LinkedIn's Candidate Interest Score Will Herd Recruiters Into 199 Profiles

On September 29, 2026 at Talent Connect NYC, LinkedIn announced Hiring Assistant 2. Rollout starts in November, automatic and free for English-language Hiring Assistant customers, and the headline feature is a "Candidate Interest" indicator that scores every profile in your shortlist on how likely they are to say yes. If you source engineers for a living, you have about six weeks to understand what that score does to your pipeline.

What the Candidate Interest signal actually measures

Candidate Interest is a global receptiveness score built from three inputs LinkedIn already owns: InMail response frequency, Open to Work status, and platform engagement. It is not a fit score. It is not a hirability score. It ranks attention, nothing else.

That distinction matters because of how LinkedIn is positioning the launch. VP of Product Dan Reid framed Hiring Assistant 2 as learning "how you hire, not how everybody hires." The Candidate Interest signal does the opposite: it is a single, global ranking that every HA2 user sees on the same profile at the same time. If you and three competitors load the same shortlist, you all see the same green light next to the same person.

The second new indicator, Fit, replaces the generic match summary with a role-specific one. Combined, the two signals are LinkedIn's bid to shorten the median pre-screen from where it already is - Hiring Assistant closes screens in a median of six minutes - to something closer to one click. Agentic hiring at LinkedIn has crossed $450M in annualized revenue, so this is not optional theater. HA2 is where the budget is going.

The math of the crowding problem

The addressable "high-interest" US software engineering pool is a few hundred profiles, against hundreds of thousands of HA2 seats. That is not hyperbole. It is what the index shows.

In Refolk's index of professional profiles, there are 562,598 US-based software engineers under the SWE, Senior, and Staff titles. Of that cohort, only 199 currently signal Open to Work in their headlines or profiles. That is 0.035%, or roughly 1 in 2,826.

199

Layer on the InMail side. The 2026 platform benchmark sits between 10% and 25% reply rates, but software and SaaS is the worst-performing industry at 4.77%. Talent Acquisition itself, where recruiters InMail other recruiters, only gets to 12.08%. If HA2 defines "high interest" as roughly Open to Work plus above-median InMail replies in SaaS, the green-light pool for US engineering roles is on the order of a few hundred to low thousands of profiles.

The last row is the one to sit with. Recruiters are already 27% more likely to contact a Hiring Assistant-sourced candidate than they were six months ago. The herding effect is pre-existing. HA2 is the accelerant.

Why Candidate Interest is a Keynesian beauty contest

The score ranks who will reply, not who you should hire, which means every HA2 user converges on the same tiny set of profiles rather than the best ones. That is the structural flaw.

InMail response frequency is self-reinforcing. The engineers who historically reply to InMails get boosted in future rankings. The engineers who ignore 95% of InMails - which is to say, every staff-level builder at Oxide, Figma, Anthropic, Datadog - get buried further. The inverse of the Candidate Interest score may actually correlate with seniority in engineering, because the senior IC's rational response to recruiter spam is to stop replying.

Open to Work has the same lagging-indicator problem. The talent you want most, the Rust engineer at Oxide or the infra IC at Figma, turns Open to Work off precisely because leaving it on draws dozens of InMails a week. HA2 institutionalizes the opposite of the sourcer's actual craft, which is finding the person who has not raised their hand.

Candidate Interest ranks who will reply, not who you should hire. In software, those are almost never the same list.

The second-order effect on reply rates

Software and SaaS already has the worst InMail reply rate on the platform at 4.77%. HA2 will not fix that. It will concentrate the finite InMail credits - a Recruiter Corporate seat is $8,999 per year for 150 credits a month - on the same tiny tail of responders, which drives effective reply rates for crowded roles toward zero within one to two quarters of rollout. The 199 Open-to-Work engineers will get flooded. They will stop replying. The score will update. A new set will get flooded. And so on.

What Hiring Assistant 2 cannot see

The arbitrage is off-platform signals: GitHub contributions, conference speaker lists, Discord and Slack community activity, patent filings, and open-source maintainer status. HA2 does not index any of these, and will not within the rollout window.

Concretely, the signals that beat Candidate Interest for engineering roles include:

• GitHub contribution patterns: commit cadence, repo ownership, language mix, and who they commit alongside.
• Conference speaker lists: RustConf is the obvious one for systems roles.
• Community rolls: Rust Foundation members, GitHub Sponsors recipients, StaffEng Slack regulars.
• Hacker News "Who wants to be hired" and "Who is hiring" threads, parsed monthly.

