# LinkedIn Hiring Assistant 2 Messages the Same 0.1%. Source the Rest

_LinkedIn Hiring Assistant 2 ranks connected, verified profiles first. Here is what it misses and a workflow to find strong engineers through commits and repos._

- Canonical URL: https://www.refolk.ai/blog/linkedin-hiring-assistant-2-source-unconnected-engineers
- Published: 2026-11-13
- Reading time: 10 min
- Keywords: LinkedIn Hiring Assistant 2, LinkedIn connected apps recruiters, LinkedIn Hiring Assistant 2 review, source engineers outside LinkedIn, GitHub sourcing passive candidates

Every Recruiter seat in your market is about to get a very similar shortlist of engineers. LinkedIn Hiring Assistant 2 rolls out from November, free and automatic, to existing Hiring Assistant customers working in English. It ranks candidates by who you've hired before, how receptive they look and whether their profile is verified and connected. The engineers that shortlist leaves out are where your next good hire is.

## What does LinkedIn Hiring Assistant 2 change?

Hiring Assistant 2 (HA2) adds recruiter-level memory and receptiveness scoring, and it gives more weight to verification and connected-app evidence. From November it ships automatically, at no extra cost, to existing Hiring Assistant customers who use the product in English. LinkedIn announced it at Talent Connect in New York.

Here is what changes:

- **Recruiter-level memory.** HA2 learns from your past hires. LinkedIn's own example has it finding a product manager "close to the office, in the retail industry and with mid-level experience, because those are all the things you've hired for in the past."
- **Receptiveness indicators.** Each candidate gets indicators of how receptive they might be, built from platform engagement, Open to Work details and InMail responsiveness.
- **Verification and connected apps.** Profiles with identity verification and connected-app data count for more.
- **Batch outreach.** The agent sends to up to 25 candidates at a time. Each message is labelled as sent by the recruiter's Hiring Assistant and costs an InMail credit.
- **ATS reach.** At launch HA2 plugs into Greenhouse, Lever, SmartRecruiters, Zoho Recruit and Tracker, and it ranks your old applicants with the same signals.

There are two hard limits. HA2 only reaches LinkedIn members and your ATS, never the open web. And instant outreach is turned off for EU-located projects.

This isn't a pilot. Agentic hiring has passed $450M in annualized revenue, so a large number of seats will be running the same agent on the same profiles.

## What does HA2 actually rank on?

For sourcing, HA2 leans on four inputs: memory of past hires, receptiveness, verification and connected apps. None of them measures the code an engineer has shipped. A connected app is a third-party tool linked to a LinkedIn profile that sends usage or proficiency data back to it.

| Signal | What it actually measures | Who it ranks low |
|---|---|---|
| Memory of past hires | Similarity to people you already hired | Career changers, people from non-target employers |
| Receptiveness | Engagement, Open to Work, InMail replies | Busy senior engineers who ignore InMail |
| Verification | Identity checked by LinkedIn | Members who never verified |
| Connected apps | Replit or JetBrains usage data | Engineers whose work sits in private repos |

One correction: GitHub wasn't a live connected app at launch. In June 2026, 19 apps were live and 10 more were announced as coming, and GitHub was in the "coming" group. By October the Connected Apps menu listed 26 partners. Check the menu before you assume any candidate's GitHub feeds into their ranking.

The developer apps that were live are Replit and JetBrains. JetBrains tracks IDE feature usage locally and sends a proficiency level to LinkedIn. That tells you how someone uses an editor. It doesn't tell you what they shipped.

## Why is LinkedIn's 2.9x connected-app figure a feedback loop?

The 2.9x figure describes who already gets messaged, not who is better, and HA2 will widen that gap. LinkedIn says profiles with a connected app receive up to 2.9x as many recruiter messages, and candidates the agent surfaces are 4x as likely to be contacted as those found through traditional sourcing.

The mechanism is simple. People who connect an app are already active, visible and curious about new jobs, and recruiter messages follow activity. HA2 then ranks those same people higher, and recruiters message whoever the agent shows them. Part of the 4x is the agent grading its own homework.

Now look at how small that group is. Over one million members had connected apps as of June 17. Against LinkedIn's 1B+ members, that's roughly one in a thousand.

