See what AI recruiting platforms sync to your ATS, from interview notes to scorecards. Learn how to verify field mapping, permissions and writeback accuracy.
Vendor demos often make sweeping claims about "seamless, two-way ATS integration." In real-world recruiting operations, however, a "synced" status can mean anything from a structured, native scorecard submission to an unformatted plain-text note dropped onto the wrong candidate application.
For talent acquisition leaders, the difference between surface-level synchronization and true field-mapping accuracy determines whether hiring managers actually adopt your ai recruiting platform—or spend hours chasing missing candidate evaluation data.
This guide breaks down exactly what writes back to your Applicant Tracking System (ATS), where data lands in systems like Greenhouse and Lever, how timing and permission models work, and how to execute a step-by-step verification audit before buying ai recruiting software.
What Is ATS Writeback?
ATS writeback is the automated transfer of candidate evaluation data, interview summaries, competency scores, and workflow triggers from an external recruiting tool back into your ATS.
A basic one-way integration only imports candidate profiles or job requisitions into the AI platform. A writeback-enabled integration pushes structured interview findings out to your ATS so hiring teams can evaluate candidates within their existing workflows.
However, actual writeback behavior varies significantly based on integration architecture, API endpoints, and granted OAuth permissions.
What Actually Syncs Back to an ATS?
Not all writeback mechanisms produce the same operational results. This table details what talent acquisition buyers must inspect during live vendor testing and provides explicit acceptance tests for each data type.
Output | Destination to Verify | Buyer’s Acceptance Test |
Screening results | Application field, note, or report | Confirm screening results map to the active application, rather than floating at the candidate profile level when multiple applications exist. |
Interview summary | Candidate/application note | Compare the AI-generated summary directly against the raw source transcript to confirm statement accuracy and readability. |
Competency ratings | Native scorecard, custom field, or attachment | Verify that custom competency ratings map directly to your ATS rubric scales (e.g., 1–4 rating scale). |
Note, report, or linked transcript | Confirm that reviewers can trace numerical ratings directly back to verbatim candidate quotes, timestamps, or question segments. | |
Recording | Attachment or external URL | Open video/audio playback URLs using standard hiring manager permission roles to verify single sign-on (SSO) and access settings. |
Stage change | Application pipeline stage | Verify which trigger initiates stage movement and confirm whether recruiter approval is required before pipeline migration. |
Is a Synced Note the Same as a Completed Scorecard?
No. Writing an interview summary as a general activity note is technically and operationally distinct from submitting a native ATS scorecard.
Where interview data lands inside your ATS directly dictates how easily your team can report on, aggregate, and review candidate performance.
1. Plain-Text Notes
How it works: Pushes raw text or markdown into the candidate's general activity log.
Limitations: Notes cannot populate structured ATS reporting filters or drive native candidate comparison tools. While simple to implement via basic POST endpoints, unstructured notes are difficult to aggregate across high-volume pipelines.
2. PDF Attachments
How it works: Generates a formatted PDF report and uploads it to the application's file attachments tab.
Limitations: PDFs preserve layout structure and full transcripts, but the underlying data remains static. ATS reporting tools cannot parse or analyze data stored inside binary attachments.
3. External Hyperlinks
How it works: Drops a URL link inside a note or custom field pointing back to the AI vendor’s proprietary web portal.
Limitations: Forces hiring managers to leave their central ATS, requiring separate authentication or user seats in the external application.
4. Custom Fields
How it works: Writes specific data points into custom candidate or application attributes (e.g., AI_Screening_Recommendation: Pass).
Limitations: Custom fields allow quick search filtering, but they rarely offer enough structural depth to present multi-tiered competency rubrics.
5. Native Scorecard Submissions
How it works: Maps AI evaluation outputs directly into native ATS scorecard objects (e.g., populating individual skill ratings, written comments, and overall hiring recommendations).
Limitations: Requires specialized API endpoints (e.g., Greenhouse Harvest API scorecard endpoints) and elevated OAuth scope permissions during setup.
Integration Case Study (Greenhouse Harvest API):
A generic claims of "Greenhouse Integration" is insufficient proof of native scorecard delivery. The Greenhouse Harvest API separates candidate notes (POST: Add Note) and file uploads (POST: Add Attachment) from formal interview scorecards (GET: Retrieve Scorecard). Vendors must explicitly demonstrate which endpoints their integration utilizes.
Does Two-Way Sync Mean Real-Time Updates?
Many vendors use "two-way sync" as a blanket term implying immediate, real-time data flow. In system integration architecture, direction, trigger mechanism, timing, and approval requirements are separate technical parameters:
Direction (One-Way vs. Two-Way): One-way sync reads data from the ATS; two-way sync reads from and writes back to the ATS.
