommendations to having none, not one count against another.

The audience reading this section is concentrated and specific. In Refolk's index of professional profiles, 112,788 US professionals hold a current title including Recruiter, Technical Recruiter, or Talent Acquisition. These are the people who read your recommendations. That number is worth internalizing, because it tells you who you are writing for and, as the next section shows, where they are.

## Who actually reads your recommendations, and where they are

Your recommendation readers are overwhelmingly recruiters and talent-acquisition staff, and they are concentrated in one market. In Refolk's index, the US recruiter-reader pool outnumbers the UK pool by roughly 15.4 times.

That concentration changes how you should think about signal. A UK job seeker is read by a far thinner recruiter pool, which raises the signal-per-reader of each recommendation: fewer eyes, so each one counts for more. A US job seeker competes for attention inside a much larger pool, where a diverse, specific set is what breaks through.

| Market | Recruiter/TA professionals | Share of the two-market total |
|---|---|---|
| United States | 112,788 | 93.9% |
| United Kingdom | 7,337 | 6.1% |

Column source: Refolk's index, one query per country for titles including Recruiter, Technical Recruiter, or Talent Acquisition. A seniority split is not available, because Director/VP and Senior cuts on the same title both returned zero in the index.

The employers behind these readers tell you what language travels. In the US sample, top employers of recruiter-readers include Experis, K2 Partnering Solutions, Robert Half, and Forrester. In the UK, they include Oceaneering, Utility Warehouse, Bloomberg, Wayfair, and Simpplr. Staffing firms and in-house teams dominate both lists, which means client and cross-functional recommendations that mirror agency-and-employer language, scope, timelines, and delivered outcomes, will read as relevant to the people screening you.

## How many recommendations, and why the number is a trap

Aim for three to five, but do not treat any number as a threshold that unlocks results. The count everyone cites traces to LinkedIn's own completeness meter, not to a study of what recruiters respond to.

The mechanics matter here. LinkedIn suggests three recommendations to contribute to a 100% complete profile, which is why three shows up everywhere as the magic number. But three is a platform-manufactured Schelling point, not an evidence-based optimum. The widely repeated claim that view lift begins at three and plateaus around five to seven is not traceable to LinkedIn and should be treated as unverified. There is no cap on how many you can hold, and one 2012 study of top-ranking profiles found some had none at all, so do not sell yourself a numeric guarantee.

| Source type | Stated target |
|---|---|
| Platform completeness rule | 3 |
| Profile-writer consensus | 3-5 |
| Optimization guide | 5-10 |
| Connection-scaled rule | 1-2 per 50 connections |

Column source: each cited page in the dossier. The targets disagree, which is the point: no single number is authoritative.

> **Rule:** Anchor on diversity, not the count
>
> The count-based view-lift claim is unverified, so do not chase a number. Three specific recommendations spanning a manager, a peer, a report, and a client beat ten thin ones that all sound alike.

So three to five is the working target because it is achievable, it clears the completeness meter, and it leaves room for the source diversity that carries the real signal. Beyond five, marginal recommendations rarely add proof and start to invite the batched-date and reciprocal-pair problems covered later.

## Who to ask, and how to spread the sources

Pick recommenders for what they can speak to, not for who likes you most. The single most valuable recommendation comes from a direct manager, because that person has the greatest visibility into your contributions. But diversity of source matters more than any one type, so build across four relationships.

The four relationship types, and what each proves:

- **Direct manager.** Proves performance and judgment from the person who set your goals and saw the results. This is your anchor; get at least one.
- **Cross-functional peer.** Proves how you work across teams, in the language of product, design, sales, or operations partners.
- **Direct report.** Proves leadership and how you develop people. Handle with care: a report faces bias and pressure, so it reads best when specific and clearly voluntary.
- **Client or customer.** Proves delivered outcomes in the buyer's own words, which mirrors the agency-and-employer framing recruiters recognize.

What a good source map looks like when it lies to you: a set that appears full but is actually all from one role or all peers. That is lopsided. A recruiter reads a cluster from a single job as a favor traded within one team, not as a broad track record. Audit for spread and fill the under-represented relationships first.

#### The recommendation set, from most self-proof-immune outward

1. **Direct manager** - Performance and judgment from the person who set your goals
2. **Cross-functional peer** - How you work across teams, in a partner's language
3. **Client or customer** - Delivered outcomes in the buyer's own words
4. **Direct report** - Leadership and how you develop people, handled with care

*The direct manager anchors the set, and each outer layer widens the kinds of proof a recruiter can read.*

Finding the right people is the part that stalls most job seekers, because former managers move companies and cross-functional partners drift out of your immediate network. This is where a search tool earns its place: instead of scrolling your connection list from memory, you can ask directly for the people who fit each relationship slot.

Ask me this: `Find cross-functional partners in product and design who worked on a named launch project with me.` - [run the search](https://www.refolk.ai/start?q=Find%20cross-functional%20partners%20in%20product%20and%20design%20who%20worked%20on%20a%20named%20launch%20project%20with%20me.).

