User Story

As a recruiter, I want the ATS to identify highly similar applicant résumés and application patterns—even when candidates use different names, formatting, or minor wording changes—so that I can efficiently review potential duplicate, templated, or unreliable applications before investing time in interviews.

Problem

Exact-match duplicate detection is insufficient for high-volume remote recruiting. We are seeing applications that appear to be generated from a common template or source:

Nearly identical résumé structure, phrasing, KPI claims, and work-history archetypes

Slight changes to names, dates, employers, and formatting

Names presented in different orders across the application, résumé, document title/metadata, email address, and LinkedIn URL

Near-identical roles and claims across multiple applicants

These submissions are difficult to identify through standard duplicate checks because the email, phone number, and primary applicant name differ.

Requested Capability

Provide fuzzy/semantic similarity matching across résumé text and applicant data, with an explainable review experience for recruiters.

The system should:

Extract and normalize résumé text from PDF, DOCX, and image-based documents using OCR where needed.

Compare a new résumé against prior submissions within the customer’s tenant.

Detect both near-exact and semantic similarity despite changed formatting, names, dates, or minor word substitutions.

Generate a configurable similarity score and flag records above a chosen threshold.

Surface the specific matched applications and the evidence driving the score, such as:

Shared phrases or repeated bullet points

Similar work-history sequences, employers, titles, or dates

Repeated KPI/statistical claims

Similar skills/certifications/education wording

Conflicting or alternate names found in document title/metadata versus application fields

Support filtering/searching for a cohort of similar applications, not just a one-to-one duplicate match.

Allow users to tag confirmed patterns for internal follow-up and export the results for investigation.

Acceptance Criteria

When a recruiter reviews a new applicant, the system displays a visible alert if the résumé exceeds the configured similarity threshold against one or more prior applicants.

The alert includes:

Similarity score

Links to matched applicant records

Highlighted matching content

Date submitted

Match type, such as exact text, fuzzy text, work-history pattern, or document-metadata discrepancy

Matching remains effective when applicants:

Reorder first, middle, and last names

Modify dates or company names

Reformat the résumé

Make minor wording changes to copied bullet points

Recruiters can tune the threshold and choose whether alerts are informational only or create a review tag/workflow.

The system does not automatically reject or disqualify an applicant. It provides evidence for human review.

Access to matched applicant information follows existing ATS role-based permissions and audit controls.

The matching logic must not use protected-class characteristics or infer protected attributes.

Example Outcome

A recruiter receives a new résumé with a different applicant name and contact information. The system flags it as 87% similar to several prior submissions because it contains the same survey-role metrics, healthcare-support bullet points, solar-sales history, local-degree language, and repeated skills wording. The recruiter can review the matched records and decide whether to proceed with a standard verification interview or disposition the application.