Consumer AI Authenticity Backlash Index (CAABI) Dataset: YouTube Audience Discourse on AI Deception Events — Deception Anger vs Trust Repair Seeking Corpus (n=2,477, April 2026)

This dataset contains 2,477 YouTube comments collected viathe YouTube Data API v3 (April 2026) across five deceptioncontext domains targeting naturalistic consumer backlashdiscourse following AI-related deception events:(1) AI generated fake reviews exposed brand 2025(2) AI influencer fake exposed brand deception 2025(3) company using AI to write fake reviews caught(4) AI generated content brand exposed misleading consumers(5) virtual influencer AI fake disclosure backlash 2025 The corpus is the primary empirical dataset for theConsumer AI Authenticity Backlash Index (CAABI) — a novelcomputational metric measuring consumer emotional andbehavioural response to discovered AI deception in brandand influencer marketing contexts. Three signal types — all conceptually distinct from priormetrics in the literature: DECEPTION ANGER (DA): betrayal emotion directed at AI deception — distinct from general AI anxiety (AAI), brand product distrust (TCIC), or governance distrust (ATBI)BRAND EXIT INTENT (BEI): active boycott/abandonment signal following discovered deceptionTRUST REPAIR SEEKING (TRS): consumer demand for verification, disclosure, and regulatory intervention CAABI = Deception Anger / (Deception Anger + Trust Repair Seeking)1.0 = pure Deception Anger (maximum brand damage, no recovery path)0.5 = anger balanced by repair seeking (crisis manageable)0.0 = pure Trust Repair Seeking (anger minimal, recovery possible) Grounded in:- Schilke & Reimann (2025) transparency dilemma- Ekinci et al. (2025) dark side of influencers- EU AI Act 2024 Article 50 disclosure requirements- Kim et al. (2004) trust repair theory- Weiner (1985) attribution theory of emotion KEY FINDINGS:- Total corpus: n=2,477 comments, ~50 videos- Categorised (CAABI computed): 128 (5.2%)- Mean CAABI: 0.9180 — Deception Anger overwhelmingly dominant- Deception Anger density: 0.3252/100 tokens (dominant)- Brand Exit Intent density: 0.0542/100 tokens- Trust Repair Seeking density: 0.0239/100 tokens (minimal)- Deception Anger cluster: 116 comments (90.6% of categorised)- Brand Exit Intent: 17 comments (13.3%)- Trust Repair Seeking: 9 comments (7.0%) — CRITICALLY LOW- Contested: 3 comments (2.3%) DOMAIN-LEVEL CAABI:- AI fake reviews exposed: CAABI=0.9706 (highest anger, minimal TRS)- AI misleading brand content: CAABI=0.9688- AI fake reviews caught: CAABI=0.9259- Virtual AI influencer backlash: CAABI=0.8750- AI influencer fake exposed: CAABI=0.6500 (LOWEST — most TRS) → Only domain where consumers seek repair over expressing anger THEORETICAL SIGNIFICANCE:Mean CAABI=0.9180 is the highest index score across allmetrics in this research programme (vs CARI=0.8224,IRI=0.9773 in resonance terms). This extreme DeceptionAnger dominance has a critical managerial implication:Trust Repair Seeking density (0.0239/100 tokens) isdramatically lower than Deception Anger (0.3252/100) —a 13.6:1 ratio. This means consumers who discover AIdeception predominantly express rage, NOT requests forrepair. Brand recovery pathways are structurally absentfrom the discourse. The AI influencer domain is the ONLYcontext where trust repair exceeds 1.0/100 tokens (3.73),suggesting virtual influencer deception is the one areawhere brands retain a recovery window. Files:- caabi_videos.csv: ~50 unique videos metadata- caabi_comments.csv: 2,477 raw comments- caabi_results.csv: CAABI scores + signal annotation Method: YouTube Data API v3. Python 3.12, April 2026.