Trigger Mechanism (Webhooks vs. Polling): Webhooks push immediate event notifications when an action occurs (e.g., Candidate has changed stage). Polling checks the ATS for changes on a fixed schedule (e.g., every 30 minutes).
Timing (Real-Time vs. Asynchronous Queue): Real-time integrations process requests instantly. Asynchronous batching holds data in queues to respect ATS API rate limits and process background tasks without connection timeouts.
Approval Requirements (Automated vs. Human-in-the-Loop): Automated sync pushes summaries immediately after an interview completes. Human-in-the-loop workflows hold evaluation drafts in a staging queue until a recruiter audits and approves the results.
Technical Architecture: A webhook is an event notification mechanism, whereas a REST API is the transport mechanism used to read or modify data. A webhook alerts the integration that an interview ended; the AI platform then executes REST API requests to populate the interview scorecard inside the ATS.
How Do You Verify an AI Recruiting Platform’s Writeback?
Before finalizing software procurement, execute this practical testing checklist in a sandbox environment to verify writeback integrity:
Phase 1: Pre-Demo Integration Requirements
Request access to a staging or sandbox environment linked to a test ATS instance (e.g., Greenhouse, Lever, or SmartRecruiters).
Review the vendor’s endpoint mapping documentation to verify which API calls are executed during writeback.
Phase 2: Live Acceptance Testing
Trigger an Evaluation Event: Complete an ai candidate screening or interview session using test candidate credentials.
Verify Application-Level Context: If the test candidate has multiple active applications, confirm that outputs land exclusively on the specific target requisition.
Audit Field Mapping Placement: Check whether feedback populates native scorecard attributes, custom metadata fields, or activity stream notes.
Inspect Evidence & Playback Links: Click embedded timestamp and audio links using a standard hiring manager account to verify authentication permissions.
Validate Pipeline Stage Movements: Confirm that automated stage changes execute smoothly without causing duplicate triggers or unauthorized candidate emails.
What Breaks—and How Should the Integration Recover?
Production integrations encounter rate limits, expired tokens, and data edge cases. Evaluate how your platform handles these common failure scenarios:
Multi-Application Conflicts: Candidates frequently apply to multiple open roles. Integrations must scope writebacks using explicit application_id handles rather than generic candidate_id keys to prevent routing evaluation scores to the wrong team.
Duplicate Event Retries: Network retries can re-fire webhook events. The writeback pipeline must use deduplication keys to prevent writing duplicate notes or scorecards into the ATS.
Permission and Key Expiration: If an API key expires or lacks required scopes, the system should retain payload data in an automated retry queue and notify system administrators without dropping records.
Overwriting Human Edits: Automated writebacks must never overwrite manual recruiter notes or existing human scorecards unless explicitly configured with overwrite permissions.
A Verified JobTwine Workflow Example
To evaluate a concrete enterprise integration flow, consider JobTwine's verified integration sequence with Greenhouse Recruiting:
Setup & Webhook Handshake: The organization creates a dedicated Harvest API key in Greenhouse with specific permissions for candidates, jobs, scorecards, and notes. Webhooks are configured for stage changes and job updates.
Candidate Export & Interview Trigger: When a recruiter moves a candidate into the designated JobTwine stage in Greenhouse, the candidate profile exports to JobTwine for scheduling.
Automated Evaluation Execution: The candidate completes their structured interview with JobTwine's AI avatar agent, JayT. The engine evaluates answers against standardized rubrics and captures fraud detection signals.
Structured Writeback Delivery:
Scorecards & Notes: JobTwine calls Greenhouse Harvest API endpoints to push structured scorecards, evidence summaries, and full transcripts directly into the candidate profile.
Evidence Access: Reviewers click timestamped links directly within Greenhouse to inspect video clips and verify candidate scoring.
Scope Limitation: Available field mappings and automation capabilities depend on your organization's specific ATS tier, API permissions, and administrative configuration.
Questions to Ask Vendors Before Buying
Ask AI recruiting vendors these direct operational questions during technical evaluations:
Where specifically does interview feedback land in our ATS layout (notes, scorecards, custom fields, or attachments)?
Does your writeback populate native ATS scorecard fields, or does it drop raw text into candidate notes?
How does your platform identify the correct application when a candidate has multiple active applications?
Do hiring managers need a separate user account on your platform to view interview playback and transcripts?
How does your integration handle API rate limits, temporary outages, or expired access tokens?
Conclusion: Verifying Real ATS Writeback
Evaluating an ai recruiting platform requires looking past demo slides and auditing actual field-level behavior in your ATS. Real two-way integration depends on precise API mapping, correct endpoint utilization, and robust recovery mechanisms when network or permission errors occur.
By requiring live sandbox demonstrations, testing multi-application edge cases, and verifying where evaluation evidence lands, talent acquisition leaders can ensure their ai candidate screening tools deliver actual workflow efficiency rather than operational overhead.