Toxic Comment Index for Influencer Commerce (TCIC) Dataset: YouTube Audience Discourse on Influencer Product Reviews — Brand Distrust vs Purchase Intent Corpus (n=2,732, April 2026)

This dataset contains 2,732 YouTube comments collected viathe YouTube Data API v3 (April 2026) across four productcategory domains targeting naturalistic consumer discourseon influencer product review content:(1) beauty product review honest influencer 2025(2) tech gadget review influencer sponsored 2025(3) fitness supplement review influencer 2025(4) food product review influencer honest opinion 2025 The corpus is the primary empirical validation dataset forthe Toxic Comment Index for Influencer Commerce (TCIC) —a novel computational metric that, for the first time,distinguishes three distinct signal types in influencerproduct review discourse: INFLUENCER-DIRECTED TOXICITY (ITox): anger/distrust aimed at the influencer person — may not damage brandBRAND-DIRECTED DISTRUST (BDist): skepticism aimed at the product/brand — direct commercial damage signalPURCHASE INTENT (PI): expressed buying intention — positive commercial signal TCIC = Brand Distrust / (Brand Distrust + Purchase Intent)1.0 = pure Brand Distrust (commercial risk dominant)0.5 = Contested commercial space0.0 = pure Purchase Intent (commercial opportunity) Grounded in:- Ekinci et al. (2025) dark side of influencers- Di Domenico et al. (2026) source-amplified toxicity- Cheng & Zhang (2024) reputation burning effect- Ajzen (1991) Theory of Planned Behaviour- Fishbein & Ajzen (1975) attitude-behaviour link KEY FINDINGS:- Total corpus: n=2,732 comments, ~48 videos- Categorised (TCIC computed): 155 (5.7%)- Mean TCIC: 0.6032 (Brand Distrust dominant overall)- Brand Distrust density: 0.2113/100 tokens (highest)- Purchase Intent density: 0.1239/100 tokens- Influencer Toxicity density: 0.1096/100 tokens (lowest)- Brand Distrust cluster: 85 comments (54.8% of categorised)- Purchase Intent cluster: 59 comments (38.1%)- Influencer Toxicity cluster: 52 comments (33.5%)- Contested: 5 comments (3.2%) CATEGORY-LEVEL TCIC:- Tech: TCIC=0.7586 (highest Brand Distrust risk) BDist=7.94/100 | ITox=0.73/100 | PI=1.49/100- Beauty: TCIC=0.6194 (Brand Distrust dominant) BDist=2.84/100 | ITox=0.08/100 | PI=1.53/100- Fitness: TCIC=0.5200 (Contested — near parity) BDist=3.13/100 | PI=2.76/100 | ITox=0.00/100- Food: TCIC=0.5000 (Perfect balance — highest PI) BDist=2.30/100 | PI=3.64/100 | ITox=0.36/100 THEORETICAL SIGNIFICANCE:Tech reviews show highest TCIC (0.7586): consumersengage in more systematic product evaluation, generatingstronger brand distrust signals when quality fails.Beauty shows mid-high TCIC (0.6194): aestheticdisappointment translates to brand rather than influencerblame. Fitness (0.5200) and Food (0.5000) occupy thecontested zone: purchase intent nearly equals branddistrust, confirming that functional product categoriesactivate dual-directional consumer discourse.Crucially, Influencer Toxicity density (0.1096) isLOWER than Brand Distrust (0.2113): consumers blamebrands more than influencers in product review contexts,challenging the dominant assumption in influencermarketing research. Files:- tcic_videos.csv: ~48 unique videos metadata- tcic_comments.csv: 2,732 raw comments- tcic_results.csv: TCIC scores + signal type annotation Method: YouTube Data API v3. Python 3.12, April 2026.