*Returns partners you collaborated with, so you can fill the cross-functional slot with someone who can speak to a specific project rather than a generic peer.*

## The full procedure, from ask to placed testimonial

Run the set as one sequence rather than a scatter of one-off asks. The whole thing takes two to four weeks in calendar time but only a few hours of your own effort, most of it spent making the ask easy for each recommender.

#### From shortlist to a placed, ordered set

1. **Set the target and map recommenders** - Decide on three to five and build a shortlist spanning a manager, a cross-functional peer, a direct report, and a client, with the specific role and project each can speak to. Takes about 30 minutes.
2. **Warm up before the formal ask** - Message each person in plain language first and get a yes-in-principle; never fire the blank request box. Takes about 5 minutes each.
3. **Send the platform request with context** - Use the Recommendations section on your profile or the More menu on theirs, set the relationship and role, and add a personalized note. Takes about 30 seconds each.
4. **Make it easy with bullets or a draft** - Give two or three specific bullets you want highlighted, or offer a full draft they can edit. Takes about 10 minutes each.
5. **Space the requests so dates look organic** - Stagger the asks across two to four weeks so the given dates do not cluster in one week.
6. **Follow up once, kindly** - Wait five to seven business days, then send a single nudge with a ready draft attached.
7. **Review, accept, and order** - Accept each recommendation, curate for diversity and strength, and order the strongest manager-first. Takes about 15 minutes.
8. **Reciprocate on a delay** - If you owe a recommendation back, wait at least a few months before posting it, so no simultaneous mutual pair appears.

### The exact request mechanics

You must be a first-degree connection to request a recommendation. There are two entry points. On your own profile, scroll to the Recommendations section and click Ask for a recommendation, then choose the role and identify the person from your connections. On their profile, click the More button in their intro section, the three dots, and select Request a recommendation. Either way you set your relationship to them and what their role was at the time. The mechanics take about 30 seconds.

You can add up to three people to a single request, but do not: customize each request individually so it reads as personal. Requests only fire in your primary profile language, so if yours is set to a secondary language the flow will silently fail.

> **Tip:** Do the writing for them
>
> The easier the ask, the faster the yes. Supplying two or three bullets or a full draft is the single strongest yes-driver, because it removes the writing burden that stalls busy people. A ready draft is also what recovers most non-responders on the one follow-up.

## The ask that gets a yes, word for word

The warm-up message is what earns the yes; the platform request is just the mechanism. Send a real message first, in your own voice, that names the specific work and offers to do the writing. Never send the default blank box.

**Warm-up message before the formal request**

```
Hi [name] - I'm updating my profile for a search and I'd value a recommendation from you, since you saw my work on [specific project or period] up close. No pressure at all, and I'll make it easy: I can send two or three bullets on what I'd love you to speak to, or a full draft you're free to rewrite however you like. Would that be alright? If so I'll fire over the LinkedIn request and the bullets today.
```

*Send this as a normal message first. Replace the bracketed parts with real specifics, then send the platform request only after they say yes.*

When they agree, send the bullets in the same thread. Give them scope and scale, because that is what makes a recommendation read as authentic rather than generic.

**The bullets you hand a recommender**

```
- We worked together on [project] from [month/year] to [month/year], where I [specific role].
- The result worth naming: [outcome with a number, e.g. cut onboarding time from 6 weeks to 3].
- The skill I'd like to stand out: [specific strength], for example when I [concrete moment].
```

*Give two or three. Each should carry a number and a specific outcome, so their recommendation cannot be about anyone else.*

Keep the target length in mind so you can tell a recommender what good looks like without micromanaging. A recommendation should be short and focused, typically three to five sentences and around 50 to 100 words, with a hard cap under 200. That is long enough to feel specific, short enough that people read it.

| Attribute | Benchmark |
|---|---|
| Length | 50-100 words / 3-5 sentences |
| Hard cap | under 200 words |
| Follow-up wait | 5-7 business days |
| Request time | ~30 seconds |

Column source: each cited page in the dossier.

> A recommendation that could describe anyone is worth nothing; the words must not be applicable to any other person.

## How this goes wrong, and the false positives to watch

Most failed recommendation sets fail in predictable ways, and several of them look fine on the surface. The dangerous ones are the false positives: a set that reads warm and complete to you but reads as coordinated or hollow to a recruiter.

- **Batched dates.** All recommendations timestamped the same week reads as coordinated. Before a job search, check the given dates and stagger any future asks.
- **Reciprocal pairs.** Two people recommending each other in the same window looks mutual and warm to you but reads as a trade to a recruiter. If you and someone agree to post at the same time, good recruiters spot it. Delay any reciprocation by months.
- **Generic or AI text.** A recommendation that could describe anyone adds nothing. The test: it should not be applicable to any other person. If it names no project, number, or moment, send it back with bullets.
- **Endorsement confusion.** Piling up one-click skill endorsements is not the same as written recommendations and carries little weight; recruiters largely disregard endorsements as a quality signal. Confirm you are collecting the written kind.
- **The blank-box ask.** Sending the default request with no context lowers your yes-rate. Warm up first and supply a draft.
- **Wrong-language request.** Requests only fire in your primary profile language; if yours is set to a secondary language the flow silently fails and no request arrives.
- **All from one role.** A cluster from a single job or all peers looks lopsided. Audit for source diversity and fill the under-represented relationships.
- **Chasing volume.** Ten thin recommendations underperform three specific ones, and the count-lift claim is unverified, so do not treat volume as a substitute for specificity.