> **~0.1%** Share of LinkedIn members with a connected app
>
> Over 1M connected members against 1B+ total, as of June 17 (derived).

Point every competitor's agent at that same thin slice and the result is predictable: those inboxes fill up with agent-labelled InMail. Your carefully written note becomes one more near-identical message in the pile.

## Why does "no connected app" quietly count against an engineer?

These signals reward presence, so an engineer with no connected app, a thin profile and no InMail history doesn't get a neutral score. They score low on every input. Missing data reads as a weak candidate.

Think about the engineer you'd most like to hire. They rarely log in to LinkedIn, never answer InMail and haven't touched Open to Work. Their best work lives in their employer's private repos. HA2's receptiveness model sees almost nothing, so it puts them near the bottom.

That isn't an edge case. It's where most work on GitHub happens.

> **81.5%** Share of GitHub contributions made in private repositories in 2025
>
> Even though 63% of all repositories on GitHub are public.

LinkedIn CEO Dan Shapero describes the goal as "new ways for members to show real, credible proof of what they're capable of." The demand behind that is real: 39% of talent acquisition professionals say their top challenge is not knowing whether a candidate or their skills are real. Verification answers that question well. But proof that someone is real isn't proof that they're good. For engineers whose code never leaves the company, no connected app will ever show their best work.

> HA2 can verify that an engineer is real. It cannot see the 81.5% of work done in private repos.

## How many US engineers advertise their code on LinkedIn?

Almost none. In Refolk's index of professional profiles, only 675 of 562,804 US software engineers (0.12%) mention GitHub in their headline. If you search LinkedIn to find people with strong code portfolios, you start from a near-empty set.

| Segment (US; SWE, Senior SWE, Staff SWE titles) | Count | Share of base |
|---|---|---|
| All matching profiles | 562,804 | 100% |
| Headline mentions "GitHub" | 675 | 0.12% |
| Headline mentions "open source" | 385 | 0.068% |
| Engineers per "GitHub" headline | ~834:1 | derived |
| "GitHub" sample who work at GitHub itself | 11 of 25 | 44% (sample) |

It's worse than it looks. In a 25-profile sample of the "GitHub" headline group, 11 people work at GitHub. A keyword search for "GitHub" mostly finds GitHub's own staff, not engineers with great portfolios.

The "open source" group is more useful as a seed list. Its 385 profiles include people at OpenTelemetry, Red Hat, Guardrails AI, Uber and Rippling, so they point to projects whose contributor graphs you can walk.

The lesson is about direction. Start from code and work towards the person. Going from profile to code fails because 833 of every 834 engineers never put the link there.

## How does HA2's memory narrow your pipeline over time?

Memory makes HA2 favour the companies, regions and profile shapes you've hired from before, so anyone outside that pattern gets ranked down a little more with each req. You won't see it happen. The shortlist just looks familiar.

LinkedIn's product manager example shows how it works. If your past hires came from retail, lived near the office and were mid-level, HA2 assumes the next one should too. Three groups slide down the list:

- **Career changers.** A firmware engineer moving into backend has no "past hire" that looks like them.
- **Engineers from non-target employers.** Someone excellent at a small company your team has never heard of doesn't match.
- **People who skipped the usual ladder.** Maintainers, contractors and founders whose titles don't map to your previous offers.

This compounds across your ATS. HA2 also ranks old applicants in Greenhouse or Lever with the same signals, so candidates you passed on can get pushed down again for not resembling the people you did hire. If you want range, you have to source outside the pattern on purpose. The agent won't do it for you.

> **Senior engineering roles open now** - [see the open roles](https://www.refolk.ai/jobs/search?family=engineering&seniority=senior)
>
> Live postings from Refolk's job board, refreshed as you read.

## How do you source the engineers HA2 ranks low?

Start from code and work out to the person. Pick the repos your role depends on, pull the people who commit to and maintain them, check their work in the commits themselves, then find a contact route and send a message that cites what they built.

