AI Recruitment Statistics: What 2,432 AI-Scored Profiles Show

A typical AI sourcing search gave a shortlist of about a dozen people who met every must-have. Across 52 Roles from 10 June to 29 September 2026, the median Role had 11.5 such profiles. Of all 2,432 profiles our AI scored, 11.5% reached 85 or above.

We build Talent Summoner, which searches public profiles and ranks each one against your must-haves. Your first Role is free with no card (pricing), so you can test these numbers on your own search.

If you hire a few times a year and can review a dozen profiles in one sitting, these numbers are for you. If you need one named person from a pool of a few dozen, skip to the section on when this approach is wrong.

AI recruitment statistics from 2,432 scored profiles

Our AI scored each profile from 0 to 100 against the brief for one Role, which is one hiring search. The bands match the labels in our app: Strong fit, Good fit, Partial fit and Weak fit.

Source: Talent Summoner production data, 10 June to 29 September 2026, external accounts only. The method is below.

What a realistic shortlist looks like

Plan for about a dozen people per Role, and expect a search to review several times that many. Of the 52 Roles run between 10 June and 29 September 2026, 43 reached at least five profiles rated 65 and above that met every must-have, and 3 reached none.

Why more candidate profiles do not mean a better shortlist explains why a long list is not the goal. The cut comes from the must-haves, so write them with our guide to must-have vs nice-to-have requirements.

Must-haves also reorder the list. Of the 885 profiles rated 65 or above in that window, 79 failed a must-have, and the app lists those below the profiles that pass every one.

How to read these numbers

The scores are our AI's, not a hiring outcome: they show how a profile matched the brief, not who would perform well, and a fit score is a prioritisation aid rather than a verdict.

The 65 line is our app's "Good fit" label, not an industry standard, so another tool's 65 means something different. We do not report how many profiles recruiters kept, because too few selections were recorded to count.

By the location text on each profile, 54.5% of the 2,432 were in Hong Kong, 4.7% in Singapore and 15.0% in North America. Early versions scored people outside the target location; since 25 September 2026 a pre-screen drops them first.

Method

• Source: Talent Summoner's production database. Every score is from our AI, not from a human reviewer.
• Counted: 2,432 scored profiles in 52 Roles, created 10 June to 29 September 2026. "Reviewed" counts every profile a Role pulled in, including ones that failed to load.
• Excluded: team and test accounts, and a batch of accounts that looked like automated sign-ups. The window includes our June beta.
• Checks: without the largest single account, no band share moved by more than 1.3 points. Profiles per Role also depend on plan limits; the free Role shows up to 50 candidates.
• Biggest limit: 96% of the profiles were scored before the 25 September 2026 pre-screen, so today's fit shares are likely higher. We re-run the same query every quarter and will update this page.

When this approach is wrong for you

An AI shortlist is a poor fit when:

• You need one named person from a pool of a few dozen; bands built on hundreds of profiles say little about so few.
• You want someone to screen candidates and sell the role for you. That is an agency's work, and our tool does not do it.
• You need proof of who will perform well; a match score is not a hiring outcome.
• You need candidates synced into an applicant tracking system today. We do not integrate with one yet.

Where Talent Summoner fits

Talent Summoner searches LinkedIn, GitHub and other public sources, over 200 million profiles. It ranks each person against your must-haves and nice-to-haves with plain-English reasoning.

You can send outreach from your own Gmail, Outlook or LinkedIn, and you review each message and confirm the send. Your first Role is free and your second is US$4, then US$29 per Role, or US$59 a month for 10 Roles if you hire often.

Trade-offs, stated plainly: no paid InMail, no always-on pipeline and no ATS integration today. It does not screen, interview or close candidates for you.

We checked these facts on 4 October 2026 and re-check them every quarter.