Skip to main content

​
Overview

Creating effective argument analyses is both an art and a science. This guide covers strategies for choosing topics, formulating inputs, and getting the most accurate and insightful results from the AI.
The quality of your analysis depends heavily on the input you provide and the availability of quality sources on your topic.

​
Choosing the Right Topic

​
Ideal Topic Characteristics

1

Genuine Controversy

Choose topics with legitimate arguments on multiple sidesGood: “Should social media platforms moderate political content?”Poor: “Is murder wrong?” (no genuine debate)
2

Sufficient Coverage

Topics need at least 5-8 quality sources for thorough analysisGood: Current policy debates, tech controversies, major cultural issuesPoor: Hyper-local issues, personal disputes, brand-new topics
3

Appropriate Scope

Neither too broad nor too narrowGood: “Universal Basic Income implementation challenges”Too Broad: “Economics”Too Narrow: “UBI pilot program in Stockton, CA, week 3 results”
4

Timely But Not Breaking

Recent enough for fresh sources, old enough for analysisGood: Topics from last 30 days with established discoursePoor: Stories from the last 2 hours (sources still developing)

​
Topic Categories That Work Well

  • AI regulation and safety
  • Data privacy and surveillance
  • Social media impact on society
  • Cryptocurrency legitimacy
  • Autonomous vehicle ethics
Why they work: High coverage, clear stakeholders, ethical dimensions

​
Input Formats

​
Format 1: Topic Query

The most common and versatile input type. 
"Should artificial intelligence be regulated by governments?"

"Universal Basic Income: Pros and cons"

"Climate change carbon tax effectiveness"

"Free speech vs hate speech on social media"

"Nuclear energy as climate solution"
Tips for queries:
  • Frame as a question OR state the controversy
  • Include key search terms (specific names, concepts)
  • Avoid yes/no questions when possible
  • Be specific but not overly technical
Avoid: “What do you think about AI?” (too vague)Better: “Should AI-generated art be copyrightable?”

​
Format 2: Article URL

Analyze a specific piece of content. 
https://www.bbc.com/news/technology-67890123

https://www.nytimes.com/2024/03/01/opinion/ai-regulation.html

https://www.reuters.com/world/climate-summit-2024
When to use URLs:
  • Analyzing a specific article’s arguments
  • Starting from a source you trust
  • Fact-checking a particular piece
System behavior:
  1. Scrapes the article content
  2. Identifies the central claim
  3. Searches for additional sources on the same topic
  4. Compares arguments across sources
The system will analyze the URL AND find supporting/opposing sources - you’re not limited to just that one article.

​
Format 3: Raw Text

Paste full text of an argument for analysis. Best for:
  • Debate transcripts
  • Academic essays
  • Opinion pieces
  • Social media threads (copy-paste)
  • Email arguments
Limitations:
  • 50,000 character limit
  • System may struggle to find corroborating sources
  • Best results when text includes citations
Paste full argument here:

"The case for universal basic income rests on three pillars.
First, automation is eliminating jobs faster than new ones are
created. Second, existing welfare systems are inefficient and
stigmatizing. Third, pilot programs in Finland and Kenya show..."

[Continue with full text]

​
Optimizing Analysis Quality

​
Pre-Analysis Checklist

1

Research Availability

Quick Google search: Does your topic have 5+ recent articles?If not, consider:
  • Broadening the topic
  • Choosing a more newsworthy angle
  • Waiting for more coverage
2

Refine Your Query

Test different phrasings:
  • “AI regulation” vs “artificial intelligence government oversight”
  • “UBI” vs “universal basic income” (use full terms)
  • Include synonyms if topic has multiple names
3

Check Recency

For time-sensitive topics, include year:
  • “2024 election polling accuracy”
  • “COVID vaccine effectiveness 2024”
4

Set Expectations

Understanding: Complex topics may take 30-45 seconds to analyze

​
During Analysis

Watch the progress indicators:
  1. 🔍 Searching web sources… (5-10s)
    • System finding articles via Firecrawl
    • Looking for 8+ diverse sources
  2. 📰 Scraping articles… (10-15s)
    • Extracting full text from URLs
    • Converting to markdown format
  3. 🐦 Fetching social sentiment… (3-5s)
    • Searching Twitter for recent tweets
    • Gathering engagement metrics
  4. 🤖 Analyzing arguments… (10-15s)
    • Gemini decomposing logical structure
    • Identifying claims and evidence
  5. ⚠️ Detecting fallacies… (5-8s)
    • Scanning for logical errors
    • Classifying fallacy types
  6. ✅ Generating blueprint… (2-3s)
    • Validating data schemas
    • Rendering visualization
If any step fails (API error, no sources found), the system will attempt fallbacks. You’ll see warning messages if the analysis is degraded.

