Automated resume screening software and automated CV screening that show the reasoning
Automated resume screening software reads each CV or profile and scores it against your requirements. RecruitingBot parses it, checks every must have, nice to have and deal breaker in a weighted rubric, quotes the evidence it found and leaves the final call to a person.
- GDPR built in
- Human approves every shortlist
- No hiring fees
Try it on a role you are hiring for
Example shortlist in about a minute, no signup needed.
The rubric decides what a good CV looks like
Keyword filters fail because a CV is not a list of keywords. Someone who "rebuilt the billing service in Go" never wrote "backend developer", and someone who lists twenty frameworks may have touched each for a week. RecruitingBot starts from a rubric instead: a short set of criteria that describe the job, split into three kinds and each carrying a weight you can change.
The rubric is drafted from your role intake together with the job description, then opened for editing before any screening runs. Keep it short. Five to eight criteria usually separate people better than twenty, because every extra line dilutes the ones that matter. If a criterion cannot be judged from a CV, such as availability or salary expectations, move it to the screening questions instead of the rubric.
| Must have | 5+ years building backend services in Python | 30% |
| Must have | Designed and ran Postgres schemas in production | 25% |
| Nice to have | Payments or billing domain | 15% |
| Nice to have | Mentored other engineers | 10% |
| Deal breaker | Needs visa sponsorship (role cannot sponsor) | stop |
| Criterion type | What it does to the score | Example for a senior product designer |
|---|---|---|
| Must have | Heavy weight, missing it pulls the score down sharply | Shipped end to end design work for a B2B product |
| Must have | Heavy weight | Ran usability research and changed the design because of it |
| Nice to have | Light weight, adds points without blocking | Built or maintained a design system |
| Nice to have | Light weight | Experience with data heavy dashboards |
| Deal breaker | Removes the candidate from the ranked list, visible in a separate tab | Cannot work in the required time zone |
Evidence quoted next to every criterion
A number on its own invites either blind trust or blanket suspicion. Each candidate in RecruitingBot gets a breakdown with one line per criterion: met, partly met, not found, and a short quote or paraphrase from the CV or profile that justifies it. "Led redesign of onboarding, cut steps from nine to four" is evidence. "Strong designer" is not, and the screening step is built to prefer the first kind.
The breakdown also makes disagreements productive. When a hiring manager thinks a candidate is underrated, they can see exactly which criterion scored low and why. Sometimes the CV really is thin on that point, which becomes a screening question. Sometimes the rubric weight is wrong, and changing it rescores everyone in the role at once, so the correction applies fairly rather than only to the person someone noticed.
How CV parsing handles messy documents
Real CVs arrive as two column PDFs, Word files with tables, image heavy portfolios and plain text pasted into an apply form. RecruitingBot extracts text and structure from each format, then maps it into work history, dates, titles, skills, education and links. Parsed fields are shown next to the original file, so if a date range or job title was read incorrectly, you see it before it affects anything.
Profiles found through sourcing go through the same analysis as uploaded CVs, which matters because a public profile is often shorter than a resume. Where a profile lacks detail on a must have, the breakdown says "not found" rather than guessing. That honesty keeps sparse profiles from being punished as failures. They are flagged as needing a question, and the outreach or screening step can ask it.
- 01 Formats. PDF and Word files, plus text submitted through the hosted apply page.
- 02 Structure. Employment periods, titles, employers, skills, education, certifications and portfolio or code links.
- 03 Gaps and overlaps. Shown as facts in the record, never scored as a negative unless your rubric explicitly says so.
Controls that keep screening fair
Screening software can repeat old hiring patterns if it is allowed to learn from who was hired before. RecruitingBot scores only against the written rubric, and the rubric is the thing you review and own. That puts the fairness question where it belongs, in the words of each criterion. Avoid criteria that describe protected characteristics or obvious proxies for them, such as graduation year standing in for age, and rewrite them toward the actual skill the job needs.
Names, photos and personal details are not scoring inputs. The evaluation looks at work done, scope, tools and outcomes. Every score change, rubric edit and approval is kept in the role history, which helps when you need to explain a process to a candidate or an auditor. On the Agency plan, the audit log records which user did what across the workspace.
