{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"tool_liarliar","slug":"liarliar","name":"Liarliar","type":"product","url":"https://liarliar.pro","page_url":"https://unfragile.ai/liarliar","categories":["automation"],"tags":[],"pricing":{"model":"freemium","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"tool_liarliar__cap_0","uri":"capability://text.generation.language.text.based.deception.pattern.analysis","name":"text-based deception pattern analysis","description":"Analyzes written text input through undisclosed machine learning models to identify linguistic patterns claimed to correlate with deceptive statements. The system processes natural language features (word choice, sentence structure, temporal references) and outputs a confidence score or binary classification. Implementation details are not publicly documented, raising questions about whether the approach uses transformer-based embeddings, rule-based heuristics, or statistical pattern matching.","intents":["I need to screen job applicants for dishonesty in written responses","I want to identify false claims in legal documents or depositions","I need to verify truthfulness of candidate statements during hiring"],"best_for":[],"limitations":["No peer-reviewed validation of accuracy; claimed capabilities lack scientific evidence and contradict established deception research showing AI accuracy barely exceeds chance (50-55%)","Produces high false positive rates that could wrongly flag truthful statements, damaging innocent individuals' careers and relationships","No transparency on training data, model architecture, or validation methodology; inability to audit or understand decision factors","Susceptible to adversarial inputs and gaming—users can learn to manipulate linguistic patterns to evade detection","No cross-cultural or multilingual validation; linguistic patterns vary significantly across languages and cultural communication norms"],"requires":["Text input in English (assumed; other languages not documented)","Internet connection for cloud-based API calls","User account with Liarliar platform (freemium tier available)"],"input_types":["plain text","written statements","interview transcripts","email or chat messages"],"output_types":["confidence score (0-100 or percentage)","binary classification (truthful/deceptive)","structured report with flagged phrases"],"categories":["text-generation-language","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_liarliar__cap_1","uri":"capability://text.generation.language.speech.to.text.deception.scoring","name":"speech-to-text deception scoring","description":"Processes audio or video input (likely through speech-to-text conversion followed by the same text analysis pipeline) to generate deception likelihood scores from spoken statements. The system presumably transcribes audio to text, then applies linguistic pattern matching. No documentation clarifies whether prosodic features (tone, pitch, pause patterns) are analyzed independently or only text-derived features are used.","intents":["I want to analyze recorded interviews or depositions for signs of deception","I need to screen video interview responses from job candidates","I want to flag suspicious statements in recorded conversations"],"best_for":[],"limitations":["Relies on speech-to-text accuracy, which introduces compounding errors—transcription mistakes propagate into deception scoring","No evidence that prosodic features (tone, hesitation, pitch variation) improve detection accuracy; most deception research shows these signals are unreliable","Audio quality degradation (background noise, accents, speech impediments) directly impacts transcription and downstream analysis","No validation on diverse speaker populations; potential bias toward certain accents, speech patterns, or demographic groups","Privacy and consent concerns: recording and analyzing speech without explicit informed consent may violate wiretapping laws in many jurisdictions"],"requires":["Audio or video file in common formats (MP3, WAV, MP4 assumed; specific formats not documented)","Internet connection for cloud processing","Liarliar account with audio analysis tier (if separately priced)"],"input_types":["audio files (MP3, WAV, etc.)","video files (MP4, MOV, etc.)","live audio stream (if supported)"],"output_types":["deception confidence score","transcript with flagged segments","timeline of suspicious statements"],"categories":["text-generation-language","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_liarliar__cap_2","uri":"capability://data.processing.analysis.batch.statement.verification.with.report.generation","name":"batch statement verification with report generation","description":"Accepts multiple text inputs (candidate responses, document excerpts, interview transcripts) in batch mode and generates a consolidated report ranking statements by deception likelihood. The system likely processes inputs asynchronously, stores results in a database, and formats outputs as downloadable reports (PDF, CSV). No details on batch size limits, processing latency, or report customization options are publicly available.","intents":["I need to screen 50+ job candidates and rank them by truthfulness","I want a summary report of suspicious statements across multiple interviews","I need to export deception analysis results for HR review or legal proceedings"],"best_for":[],"limitations":["Batch processing introduces latency; no SLA or performance guarantees documented","Reports lack explainability—users cannot understand why specific statements were flagged, making results legally indefensible","No audit trail or versioning; impossible to track how results were generated or reproduce analysis","Batch results may be used to make discriminatory hiring decisions, creating legal liability for employers","No integration with HR systems (ATS, HRIS); requires manual data entry and export workflows"],"requires":["Liarliar account with batch processing capability","CSV, JSON, or text file with statements to analyze","Internet connection for upload/download"],"input_types":["CSV with statement column","JSON array of text objects","plain text file with line-separated statements","uploaded interview transcripts"],"output_types":["PDF report with rankings and scores","CSV export with statement-level deception scores","HTML dashboard with visualizations"],"categories":["data-processing-analysis","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_liarliar__cap_3","uri":"capability://automation.workflow.freemium.tier.access.with.usage.limits","name":"freemium tier access with usage limits","description":"Provides free trial access to core deception analysis features with rate-limiting and feature restrictions (e.g., limited analyses per month, no batch processing, no report exports). Paid tiers unlock higher quotas and premium features. The freemium model is implemented via API key-based quota tracking and feature flag gating, allowing users to trial the tool before commitment.","intents":["I want to try lie detection without paying upfront","I need to test if this tool works for my use case before buying a license","I want occasional deception analysis without a subscription"],"best_for":["Individual users or small teams evaluating the tool","Non-technical users who want to trial before commitment"],"limitations":["Free tier quotas are likely insufficient for production HR workflows (e.g., 10-50 analyses/month)","Freemium model may