LinkedIn's 2.9x InMail Badge Is a Selection Effect. Source the Unbadged

LinkedIn says members who connect an app like Replit or a JetBrains IDE get up to 2.9x more recruiter messages, and it is expanding that program at the same moment it ships Hiring Assistant 2, an agent that also scores candidates on how active they are on LinkedIn. If you hire engineers, that number should make you suspicious. Mostly it tells you who chose to show off for recruiters, and it pushes every agent-assisted Recruiter seat toward the same small group of candidates who already get plenty of messages.

What LinkedIn actually shipped with Hiring Assistant 2

Hiring Assistant 2 is the next generation of LinkedIn's AI recruiting agent, announced at its Talent Connect event in New York, alongside a wider Connected Apps program. Connected Apps is the opt-in program that lets a member link a third-party tool to their profile so LinkedIn can display a validated capability from that tool.

The program went from four launch partners to 26 catalog entries in about seven months:

• January 2026: It launched as "Verified AI Skills" with Descript, Lovable, Replit and Relay.app. TechCrunch framed it as a way to show off "vibe-coding chops" and noted that LinkedIn didn't say what a user had to do to count as proficient.
• August 7, 2026: The catalog had 26 entries, including Buffer, Descript, HubSpot, beehiiv, Replit Agent, HeyGen, Fiverr Freelancers and nine JetBrains IDEs.
• Hiring Assistant 2 announcement: LinkedIn expanded the program to partners including Chess.com, ElevenLabs and Medium, made Behance and Replit work displayable, and claimed up to 2.9x more recruiter messages for members with a connected app.

The catalog isn't only vibe-coding toys. Nine JetBrains IDEs is a real engineering presence. The problem for anyone doing LinkedIn connected apps recruiting is what the program measures, and who it measures.

Is the 2.9x a skill signal or a selection effect?

The 2.9x is mostly a selection effect. People who connect apps are already heavy LinkedIn users, so the badge marks LinkedIn activity at least as much as ability. As HeroHunt puts it, the candidate has to opt in, so the evidence only covers members who know the feature exists and decided to use it.

Hiring Assistant 2 adds a second layer. It also rates how receptive a candidate is, using signals that include platform engagement, Open to Work details and InMail responsiveness. Each of those rewards time spent on LinkedIn. So the agent has two separate reasons to surface the same person, and each one feeds the other:

1. An active member connects an app.
2. The badge and their engagement push them up the agent's ranking.
3. They get more recruiter messages.
4. They reply, so their InMail responsiveness score goes up.
5. Next time, they rank higher still.

None of those five steps checks whether the person writes code that holds up in production. The loop rewards being visible on LinkedIn, then counts the extra messages as proof the signal works.

2.9x

What does a "verified" badge actually verify?

A connected-app badge shows how someone uses a tool. It doesn't show what they shipped. A JetBrains proficiency summary can tell you a person spends real time in a JetBrains IDE. It can't tell you whether their pull requests got merged, whether they own a service on call, or whether anyone trusted their code enough to depend on it.

"Verified" is also weaker than it sounds, for three reasons:

• Undefined thresholds. At launch, LinkedIn didn't publish what counts as proficient (TechCrunch).
• Stale badges. Descript removes the capability when a member disconnects. JetBrains may leave the existing badge up after proficiency updates stop. So a badge can describe habits the person no longer has.
• Usage over output. Replit and Lovable summaries measure activity inside a product built for fast prototyping. That's a real skill, just a different one from maintaining a production system.

The problem the badges are supposed to fix is real. LinkedIn and Censuswide found that 39% of talent acquisition professionals say their top challenge is not knowing whether a candidate or their skills are real. Gartner director analyst Deepak Seth told Computerworld that showing an app or repository someone built "should be valuable." I agree, and the important word there is built. The repository is the evidence. The badge is a summary that someone else wrote about it.

A badge proves someone opened the tool. A merged pull request proves someone else trusted the code.

Where do production engineers' signals actually live?

Mostly outside LinkedIn: in commit history, maintainer roles, issue threads and conference talks. Engineers barely describe that work in their headlines. In Refolk's index of professional profiles, 353,826 people in the US hold a "Software Engineer" title. Only 205 of them have "open source" in their headline. That's 0.058%, about 1 in 1,726.

My read is that for most engineers, a LinkedIn headline is just a job title, filled in once and rarely touched. People who maintain real infrastructure put their effort into the repo, not the profile. Look at the 205 who do mention open source and you find current employers like OpenTelemetry, Docker, Zilliz, Cloudera, NetApp and Google, with titles like "Contributor" and "Tech Fellow." Those are exactly the organizations where the proof is on GitHub, and the headline is almost an afterthought.

