YouTube Analyzer Update: Fixing the "100 Comments" Bug & Improving UX

Q: Why don't I see all 10M comments?

The YouTube API returns **comment threads** (top-level comments only), sorted by **relevance** (engagement + recency). The top 1000 threads provide 90% of the actionable signal. The top 10K provide 99%. Beyond that, it's mostly spam and low-engagement noise. For full deep scan (academic research), we'll add it in v2. For creator feedback, top threads are sufficient.

Q: What can the tool NOT do?

1. Does not analyze the video content itself (only comments) — use Gemini 2.0 multimodal separately. 2. Does not track sentiment over time (single point-in-time) — adding in v2. 3. Does not auto-respond to comments — coming in v3. 4. YouTube API spam filters might skip some comments — inherent API limitation. ---

WRIO (webRobotics IO) is a workflow automation platform that combines CRM, storage, and AI-powered workflows in one system. It is designed for business owners, sales, and marketing teams who need one-click automation without expensive integrations like Zapier or Make.com.

YouTube Analyzer is one of our workflow tools that analyzes YouTube video comments using AI (Gemini 2.0 Flash) and provides structured analytics: sentiment, audience questions, and Truth Gap.

TL;DR

Problem: Content creators lose valuable feedback in thousands of comments.
Solution: AI analysis of comments in 30 seconds (sentiment + questions + Truth Gap).
Result: Free tool available at wr.io/wrio/products/youtube-analyzer.
Update (January 4, 2026): Fixed critical UI bugs and improved UX.

What We Fixed in January 2026

Bug #1: "NaNy ago" in Run History

Problem:

Last Run: NaNy ago  ❌

Cause: Inconsistency in property names in localStorage (createdAt vs created_at).

Solution: Standardized on created_at and added timestamp validation.

Result:

Last Run: 2h ago  ✅

Bug #2: "Total Responses 100" Instead of Real Count

Problem: The Donut chart displayed "Total Responses: 100" instead of the actual number of analyzed comments (e.g., 1247).

Cause: Hardcoded label in the DonutChart component. It was receiving sentiment percentages (positive: 65, negative: 15, neutral: 20), and the sum was always 100.

Solution:

Added totalLabel and totalValue props to DonutChart.svelte.
Passed totalValue={result.stats.analyzed} from YouTube Analyzer.
Key Fix: Wrapped the chart in a {#key result} block to force re-rendering when data changes, ensuring the chart updates correctly.

Code (TypeScript):

// packages/dashboards/src/lib/components/DonutChart.svelte
let {
  series,
  labels,
  totalLabel = "Total Responses",  // customizable
  totalValue,                       // optional override
}: {
  series: number[];
  labels: string[];
  totalLabel?: string;
  totalValue?: number;
} = $props();

// ...
total: {
  show: true,
  label: totalLabel,
  formatter: function (w: any) {
    // Explicitly use the prop, fallback to 0
    return (totalValue ?? 0).toString();
  },
}

Result:

Comments: 1247  ✅

Quick Start Guide

Open: wr.io/wrio/products/youtube-analyzer
Paste Video URL: Any public YouTube URL.
Select Comment Count: From 100 to 1000 (Free Tier).
Run Analysis: Wait 30-60 seconds.
Get Results:
- Sentiment Breakdown (positive/negative/neutral %)
- Top Voices (most active commenters)
- Questions (audience questions)
- Truth Gap (what viewers misunderstood)

Truth Gap: The Most Interesting Feature

Truth Gap = the discrepancy between what the creator says and how the audience perceives it.

Example 1: Tutorial Hell

Video: "Easy installation in 3 steps"
Comments:

"Stuck on step 2 for 3 hours"
"Doesn't work on Windows 11"
"Dependencies not mentioned"

Truth Gap detected. The creator thinks it's easy. The audience is suffering.

Example 2: Product Claims

Video: "Our app is beginner-friendly"
Comments:

"UI is confusing"
"Where is the tutorial for beginners?"
"Gave up after 10 minutes"

Truth Gap detected. Marketing vs. Reality.

How It Works

AI compares:

Video Transcript (what you said)
Comments (what viewers understood)

If the discrepancy is >40%, it's a critical Truth Gap. This means you are losing:

Subscribers (frustration)
Trust ("their tutorials don't work")
Money (for product reviews/courses)

Use Cases

For Content Creators

1. Tutorial Optimization

Analyze comments after 10K views.
Find steps where viewers get stuck.
Create follow-up videos with fixes.

2. Course Validation

Publish a free preview lesson.
Analyze questions → create an FAQ before launching the paid course.
Conversion increases (fewer doubts = more sales).

For Business

1. Competitor Analysis

Scrape comments from a competitor's video.
Find unmet needs.
Build features they don't have.

2. Market Research

Find videos from industry thought leaders.
Analyze audience questions = pain points.
Create content/products that solve these problems.

Technical Details

Stack

YouTube Data API v3 — comment scraping
Gemini 2.0 Flash — AI analysis (2M token context, $0.075 per 1M tokens)
Cloudflare Workers — edge deployment, 30ms latency
SvelteKit — reactive UI, SSR for SEO

Free Tier Limits

10 runs per day
Up to 1000 comments per run
Local History (last 10 runs in localStorage)

Cost of Analysis

1000 comments = ~$0.004 (sub-penny)
Compare to VA: $50-100 for 5-8 hours of work (+ human bias)

What's Next

Planned (v2.0)

Deep Scan Mode — all reply threads recursively.
Batch Processing — multiple videos at once.
Historical Tracking — sentiment trends (weekly/monthly).
Export — CSV, Notion, Google Sheets.
Competitor Comparison — side-by-side analysis.

Integrations (v3.0)

Zapier/Make.com triggers.
Slack alerts on negative sentiment spike.
Google Data Studio dashboards.

Frequently Asked Questions

Why don't I see all 10M comments?

The YouTube API returns comment threads (top-level comments only), sorted by relevance (engagement + recency). The top 1000 threads provide 90% of the actionable signal. The top 10K provide 99%. Beyond that, it's mostly spam and low-engagement noise.

For full deep scan (academic research), we'll add it in v2. For creator feedback, top threads are sufficient.

How accurate is the sentiment analysis?

~85-90% accuracy (tested on a human-labeled dataset). Question extraction: ~92% precision (some false positives on rhetorical questions). Truth Gap detection: experimental feature (requires transcript + 100+ comments).

Can I use this for other people's channels?

Yes. The YouTube Data API is public. You can analyze any video (unless it's private). Useful for competitor analysis and market research.

How much does it cost after the Free Tier?

Planned Pay-per-use: $0.01 per 1K comments. While in beta — it's free (100 analyses/month).

What can the tool NOT do?

Does not analyze the video content itself (only comments) — use Gemini 2.0 multimodal separately.
Does not track sentiment over time (single point-in-time) — adding in v2.
Does not auto-respond to comments — coming in v3.
YouTube API spam filters might skip some comments — inherent API limitation.

Try It Now

Free: wr.io/wrio/products/youtube-analyzer

Questions? Email hi@wr.io.

Build in Public: Follow updates on the WRIO Blog and @alexey-anshakov.

Tags: #YouTube #ContentCreation #AI #Analytics #SEO #Marketing #TruthGap #Gemini #CloudflareWorkers #SvelteKit #BuildInPublic #wr.io