Capifiq is the patent-protected integrity layer for remote technical interviews. We produce deterministic, binary evidence that AI-assist tools were running — not a probability score, not a behavioral hunch.
Early access to investor materials is available to accredited investors and qualified institutional buyers.
Request Investor AccessA new class of AI-assist tools has made traditional remote interviews unreliable. The problem emerged fast, and no incumbent has solved it.
AI-assist adoption — up from 15% in six months
Fabric analysis of 19,368 live interviews, July 2025 – January 2026. The rate of flagged interviews jumped from 9% to 45% in that window.
Major employers reverting to in-person interviews
Google and McKinsey reintroduced mandatory in-person rounds specifically to counter AI interview fraud. The FBI issued warnings about state-sponsored actors using interview fraud to infiltrate corporate networks.
Job fraud losses in 2024 — up from $90M in 2020
FTC data. Gartner projects that by 2028, one in four candidate profiles will be entirely fabricated using generative text, synthetic voice, and deepfake video.
No existing solution delivers deterministic, evidence-based detection without disrupting the interview workflow. Capifiq fills the gap.
A subscription industry has emerged around AI-assist tools marketed to candidates for use during live interviews. They render invisible overlays, capture interviewer audio through virtual drivers, and generate responses in one to two seconds. They are engineered specifically to evade screen-sharing, browser proctoring, and behavioral monitoring. Capifiq detects all of them.
Desktop overlays
Native apps that render AI answers as a transparent overlay on the candidate's screen. Invisible to screen-sharing APIs, hidden from the system tray, disguised process names. Operate below the browser, where proctoring cannot see.
Hybrid stealth apps
Ship both a stealth desktop app and a browser extension. Market themselves as “undetectable” across Zoom, Teams, Meet, and coding platforms. Claim OS-level invisibility on shared screens.
Browser copilots
Browser-based AI assistants that transcribe the interviewer and display answers in a side panel or voice feed. Frequently marketed as “interview copilots.”
Why this surface keeps growing
Final Round AI reports 10 million users. ParakeetAI has driven 500 million social views in a year. LockedIn AI, Sensei AI, Beyz AI, Verve AI and a dozen other entrants launched in the past eighteen months. Each retails for $20–$60 per month against a $150,000+ target salary, so the economics make this a permanent category, not a passing trend. Every new tool that ships expands the surface defenders have to cover. Our detection core is tool-agnostic by design: it does not match known binaries or signatures. It targets the architectural technique the entire category depends on — which is why it continues to work against tools that have not been built yet.
Top-tier VCs have funded the offensive side of this market at nine-figure valuations. The defensive side is wide open — and structurally worth more.
The bellwether
raised at ~$120M valuation
Cluely closed a $15M Series A led by Andreessen Horowitz in June 2025, on top of a $5.3M seed from Abstract Ventures and Susa Ventures. Valuation reported at ~$120M post-money. When a16z backs the offensive side at that valuation, the category signal is unambiguous.
The category
valuations across adjacent interview AI
Yoodli (interview practice & coaching) tripled its valuation to $300M+ in December 2025 on a $40M Series B led by WestBridge, with Madrona and Neotribe participating. Final Round AI reports 10M+ users. Interview Sidekick reports 721K. The category is attracting both capital and consumer scale at the same time.
The asymmetry
vs. the B2C offensive side
Every AI-assist tool sells one license at a time to individual candidates. Integrity sells into the Fortune 5000, staffing firms, universities, and every regulated vertical — annual enterprise contracts, not monthly consumer subscriptions. The buyer population for defense is multiples larger than the buyer population for offense.
The thesis in one line
If a16z believes an undetectable AI-assist tool is worth $120M at Series A, the company that reliably detects it — with a patent-protected method and deterministic, binary evidence — is a materially larger opportunity. Every dollar of funding into the offensive side expands our addressable market.
Sources: TechCrunch (Jun 2025, Dec 2025), PitchBook, Crunchbase, PR Newswire, GeekWire. Figures reflect publicly disclosed funding rounds as of April 2026.
Interview integrity is a layer over the entire online recruitment stack. As hiring continues its shift online, every remote interview becomes a candidate for integrity monitoring.
TAM
online recruitment technology by 2034
The global online recruitment technology market is projected to grow from $17.48B in 2026 to $46.07B by 2034 at a 12.9% CAGR. Every remote interview conducted inside that market is a potential integrity event.
SAM
recruitment software by 2033
The recruitment software segment — ATS, assessments, and adjacent tooling where integrity slots in as a native layer — is on a path from $3.05B in 2025 to $6.48B by 2033 at a 9.9% CAGR. Four independent research houses converge on the same trajectory.
SOM
enterprise & staffing targets
Fortune 5000 companies, the top 500 staffing and consulting firms, and every university running remote technical screens. Each runs hundreds to thousands of interviews annually. Integrity is not a nice-to-have for this buyer — it's a prerequisite for defensible hiring.