None of these show up in a Candidate Interest score. All of them correlate with the engineers actually shipping the systems your company wants to buy or build. Describing the person you want in plain English and letting Refolk resolve the match across GitHub, LinkedIn, and the open web is the exact workflow that routes around HA2's blind spot, because the inputs it reads are the ones LinkedIn cannot see.

The ATS pipeline is about to get the same treatment

HA2 ships with launch integrations into Greenhouse, Lever, SmartRecruiters, Zoho Recruit, and Tracker, with iCIMS and Workday coming after. The Candidate Interest score will rank candidates already inside your own ATS using the same signals, extending the crowding problem to the pipeline you paid to build.

The practical consequence: a candidate who applied directly, interviewed well, and is sitting in your Greenhouse pipeline at "silver medalist" will be ranked by the same global receptiveness metric as a cold LinkedIn profile. If they stopped replying to InMails two years ago because they got a job, their score drops, and the HA2 UI will quietly deprioritize them in your view. The sourcer who re-engages silver medalists manually will outperform the sourcer who trusts the ranking.

A playbook for the November rollout

Treat the November launch as a 90-day window to rebuild your sourcing around signals HA2 does not read. Five moves worth running this quarter:

1. Build a non-LinkedIn shortlist for your top 3 reqs. Pull GitHub, conference speakers, and community membership for each. Compare it to what LinkedIn Recruiter returns for the same req. Note the overlap. It will be small.
2. Tag your ATS silver medalists manually. Before HA2 ranks them against new inbound, flag the top 50 per req as "re-engage directly." Bypass the Candidate Interest ordering.
3. Rewrite your first-touch message off-platform. An email referencing a specific pull request lands better than a generic InMail. The 4.77% ceiling is a LinkedIn ceiling, not a candidate ceiling.
4. Audit your Open to Work filter dependency. If a meaningful share of your sourced pipeline currently relies on OTW signals, you are already crowded. HA2 makes it worse.
5. Price a LinkedIn Recruiter alternative for 2026. At $8,999 per seat, Recruiter Corporate's value calculation flips the moment reply rates on contested roles drop below 2%. Alternatives that work from GitHub and the open web stop looking exotic and start looking cheaper.

Who is still invisible to HA2, by employer

The specific employers whose engineers HA2 will systematically underrank are the ones with the strongest off-platform footprint and the lowest Open-to-Work rates.

In Refolk's index, Oxide Computer Company has roughly two US-based Rust SWEs with any Open-to-Work signal at all. The 1,094 US Rust-skilled engineers in the index cluster at Oxide, Figure, Shopify, Uniswap, and Helius. Figma, Anthropic, and Datadog show up repeatedly as current employers of senior engineers who maintain active GitHub profiles but minimal LinkedIn activity. These are not edge cases. They are the modal profile for the people you actually want to hire into a staff systems role.

The sourcer's job in 2026 is to notice that the median LinkedIn recruiter will stop seeing these engineers in November, and to decide whether that is a problem or an opportunity.

FAQ

When exactly does Hiring Assistant 2 ship?

LinkedIn announced HA2 on September 29, 2026 at Talent Connect NYC, with English-language rollout beginning in November 2026. It will be automatic and free for existing Hiring Assistant customers, so you do not opt in. The Candidate Interest and Fit signals replace the generic match summary in the same release.

Does the Candidate Interest score replace Open to Work?

No. Candidate Interest incorporates Open to Work as one of its three inputs, along with InMail response frequency and platform engagement, but does not replace the standalone filter. The practical effect is that Open to Work's weight increases, because the signal now shows up in the ranking even for recruiters who were not actively filtering on it.

Is this a reason to drop LinkedIn Recruiter?

Not immediately, but it is a reason to stop treating Recruiter as your only pipeline. At $8,999 per seat per year for 150 InMail credits, Recruiter's value depends on reply rates holding up. If HA2 pushes software and SaaS reply rates below the current 4.77% floor, the per-hire economics shift fast. The sensible move is to run GitHub, conference, and open-web sourcing in parallel this quarter and measure the delta yourself.

What should I do with my existing Open-to-Work shortlists?

Treat them as the most crowded pool in your pipeline, not the least. Every HA2 user will see the same green light on the same 199 US engineers in November. Prioritize speed on those profiles if you must contact them, but assume reply rates fall sharply within 60 days and build your real pipeline from signals - GitHub, patents, community activity - that HA2 cannot read.