> **Watch out:** The two coordination signals cost more than they earn
>
> Batched dates and reciprocal pairs are both flagged by recruiters. In an AI-generated-profile era the organic-looking, staggered, specific set is the scarce trust signal, so a coordinated set actively works against you even when every individual recommendation is glowing.

The judgment call most people get wrong is which recommendations to keep and which to hide. Weigh each on two axes: how specific it is, and how well it fills a relationship slot you do not already cover.

#### Keep, hide, or request another

Horizontal axis runs from Fills a slot you already have to Fills a missing slot. Vertical axis runs from Generic, could be anyone to Specific, names real work.

| Quadrant | What it means |
| --- | --- |
| Specific but duplicates a source | Keep, but stop asking this relationship type |
| Specific and fills a gap | Keep and order it early |
| Generic and duplicates | Hide it; it dilutes the set |
| Generic but fills a gap | Send it back with bullets and re-request |

*Judge each recommendation on specificity and whether it fills a source you are missing.*

## What "done" looks like, and how to keep it current

You are done when you have three to five live recommendations that span different relationships, each names specific work, the dates do not cluster, and the set is ordered manager-first. If any of those fails, the set is not finished, however many recommendations you have.

#### Before you call the set done

- [ ] The section shows at least three accepted recommendations and no more than five.
- [ ] The set spans at least three of the four relationship types: manager, peer, report, client.
- [ ] Every recommendation names a specific project, number, or moment and could not be about anyone else.
- [ ] No two recommendations share the same or a near-identical given date.
- [ ] No reciprocal pair posted within the same window; any recommendation you owe back is delayed by months.
- [ ] The strongest, most specific recommendation, ideally a manager's, appears first.
- [ ] Your profile's primary language matches the language your requests were sent in.
- [ ] You collected written recommendations, not one-click endorsements.

Keeping the set current is a light, recurring job rather than a one-time build. Every time you finish a notable project or change roles, add one recommendation from that chapter while the work is fresh and the person still remembers the numbers. Retire the oldest generic ones as stronger, more specific ones arrive. Before each new search, re-check the given dates so an old cluster does not undercut an otherwise strong profile.

If you are running several applications at once, the same relationship-mapping work that fills your recommendation set also feeds the rest of your materials. [Refolk](/candidates) writes your resume from your own history, tailors it to each posting, drafts the cover letter, and scores how well you fit, so the specifics you gather for a recommender, the projects, numbers, and outcomes, do double duty across your whole search. Refolk also helps you find the former managers and cross-functional partners who left your network, so the shortlist in step one is a search rather than a memory exercise. The recommendation set is the one profile element you cannot write yourself, and treating it as a sequenced job, not a favor you hope for, is what turns it into proof.

## Frequently asked questions

### How many recommendations should my profile have?

Three to five quality recommendations is the practitioner target for a strong profile, and three is the count that marks the section complete on LinkedIn's own meter. Alternate targets exist, from five to ten across recent roles to one or two per fifty connections, but the claim that view lift begins at a specific count is not traceable to a primary source. Anchor on diversity of source, not the number, because ten thin recommendations underperform three specific ones.

### Who should I ask for a recommendation?

Start with a direct manager, who usually has the greatest visibility into your work, then add a cross-functional peer, a direct report, and a client or customer. The best set spans different relationships and roles rather than clustering in one job. Recruiters read a lopsided cluster from a single team as weak, so audit for source diversity and fill the under-represented relationships first.

### What is the best LinkedIn recommendation request template?

A good request warms the person up in plain language, names the specific project or result you want them to speak to, and offers two or three bullets or a full draft they can edit. Never send the default blank box. Supplying the raw material is the single strongest driver of a fast yes, because it removes the writing burden that stalls busy recommenders.

### How long should a recommendation be?

Aim for 50 to 100 words, or three to five sentences, with a hard cap around 200 words. That is long enough to feel authentic and specific but short enough that people actually read it. Structure it as relationship and duration, one or two concrete achievements with numbers, standout skills, and a clear close.

### Is it bad to recommend someone who recommended me?

Reciprocating simultaneously is a recruiter red flag, because two people recommending each other in the same window reads as a trade rather than genuine praise. If you owe a recommendation back, wait at least a few months before posting it. Staggered, non-mutual recommendations read as organic, which is the scarce trust signal in an era of AI-generated profiles.

### Why isn't my recommendations section showing up?

If you have no accepted recommendations, the section does not display on your profile at all. Also check that your profile's primary language matches the language you are requesting in, because the request flow only fires in your primary profile language and can silently fail otherwise. Once you accept at least one recommendation, the section appears.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/candidates/guides/profile-recommendation-set*