Sources: Fortune Business Insights (online recruitment technology, 2026), Coherent Market Insights / SkyQuest / Mordor Intelligence (recruitment software, 2025–2033).
The only interview integrity platform combining a native agent, dual-mode AI, and a patent-protected deterministic detection core.
Pillar 1
A lightweight desktop application the candidate installs in under 30 seconds. It runs alongside Zoom, Teams, or Meet — no workflow change for the employer. Today's AI-assist tools are native desktop applications that operate below the browser. You cannot detect them from inside a browser. This is architectural, not a feature choice. No incumbent has a native agent; building one is a 6–12 month engineering effort.
Pillar 2
Active detection monitors in real time during the live interview and pushes an integrity signal — green, yellow, red — to the interviewer's dashboard. After the session ends, the passive layer ingests the full recording: transcription, speaker diarization, audio-visual coherence mapping, answer-originality scoring, and temporal pattern correlation across the entire session. Patterns invisible in a five-second real-time window become obvious when AI reviews forty-five minutes with no time constraint. No competitor runs both.
Pillar 3
The detection method at the heart of the active engine is covered by a filed US provisional patent. It defeats every known AI-assist tool on the market and produces deterministic, binary evidence — not a probability score. The tool was either running, or it wasn't. Claims cover the fundamental technique and its variations. Technical detail is disclosed under NDA during due diligence.
We compared Capifiq against every category of incumbent: full-platform replacements, browser extensions, proctoring suites, assessment platforms, and human interviewers. None detect invisible-overlay AI-assist tools architecturally. None produce deterministic evidence. None combine active and passive AI in a companion model.
| Competitor | Native desktop agent |
Detects invisible overlays |
Deterministic evidence |
Active + passive AI |
Companion (keep Zoom) |
Patent- protected core |
|---|---|---|---|---|---|---|
| Capifiq | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Fabric | — | — | — | — | — | — |
| Talview | — | Partial1 | — | Partial | — | — |
| Sherlock AI | — | — | — | — | ✓ | — |
| Truely | — | Partial1 | — | — | — | — |
| HackerRank | — | — | — | — | — | — |
| HackerEarth | — | — | — | — | — | — |
| Proctorio / Honorlock | — | — | — | — | — | — |
| Karat | N/A2 | N/A2 | N/A2 | N/A2 | N/A2 | N/A2 |
1 Talview and Truely offer behavioral and environmental monitoring (secure browser, dual camera, audio-anomaly detection). These are useful but operate above the layer where modern AI-assist tools render. They cannot architecturally see a below-browser invisible overlay.
2 Karat is a human-interviewer service, not a software platform. Included for completeness; feature comparison does not apply.
Fabric
Full platform replacement · Behavioral analysis (self-reported 85% accuracy)
Our edge: companion model preserves Zoom/Teams/Meet. Deterministic proof instead of probability.
Talview
Secure-browser lockdown · Seven-layer proctoring · Deepfake & voice-anomaly detection
Our edge: invisible overlays render below the browser. Secure-browser enforcement cannot see them; a native agent can.
Sherlock AI
Browser extension companion · Behavioral-only signals · Low friction
Our edge: a browser extension cannot inspect native desktop processes. Patent-protected detection core.
Truely
Desktop monitoring for exam integrity · Environmental scanning
Our edge: signature-based process detection is evaded by every new tool. Our core is architecture-based, not signature-based.
HackerRank · HackerEarth
Browser-based assessment & proctoring · Tab-switch & copy-paste flags
Our edge: AI-assist tools are native apps that bypass the browser entirely. Browser proctoring is architecturally blind to them.
Proctorio · Honorlock
Exam-focused proctoring (academic) · Browser lockdown · Recording review
Our edge: built for academic test integrity, not live technical interviews. No desktop agent, no active/passive AI, no overlay detection.
Capifiq is the only company on the planet combining a native desktop agent, patent-protected deterministic detection, and dual-mode AI in a companion model.
No incumbent has a native agent. No incumbent has a patent on deterministic detection. The technology is built and the priority date is filed. What remains is speed.
Reduction to practice
Proven against live tooling
The core detection method has been validated against the current generation of invisible-overlay tools on the latest Windows platform. Not a lab demo — working agent, real targets.
Priority date
US provisional patent filed
Three inventions covered, claims drafted, prior art analyzed. Assignment to the operating company will follow corporate formation.
The category window
12 to 18 months to own it
Cluely is hiring growth teams. Fabric is consolidating share. Incumbents are quietly evaluating builds. Whoever establishes the integrity category first establishes it for a decade.
We have the technology. We have the patent. We have the category framing. We are raising to move fast enough to win it. Round structure, valuation, and use of proceeds are shared under NDA with verified accredited investors.
Documents are delivered by email after identity verification and NDA execution. We do not distribute investor materials without a verified record.
Verify your identity to access the investor portal. All materials are delivered via email after verification.
Loading form…