People decide, the software ranks
Automated resume screening in RecruitingBot produces a ranked list and an explanation. It does not send rejections on its own based on a score, and it does not move anyone to an offer. A recruiter or hiring manager approves every shortlist, which keeps the process outside fully automated decision making in the sense of Article 22 GDPR and keeps accountability with the people who own the hire.
In practice that means a reviewer spends minutes on the top of the list instead of hours on the whole pile. They can open any lower ranked candidate, read the breakdown and pull them forward. Rejection emails go out only for people a human marked as rejected, written in your tone and sent from your connected mailbox. Data retention rules described on the GDPR compliant candidate sourcing page apply to everyone screened.
Screening is one step in a longer pipeline
A strong CV score tells you who is worth contacting, not who wants the job. That is why scoring sits between sourcing and outreach in RecruitingBot. People who pass the rubric are contacted, interested replies get knockout questions and a short skills test, and the shortlist combines all of it. The step after interest is covered on the AI candidate screening page.
Because every stage reads the same rubric, nothing has to be retyped. A must have that the CV could not confirm becomes a screening question automatically, and the answer feeds back into the candidate record. For teams that want the whole flow explained end to end, including outreach and scheduling, the AI recruiting software page walks through each part.
-
Day 1
First messageSent
Personal note on their payments work, role facts, salary range, opt out link
-
Day 4
Follow upSent
Short nudge with one new detail about the team
-
Day 9
Last noteScheduled
Polite close, door left open
Volume limits and what they mean for screening
Every plan includes a monthly number of candidates analyzed: 1,500 on Starter, 5,000 on Growth and 12,000 on Agency. An analysis is one person scored against one role. Merged duplicates count once, so the same person found in two places is not paid for twice. Removing a candidate later does not hand the analysis back, because limits count consumption during the billing period and the work was already done.
For a startup with two or three open roles, Starter covers a few hundred applicants per role with room to spare. Agencies screening inbound for many clients at once usually sit on Growth or Agency. A meter in the dashboard shows usage, a prompt appears at 80 percent, and upgrading charges the difference for the current period straight away so screening never pauses mid role.
Questions buyers ask
Anything else goes to support@recruitingbot.com or the help panel inside your dashboard.
How accurate is automated resume screening?
Accuracy depends mostly on the rubric. Clear, specific criteria produce consistent scores, while vague ones like "great communicator" produce noise. RecruitingBot shows evidence for each criterion so you can check its reading of any CV in seconds, and editing a weight rescores the whole role. Treat the ranking as a well organized first read, not a verdict.
Can candidates game the screening by stuffing keywords?
Keyword stuffing works on filters that count words. Rubric scoring looks for evidence of the work itself, such as what was built, owned or changed and over what period. A list of tools without context earns little, and the breakdown makes padding easy to spot because the quoted evidence is thin.
Does RecruitingBot reject candidates automatically?
No. It ranks and explains. Deal breakers move people into a separate tab you can review, but no rejection email goes out until a person marks someone as rejected. Every shortlist is approved by a human, so final decisions stay with your team.
Can I screen applicants who are already in my ATS?
Yes. Upload a CSV on any plan, or import from Greenhouse, Lever or Workable on Growth and Agency. Imported candidates are parsed, matched against existing records to remove duplicates and scored against the role rubric just like new applicants and sourced profiles.
What file types can it read?
It reads PDF and Word files uploaded by you or by applicants, plus the text candidates submit through the hosted apply page. Profiles found through sourcing are analyzed directly. Parsed fields are always shown next to the original document, so a reviewer can confirm dates, titles and employers before trusting the score.
Who owns the rubric and the screening results?
Your workspace does. You are the controller of candidate data and RecruitingBot acts as processor. You can export or delete any candidate in one click, retention rules delete sourced non responders automatically, and rubric changes and approvals are kept in the role history for later review.
Keep reading
Give it a role and see the shortlist it builds
Type a job title on the homepage. The demo runs instantly without an account and shows an example pipeline, job post and shortlist for that role.