1 in 1,726

Here is the scale problem in one table:

The GitHub figure is about 79x the US "Software Engineer" count in the index. They aren't like-for-like (GitHub counts accounts, including students and hobbyists), but the gap shows how much engineering activity never gets summarized on a LinkedIn profile. One more caveat: a headline search in the index for "Lovable Replit vibe coding" returned 0 profiles, but that query required all three terms to match, so it doesn't prove nobody lists them.

Why is the unmessaged engineer the cheaper reply?

The engineers nobody is messaging are the cheapest place to get a reply, because the badged group is about to get crowded. Each LinkedIn Recruiter seat can send up to 150 InMails a month. If every agent-assisted seat ranks the same 2.9x cohort near the top, those inboxes fill up, each message looks like all the others, and response rates are likely to fall for everyone sending them.

The unbadged engineer is in a different position. A staff engineer maintaining OpenTelemetry code probably hasn't connected anything, rarely updates their headline and isn't flagged as receptive. So they get fewer pitches, and one specific message about the code they actually maintain stands out instead of disappearing into a pile. Writing that message takes more work up front, but you're likely to burn fewer of your 150 InMails per reply.

The hard part is finding these people, because keyword search on LinkedIn and the connected-app filter both miss them. That's the gap Refolk closes. You describe the engineer in plain English, and I search GitHub, LinkedIn and the open web together, so a maintainer with a bare profile still shows up based on what they've shipped.

Would connecting GitHub to LinkedIn fix it?

No. Even if GitHub became a fully live Connected Apps partner, most production work would stay hidden, and the useful public signals still need open-web sourcing. Reports don't agree on GitHub's status yet. TNW's summary lists it as supported, while the body of the same article says Adobe Express, Adobe Firefly and GitHub "are on the way." Check before you rely on it.

The bigger problem is what GitHub can show. A secondary summary of Octoverse 2025 says 81.5% of contributions now go to private repos. Treat that figure as directional until you've checked it against the original report, but the direction makes sense. Most company code is private, so a profile toggle would mostly display the squares of a contribution graph and very little of the substance.

What is public, and worth a lot to anyone doing GitHub sourcing for passive engineers:

• Maintainer status. Being a named maintainer on a widely used project.
• Review activity. Thoughtful reviews on well-known repos, which show judgment as well as output.
• Issue threads. How someone debugs in public, explains tradeoffs and handles pushback.
• Release credits. Names in changelogs and release notes for projects your team depends on.
• Talks and writeups. Conference sessions and engineering blog posts on the open web.

GitHub now has 180M+ developers. The signal is there, but it's spread across repos and pages that no single profile filter will ever bring together.

How to source engineers outside LinkedIn's connected-apps filter

Start from the work, then go find the person. Don't start from the profile and hope the work is there. Here's a workflow that holds up:

1. List the open-source projects your stack depends on. OpenTelemetry, Docker, or whatever your services actually run on.
2. Pull contributors with depth, not volume. Look for reviewers, maintainers and people who keep showing up over several releases. Skip one-off typo fixes.
3. Check commit authors properly. A commit author name is just a string, so confirm identity through linked accounts, consistent email domains and talks or posts that point back to the same handle.
4. Find the LinkedIn profile afterwards. Use it for current employer and location, not as the reason they made your list.
5. Open with their code. Mention a specific PR or design discussion. That's the advantage over a recruiter who found them through a badge.
6. Use connected-app badges as a tiebreaker only. A JetBrains summary on someone who already has strong public work is a nice extra. It isn't a reason to put someone on the list.

Steps 1 through 4 are where most sourcers lose hours switching between GitHub search, Google and LinkedIn. With Refolk, that becomes one plain-English request, and you spend your time on step 5, which is the step that actually gets replies.

FAQ

Are LinkedIn connected apps useless for technical hiring?

No. They're a weak signal sold as a strong one. A JetBrains or Replit summary shows real tool use, and it can help you choose between two candidates you've already qualified. It doesn't show shipped, reviewed code. The 2.9x message lift also comes from a group that opted in to display itself, so don't read it as proof that badged engineers are better.

Is GitHub a LinkedIn Connected Apps partner?

Reports conflict. TNW's summary lists GitHub as supported, but its body text says GitHub is "on the way." Check your own Recruiter view before you build a process around it. Even when it's live, a secondary summary of Octoverse 2025 puts 81.5% of contributions in private repos that a profile link can't show.

Does Hiring Assistant 2 rank down candidates who don't connect apps?

I haven't seen LinkedIn say it penalizes them directly. But the agent's receptiveness rating draws on platform engagement, Open to Work details and InMail responsiveness, all of which reward heavy LinkedIn users. Engineers who rarely log in will tend to rank lower relative to active members, whatever their actual skill.