encourage users to make hiring decisions based on limited, unvalidated data","No clear upgrade path or pricing transparency; users may discover hidden costs after trial","Free tier may be intentionally crippled to drive conversions, not to provide genuine value"],"requires":["Email address to create account","No credit card required for free tier (assumed)"],"input_types":["text statements","audio/video (if supported on free tier)"],"output_types":["deception score","basic report (limited on free tier)"],"categories":["automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_liarliar__cap_4","uri":"capability://automation.workflow.hr.workflow.integration.and.candidate.screening","name":"hr workflow integration and candidate screening","description":"Positions the tool as part of HR hiring workflows, allowing recruiters to analyze candidate responses (written applications, video interview answers) and flag suspicious statements. The system likely provides a web dashboard or API for HR teams to upload candidate data and review deception scores alongside other evaluation criteria. No documented integrations with ATS (Applicant Tracking System) platforms like Workday, Greenhouse, or Lever.","intents":["I want to identify dishonest candidates during the hiring process","I need to flag candidates who lie about experience or qualifications","I want to reduce hiring risk by screening for deception"],"best_for":[],"limitations":["Using AI lie detection for hiring decisions exposes employers to discrimination lawsuits and EEOC violations—no AI system can reliably detect deception, and false positives disproportionately harm protected groups","No ATS integrations; requires manual data entry and export, reducing adoption in enterprise HR workflows","Results lack legal defensibility—cannot be used as sole basis for hiring decisions and may be challenged in court","Potential for algorithmic bias: if training data is skewed toward certain demographics, the tool may systematically flag truthful statements from underrepresented groups","No audit trail for hiring decisions; difficult to demonstrate fair, consistent evaluation if challenged"],"requires":["Liarliar account with HR tier","Candidate data (written responses, video transcripts, or audio files)","HR team with access to dashboard"],"input_types":["candidate application text","video interview responses","phone screening transcripts","reference check notes"],"output_types":["deception score per candidate","ranked candidate list by truthfulness","flagged statements with context"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_liarliar__cap_5","uri":"capability://text.generation.language.legal.document.and.deposition.analysis","name":"legal document and deposition analysis","description":"Analyzes written legal documents, witness statements, and deposition transcripts to identify potentially false or deceptive claims. The system processes legal text and outputs deception likelihood scores, presumably flagging statements that contradict known facts or exhibit linguistic patterns associated with deception. No documentation clarifies how the tool handles legal jargon, formal language, or the adversarial nature of legal proceedings.","intents":["I want to identify false statements in witness depositions","I need to flag suspicious claims in legal documents or contracts","I want to analyze opposing counsel's statements for deception"],"best_for":[],"limitations":["Legal proceedings require evidence that meets strict admissibility standards; AI deception scores are not scientifically validated and would likely be excluded under Daubert standards","False positives in legal contexts have severe consequences—innocent witnesses could be wrongly accused of perjury","Legal language and formal statements are fundamentally different from casual speech; the tool's training data likely doesn't account for legal register and conventions","Adversarial legal contexts involve strategic communication, not simple truth/deception; the tool cannot distinguish between legal strategy and dishonesty","No chain-of-custody or audit trail; results cannot be authenticated for legal proceedings"],"requires":["Liarliar account","Legal document or deposition transcript (text or PDF)","Understanding that results are not legally admissible evidence"],"input_types":["deposition transcripts","witness statements","legal briefs or motions","contract language","email correspondence in legal discovery"],"output_types":["deception score per statement","flagged passages with scores","summary report of suspicious claims"],"categories":["text-generation-language","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":22,"verified":false,"data_access_risk":"high","permissions":["Text input in English (assumed; other languages not documented)","Internet connection for cloud-based API calls","User account with Liarliar platform (freemium tier available)","Audio or video file in common formats (MP3, WAV, MP4 assumed; specific formats not documented)","Internet connection for cloud processing","Liarliar account with audio analysis tier (if separately priced)","Liarliar account with batch processing capability","CSV, JSON, or text file with statements to analyze","Internet connection for upload/download","Email address to create account"],"failure_modes":["No peer-reviewed validation of accuracy; claimed capabilities lack scientific evidence and contradict established deception research showing AI accuracy barely exceeds chance (50-55%)","Produces high false positive rates that could wrongly flag truthful statements, damaging innocent individuals' careers and relationships","No transparency on training data, model architecture, or validation methodology; inability to audit or understand decision factors","Susceptible to adversarial inputs and gaming—users can learn to manipulate linguistic patterns to evade detection","No cross-cultural or multilingual validation; linguistic patterns vary significantly across languages and cultural communication norms","Relies on speech-to-text accuracy, which introduces compounding errors—transcription mistakes propagate into deception scoring","No evidence that prosodic features (tone, hesitation, pitch variation) improve detection accuracy; most deception research shows these signals are unreliable","Audio quality degradation (background noise, accents, speech impediments) directly impacts transcription and downstream analysis","No validation on diverse speaker populations; potential bias toward certain accents, speech patterns, or demographic groups","Privacy and consent concerns: recording and analyzing speech without explicit informed consent may violate wiretapping laws in many jurisdictions","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.05,"quality":0.22999999999999998,"ecosystem":0.25,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.35,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:31.446Z","last_scraped_at":"2026-04-05T13:23:42.564Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=liarliar","compare_url":"https://unfragile.ai/compare?artifact=liarliar"}},"signature":"gEVWkdVRmaNPg/DgDtiwhVmd2mI6V5uyhal1erJn74CyOfJHMp01NzPpeu0pkv789j3mUS9s8RggkrlMcQQTAw==","signedAt":"2026-06-22T05:35:49.680Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/liarliar","artifact":"https://unfragile.ai/liarliar","verify":"https://unfragile.ai/api/v1/verify?slug=liarliar","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}