​
Common Pitfalls & Solutions

Problem: “Web search returned 0 results”Causes:
  • Topic too niche
  • Misspelled terms
  • Very recent topic (< 24 hours)
  • Firecrawl API quota exceeded
Solutions:
  1. Rephrase with more common terms
  2. Broaden the topic slightly
  3. Try a related angle with more coverage
  4. Use URL input with a known article
Problem: Analysis returns but credibility is very lowCauses:
  • Few sources available (< 3)
  • Sources are opinion pieces, not news/research
  • Multiple critical fallacies detected
  • Echo chamber (all sources from one outlet)
Solutions:
  1. Choose a more mainstream topic
  2. Wait for more coverage to emerge
  3. Accept that topic genuinely lacks strong evidence
  4. Use Narrative Radar for pre-vetted topics
Problem: Blueprint shows mostly “for” or mostly “against”Causes:
  • Topic isn’t genuinely controversial (consensus exists)
  • Search query biased (“benefits of X” vs “X pros and cons”)
  • Sources happen to lean one direction
Solutions:
  1. Rephrase as neutral question
  2. Check if consensus is real (maybe issue is settled)
  3. Manually search for opposing views and analyze via URL
Problem: 10+ fallacies detected, feels excessiveCauses:
  • AI is overly sensitive (flagging rhetorical flourishes)
  • Sources genuinely use emotional/manipulative language
  • Confidence scores are high (AI is confident in detections)
Solutions:
  1. Filter by confidence (hide < 70%)
  2. Read the fallacy explanations - are they valid?
  3. Accept that some topics invite emotional arguments
  4. Use fallacies as learning, not absolute truth
Problem: “No tweets found for this topic”Causes:
  • Topic not discussed on Twitter recently
  • Search query doesn’t match Twitter language
  • Twitter API rate limit hit
Solutions:
  1. This is OK - not all topics need social data
  2. Try broader search terms
  3. Check if topic is genuinely niche
  4. Analysis still valid without social pulse

​
Advanced Techniques

​
Comparative Analysis

Create multiple analyses to compare:
  1. Same topic, different angles:
    • “UBI economic impact”
    • “UBI social impact”
    • “UBI political feasibility” Compare credibility scores and evidence types
  2. Evolution over time:
    • Analyze historical articles (paste text)
    • Compare to current Narrative Radar topic
    • Track how arguments evolved
  3. Source triangulation:
    • Analyze specific left-leaning source URL
    • Analyze right-leaning source URL
    • Analyze neutral topic query
    • Compare blueprints for bias

​
Using AI Chat for Depth

After generating analysis, use “Ask More” to: 
"What's the strongest evidence for the opposing side?"

"Which claims lack sufficient evidence?"

"Are there any unstated assumptions in claim #3?"

"What questions does this analysis leave unanswered?"

"Which fallacies are most damaging to the overall argument?"

"Compare the quality of sources for each side"

​
Combining with External Research

  1. Start with Argument Cartographer for overview
  2. Identify gaps (claims without evidence, missing perspectives)
  3. Research those gaps using academic databases, expert sources
  4. Create follow-up analysis with URL input of new sources
  5. Compare blueprints to see how picture evolves

​
Best Practices Summary

Choose Wisely

Controversial + timely + well-covered topics work best

Phrase Neutrally

“X vs Y” not “Why X is wrong”

Iterate

Try different phrasings if first attempt disappoints

Verify Sources

Click through and read original articles

Use Multiple Views

Switch visualization modes for different insights

Export & Share

Save high-quality analyses for future reference

​
Next Steps

Visualization Modes

Master all 6 ways to view your analysis

Exporting Results

Learn to export and share your findings

Narrative Radar

Explore pre-analyzed trending topics

Credibility Scoring

Understand how analyses are scored

Build docs developers (and LLMs) love

Get started for freeTalk to us