Face Verification adds an identity signal to AI Calls. When a candidate completes an AI Call, Talkpush automatically compares their face from the call video against a reference image or video they submitted earlier in the process. The result — a match, review, or no-match — appears as a badge on the candidate profile, with no manual work required from the recruiter.
Note: To activate this feature, please contact your Account Manager.
Why It Matters in Hiring
When hundreds of candidates are completing AI interviews remotely, there is no proctor in the room. Without an identity check, there is no reliable way to know whether the person on the call is the person who applied.
Impersonation is a real and low-effort risk — a stronger candidate can complete an assessment on behalf of someone else with no special tools and no visible trace in the recording. Face Verification is designed to catch exactly this, without adding friction for the candidates who are not cheating.
The feature is particularly valuable for organizations in BPO, staffing, financial services, healthcare, and other regulated or high-stakes hiring environments where the cost of placing the wrong person is high.
How It Works
Face Verification runs in two stages: enrollment and verification.
Enrollment happens when a candidate submits a video or photo question that has been marked as a face capture source. Talkpush extracts a face template from that submission and stores it against the candidate's profile. This step is silent — the candidate experience does not change.
Verification happens automatically when an AI Call completes. Talkpush compares the face captured in the call video against the stored reference template. The result and a confidence percentage are written back to the candidate's profile as badges and candidate attributes.
Note: Face Verification compares the face in the AI Call video against a reference the candidate previously submitted in Talkpush — it is not a government identity document check.
A few important behaviors to know:
One reference per candidate. The first enrolled face template is used for all future AI Calls on that profile. Repeat applications do not overwrite the existing reference.
The feature never blocks a call. If no face reference exists, or if the verification service is unavailable, the AI Call completes normally and the badge reflects that verification could not run.
Multiple faces = no match. If the call video contains more than one distinct face in any frame, the result is automatically set to no match, regardless of confidence score.
Retroactive verification is not supported. Only calls completed after the feature is enabled for your company are verified.
Reading the Face Verification Badge
The badge appears on the candidate profile, next to the TalkScore pill and any voice verification badge. There are four possible states:
Badge | What it means |
✅ Face verified · N% | High-confidence match — face in the AI Call video matches the enrolled reference above the confidence threshold. No action needed. |
🟡 Face review · N% | Borderline match — recruiter judgment recommended before proceeding. |
🔴 Face mismatch · N% | Face did not match the reference, was below the confidence threshold, or multiple faces were detected in the call video. |
🔴 Face mismatch | Verification could not run — no face reference was enrolled, or the service was unavailable. The AI Call completed normally. |
Note: If you see an enrollment warning on an answer card, it means the face reference capture failed for that specific submission. No comparison was possible for that AI Call. Check that the question is correctly configured as a face capture source (see Setting Up Face Verification below).
Setting Up Face Verification
Before any verification can run, two things need to be in place: the feature must be active for your company, and at least one question in your campaign must be configured to capture a face reference.
Step 1 — Feature activation
A Talkpush admin enables Face Verification for your tenant from the backend. This is a one-time step — once it is on, it applies to all campaigns for your company. If the feature is not yet active, contact your Account Manager to request activation.
Step 2 — Flag a question as a face capture source
You need to mark a video or file upload question in your question library as the face capture source. This is the question whose response Talkpush will use to extract and store the candidate's face template — it runs silently in the background when the candidate submits their answer.
Open the Templates Tab and go to Questions.
Open the Video or File Upload question you want to use, or create a new one.
Enable the Capture Face ID setting.
Click Save.
Note: Only Video and File Upload questions support the Capture face toggle. Audio-only and text questions cannot be used as a face enrollment source.
Step 3 — Add the question to your campaign
A flagged question only captures a face reference if the candidate actually sees and answers it. Make sure the question is included in the relevant campaign — either added directly or via a question set — so candidates encounter it before their AI Call.
For help adding questions to a campaign, see the Questions article in the Templates Tab guide.
Step 4 — Results appear automatically
Once a candidate submits the face-capture question and completes their AI Call, Face Verification runs in the background and the badge populates on their candidate profile. No further action is needed from the recruiter.
How It Looks for the Candidate
The candidate experience does not change. Whether the feature is on or off, the application flow looks identical — no extra prompt, no additional disclosure, no new step. Candidates submit their video or photo as they normally would, and their AI Call proceeds as usual.
Face Verification runs in the background after the call ends.
Webhook and Integration
For companies with Face Verification enabled, the AI Call result webhook automatically includes three new fields:
Field | Definition |
face_verification | The verification outcome: match, review, or no_match. |
face_verification_confidence | Confidence score as a percentage (e.g. 94). |
face_verification_at | Timestamp of when the verification ran (ISO 8601). |
These fields are present only for opted-in tenants. If your company has Face Verification disabled, these fields will not appear in the webhook payload. Partners consuming the AI Call webhook should handle the case where these fields are absent.
Note: Face Verification results are also available as candidate attributes in Talkpush, so they can be used in Autoflow rules and exports without any additional integration work.
Q: What happens if face enrollment fails on a candidate's answer? A: A short warning appears under that specific answer in the candidate profile. The AI Call pipeline is never blocked — if no face reference exists when the call completes, the call finishes normally and no badge is shown.
Q: Does Face Verification replace government ID checks? A: No. Face Verification confirms that the face in the AI Call video matches the reference the candidate submitted earlier in Talkpush. It is not connected to any government identity database and should not be treated as a substitute for official identity document verification.
Q: Do existing integrations need to be updated? A: No. If you receive the AI Call webhook, the three face verification fields appear automatically in the payload once the feature is enabled for your tenant — no changes required on your end. Partners should handle the case where these fields are absent, for tenants where the feature is off.
Q: Can I use Face Verification for non-AI-Call interviews? A: Not currently. Face Verification applies only to AI Calls. WhatsApp interviews and landing page async video submissions are not in scope for this feature.
Q: Is any legal review needed before enabling this feature? A: Facial recognition data is biometric data in several jurisdictions. If you operate in the EU, UK, or other regions with biometric data regulations, consult your Account Manager before enabling Face Verification.
|