1. **List the dependencies your team actually runs.** Ask the hiring manager for the five libraries or services they'd most want a new hire to know deeply. Each one becomes a search space.
2. **Pull commit authors, maintainers and issue openers from the last 12 months.** Issue openers are underrated. They're often engineers at companies running the tool in production, which is exactly the private-repo crowd HA2 can't see.
3. **Seed from open-source employers.** OpenTelemetry and Red Hat show up in the "open source" headline group. Their maintainer lists and contributor graphs are a ready-made starting pool.
4. **Read commits, not stars.** One substantive fix to a hard subsystem tells you more than a popular tutorial repo. Look at review comments too, because they show judgment.
5. **Match the identity.** Link the commit author to a LinkedIn profile, a personal site, a conference talk or a public email. This is the slowest step by hand, and it's the one GitHub sourcing for passive candidates usually stalls on.
6. **Message outside the batch.** Send through email or the open web, not as the 26th agent-labelled InMail.
7. **Re-read your ATS rejects.** Don't let HA2 rerank them. Filter old applicants by stack and read them yourself.

Steps 2 through 5 are where the hours go. Refolk collapses them into one request: you describe the person in plain English and get back a ranked shortlist with the commits and profiles that qualified each name.

> **Example search:** "Find US backend engineers who committed to the OpenTelemetry Collector in the past year and don't mention GitHub in their LinkedIn headline." - [run it on Refolk](https://www.refolk.ai/start?q=Find%20US%20backend%20engineers%20who%20committed%20to%20the%20OpenTelemetry%20Collector%20in%20the%20past%20year%20and%20don't%20mention%20GitHub%20in%20their%20LinkedIn%20headline.)
>
> A ranked shortlist of contributors matched across GitHub, LinkedIn and the open web, each with the commits that put them on it.

### What the message should say

A hand-researched message has one advantage over agent outreach: it can prove someone read the work. Use it.

- **Name the commit or issue.** "Your fix to the batching processor's memory leak" beats "your impressive background."
- **Connect it to the job.** Say which problem on the team that work maps to.
- **Say why you're writing outside LinkedIn.** One line is enough, and it explains why your note doesn't look like everyone else's.
- **Keep it short.** Four sentences and a clear question.

The EU limit helps you here. HA2's instant outreach is disabled for EU-located projects, so an EU-based engineer who gets a specific, hand-written note isn't competing with the agent's instant outreach.

Done properly, this takes real time per req. For a Staff role in a narrow stack, the code-first workflow can still mean a long stretch of reading commits before a single intro, and that's often the point where teams start pricing an agency.

## Where does HA2 still earn its seat?

HA2 is good at the visible, active, verified group, so use it there and spend your human hours on everyone else. With 115 million members now verified, the agent's slice is large in absolute terms, and the 4x contact rate holds for the people it can see.

Split the work:

- **Let HA2 handle the active market.** Open to Work candidates, recent job-changers, people who answer InMail. That's high volume and low effort.
- **Keep code-first sourcing for the roles that matter most.** Staff and principal hires, niche infrastructure, anything where the best candidates are quiet.
- **Watch your diversity of source.** If every finalist this quarter came through the agent, memory has already narrowed your pipeline.

The verdict on LinkedIn Hiring Assistant 2 comes down to one point. The tool makes the crowded part of the market more efficient. It does nothing for the 99.9% of members with no connected app, or the 81.5% of GitHub contributions made in private repos. That's where your competitors' agents aren't looking.

## FAQ

### Is LinkedIn Hiring Assistant 2 free?

Yes, for existing Hiring Assistant customers using the product in English. It rolls out automatically from November at no added cost. Outreach isn't free, though: every message the agent sends uses an InMail credit, in batches of up to 25 candidates.

### Is GitHub a LinkedIn connected app?

Not at launch. In June 2026, GitHub was among the 10 apps announced as coming, while 19 were live, including the developer tools Replit and JetBrains. By October the Connected Apps menu listed 26 partners. Check the current menu before telling a candidate their GitHub activity affects their ranking.

### Can Hiring Assistant 2 search GitHub or the open web?

No. HA2 only reaches LinkedIn members and the candidates in your connected ATS, such as Greenhouse or Lever. To source engineers outside LinkedIn, you need to start from repositories, commit history and public sites, then match people back to a profile or email.

### Why not just search LinkedIn for "GitHub"?

Because almost nobody puts it there. In Refolk's index, only 675 of 562,804 US software engineers mention GitHub in their headline, and in a 25-profile sample, 11 of them worked at GitHub itself. Search the code first, then find the person.

---

_From the Refolk blog. More posts: https://www.refolk.ai/blog/index.md_