Why and how it works.

Defensible in a Title VI audit

Every certificate arrives with a named human evaluator, timestamps, evidence notes, and instrument version. Your general counsel can walk any decision from prompt to signature — the paper trail is the product.

Cut miscommunication risk in the field

Agencies replacing self-report or checklist screening with Certifiably Human typically eliminate the 'my staffer said they were bilingual' failure mode that drives grievances, sentinel events, and OCR complaints.

From invitation to certificate in under 48 hours

Candidates complete the assessment on their own device. Evaluators receive AI-focused hot points on submission. Most decisions clear the human queue the next business day — not the next quarter.

Human-in-the-loop by design, not by policy

AI advises; a credentialed human decides. That single architectural choice is what makes the certificate admissible where AI-only scores are not — and what protects the agency from algorithmic-bias litigation exposure.

Equity for Deaf and ASL communities

ASL is scored on video by Deaf or CODA raters against the ASLPI rubric — never by audio transcript, never by AI alone. Section 504 and ADA Title II obligations are met with evidence, not attestation.

Procurement-ready evidence package

Signed weekly exports, per-candidate audit files, rater rosters, and instrument versions are formatted for CMS Section 1557, HHS OCR, and state administrative office reporting — no consultant middle-layer required.

Revocable, verifiable, expirable

Every certificate has a public verify URL, a two-year expiry, and a documented revocation path. Employers and courts confirm status in one click; expired or revoked credentials show that status instantly.

Built for the agency, not the vendor

Rubrics align to ILR, ACTFL, and ASLPI — the frameworks state and federal programs already cite. Agencies can raise the qualifying floor per contract; the platform enforces it uniformly across candidates.

01

Invitation and permissions

The candidate opens a private invitation link and grants camera and microphone access inside the embedded assessment. Integrity signals are reviewed with the submission.

02

Embedded assessment

Candidates complete the assessment for their selected track in the embedded the assessment provider experience. Their responses and assessment results are returned for human review.

03

Assessment evidence

Available transcripts, scores and observations can help the evaluator focus their review. AI assistance never issues the final decision.

04

Qualified human evaluator

An assigned, language-qualified evaluator reviews the evidence and either approves with a level and reason or declines with a reason. ASL submissions require human review of the video.

05

Integrity and conflicts

Flagged submissions first go to an administrator, who clears the flag for evaluation or declines with a reason. Evaluators can disclose a conflict and return an assignment for reassignment.

06

Decision and certificate

The evaluator's approval or decline is recorded with a reason. Approval issues a certificate; administrators can revoke or override with an audited reason. Decisions remain part of the record.

Video, not audio

ASL is a visual language. All ASL sections are recorded on camera with the candidate framed shoulders-up. Audio-only submissions are automatically invalidated.

Eye-gaze on camera

The candidate must look at the lens during signed production. Sustained off-camera gaze (>2 seconds), off-frame drift, or another face appearing behind the candidate is flagged for evaluator review.

Single-tab, locked session

The assessment runs in a locked browser session. Tab blur, window resize below the recording frame, screen-share activation, or a second display all raise integrity flags with time stamps.

Full-hand visibility

Both hands must remain within the on-screen frame. Signing that leaves the frame, or hands obscured by clothing, jewelry, or props for more than one sign is flagged and the clip is marked for re-record.

No interpreter, no coaching

No third party may appear or speak in the recording. Voices, off-screen prompts, or a person visible in the background invalidates the take. Environmental audio is retained for forensic review only.

Prompt-level uniqueness

Each receptive prompt is delivered by a live-signed avatar unique to that session. Candidates cannot replay a prompt more than twice, and each replay is logged with timestamps.

Deaf/CODA rater required

Every ASL submission is scored by a Deaf or CODA evaluator on the ASLPI rubric. AI provides advisory hot/cold indicators only — it never scores ASL and never issues the final decision.

AI as a spotlight, not a judge

The AI pass highlights up to five moments per clip (non-manual markers, handshape ambiguity, spatial referents) for the evaluator to look at closely. The rater may accept, dismiss, or override every AI observation with a note.

Keyboard-operable end to end

Every candidate, evaluator, and administrator flow — including video capture, playback, and hot-point navigation — is reachable and operable by keyboard alone, with a visible focus indicator on every control.

Screen reader and semantic structure

Landmarks, heading order, form labels, and live-region status messages are exposed to assistive technology. Status changes such as recording start, upload, and decision confirmation are announced rather than shown by color alone.

Contrast and text sizing

Body and interface text meets WCAG 2.1 AA contrast at default settings, supports 200% zoom without loss of function, and never uses color as the sole carrier of meaning in flags, bands, or integrity states.

Deaf and hard-of-hearing candidates

ASL prompts are delivered in signed video with English gloss and captions. Written instructions accompany every spoken prompt, and no task requires hearing to complete.

Motion and timing

Animations respect the operating system reduced-motion preference. Recording limits are generous, retakes are unlimited before submission, and no task fails on a hidden timer.

Accommodations on record

Requested accommodations — extended time, interpreter presence, alternate input — are recorded on the submission and visible to the evaluator, so the decision is documented in the context it was made.

01

Layer 1 — Human review only

Capture, route, and human decision. No AI output is generated or shown anywhere in the workspace.

• Video and audio capture
• Qualified evaluator scoring against ILR / ACTFL
• Signed decision, audit trail, revocable credential

02

Layer 2 — + AI advisory for Spanish

Adds transcription and advisory observations for Spanish–English submissions. Evaluator decision is unchanged.

• Everything in Layer 1
• Spanish–English transcription with confidence bands
• Advisory rubric observations, never a score

03

Layer 3 — + AI advisory for ASL

Adds video analysis and hot points for ASL submissions, reviewed by a qualified Deaf or CODA rater.

• Everything in Layer 2
• ASL video hot points and integrity trace
• Deaf / CODA rater retains the decision

04

Layer 4 — + AI advisory across all languages

Full advisory coverage across every supported language pair, with the same human decision gate.

• Everything in Layer 3
• All supported language pairs
• Fairness monitoring across the full portfolio

Title VI of the Civil Rights Act of 1964

Requires meaningful access to services for individuals with limited English proficiency. Certifiably Human documents rater qualifications and language pair coverage in a form suitable for LEP compliance filings.

Section 504 & ADA Title II

Requires effective communication with individuals who are Deaf or hard of hearing, including qualified ASL interpretation. Every ASL certificate carries the ASLPI level, evaluator credentials, and integrity status.

State court interpreter programs (NCSC framework)

Rubrics align to the ILR 0–5 scale used by the National Center for State Courts consortium. Instrument versions, prompts, and rater rosters are exportable for state administrative office of the courts audits.

CMS & HHS language access guidance

Health and human services programs receiving federal funds must document language assistance. Weekly compliance exports are formatted for CMS Section 1557 and HHS OCR reporting workflows.