Skip to main content

Getting Started with VendorTruth

What VendorTruth is, what it's not, and how to actually use it.

What is VendorTruth?

Platform that helps B2B buyers see both sides of vendor claims before purchasing decisions.

What it does:

  • Runs parallel adversarial AI agents (prosecution + defense) to investigate vendor claims
  • Generates truth reports in 2-5 minutes with citations to public sources
  • Monitors vendor pricing/policy pages for changes (hourly for Pro, daily for Free)
  • Knowledge graph connects related vendor topics for discovery

What it doesn't do:

  • Replace due diligence—supplements it by automating research you'd manually do anyway
  • Verify claims requiring private data (customer testimonials behind login walls, confidential performance metrics)
  • Work well for new vendors with <6 months online presence (thin reports)

Bottom line: Accelerates vendor research by surfacing gotchas faster than manual Googling. Best for established B2B vendors with public footprints. Weak for stealth startups or vendors with sparse online presence.

How Dialectical Verification Works

AI spawns two research agents with opposing goals:

  1. Prosecution Agent - seeks evidence that challenges vendor claims (hidden costs, limitations, failures)
  2. Defense Agent - seeks evidence that supports vendor claims (innovations, advantages, success stories)
  3. Synthesis - combines findings into balanced verdict with recommendations

How it works:

  • Agents run in parallel (30-60 seconds each, streaming progress updates)
  • Recursively explore topics 1-5 levels deep using Exa API for web search
  • Generate follow-up questions using a small, fast language model optimized for speed
  • Cite specific URLs for every factual claim

Limitations:

  • Agents share the same underlying LLM and search API
  • Quality depends on public data availability
  • No human fact-checking before report publication
  • "Balanced" means both perspectives explored, not that both are equally valid

When this works:

  • Vendor has public documentation, blog posts, community discussions
  • You're evaluating claims with verifiable evidence ("supports SSO" vs vague "enterprise-grade")
  • You want fast adversarial research (minutes vs days)

When this doesn't work:

  • Vendor is new (<6 months) with minimal online footprint
  • Claims require insider knowledge or private benchmarks
  • Non-English vendor documentation (English bias in search)

Quick Start

1. Browse Existing Reports

Action: Visit Vendors to explore pre-generated truth reports.

What to expect:

  • ✅ Reports for established B2B vendors (AWS, MongoDB, Datadog, etc.)
  • ⚠️ Coverage gaps for niche vendors or new entrants
  • Reports are point-in-time snapshots (vendor situations change—check report date)

2. Generate New Truth Report

Action: Use Check a Vendor to verify a claim or vendor.

How it works:

  • Takes 2-5 minutes for comprehensive report
  • Streams live progress (you see questions being researched)
  • Free tier: 10 reports/month, Pro: 500 reports/month

Limitations:

  • Report quality varies based on vendor's public data availability
  • Counted against your quota regardless of report quality

When to use:

  • Pre-purchase evaluation of specific vendor
  • Validating bold claims from vendor sales calls
  • Building decision documentation for stakeholders

When to skip:

  • Time-critical decisions (<5 minutes to decide)
  • Vendor with zero public presence (you'll get "insufficient data" report)

3. Set Up Vendor Monitoring

Action: Create account → Add vendors to watchlist → Receive email alerts when changes detected.

How it works:

  • Monitors pricing pages, documentation, terms of service
  • Pro tier: hourly checks, Free tier: daily checks
  • Email notifications when changes detected

Limitations:

  • Can't monitor changes behind login walls (vendor dashboards, customer portals)
  • Alert significance is AI-judged—may flag trivial changes or miss critical ones
  • No built-in diff viewer (alert says "changed" but you manually compare)

When to use:

  • Tracking vendors you're already using (catch price increases before renewal)
  • Long evaluation periods (weeks/months) where vendor situation might change
  • Compliance need to document vendor policy updates

When to skip:

  • Short evaluation cycles (days)—manual checking is faster
  • Vendors with frequent changes (alert fatigue)

What You Actually Get

Truth Reports Include:

✅ Included:

  • Claim rating (True / Mostly True / Misleading / False / Unverified / Mixed)
  • What's True (validated facts from defense research)
  • What's False or Unclear (limitations from prosecution research)
  • Strengths and Weaknesses
  • When to choose vendor vs alternatives
  • Source URLs for all factual claims

❌ Not Included:

  • Market share data (not in our sources)
  • Private customer testimonials (can't access)
  • Historical pricing trends (no time-series tracking)
  • Direct vendor response to findings (no vendor participation)

Monitoring Alerts Include:

✅ What you get:

  • Notification that vendor page changed
  • Link to current page
  • AI summary of change significance

❌ What you don't get:

  • Before/after diff view
  • Root cause analysis (why vendor made the change)
  • Impact assessment specific to your use case
  • Automatic cancellation suggestions (just FYI alerts)

Knowledge Graph Includes:

✅ What works:

  • Discover related vendors organically
  • Click inline links to auto-generate deeper reports
  • Full-text search across published reports

❌ What's missing:

  • Comprehensive pre-seeded vendor coverage (graph grows based on user clicks, has gaps)
  • Advanced filtering (basic force-directed graph, no clustering/facets)
  • Semantic search (keyword-based, may miss conceptually related content)

Common Pitfalls

Pitfall 1: Expecting instant results

Reality: Reports take 2-5 minutes to generate. This is fast compared to manual research (hours/days) but not instant.

Pitfall 2: Trusting reports blindly

Reality: Reports synthesize public data using AI. You should validate critical claims by checking cited sources.

Pitfall 3: Using for brand-new vendors

Reality: Vendors <6 months old often lack sufficient public data. You'll get "insufficient evidence" reports. Use customer references instead.

Pitfall 4: Expecting "real-time" monitoring

Reality: Pro tier checks hourly. Daily for Free tier. Not sub-second real-time. Critical changes may have hours of latency.

Pitfall 5: Assuming comprehensive coverage

Reality: We only monitor pages we know about (pricing, docs, ToS). Vendor changes in dashboards, private forums, or unlisted pages won't trigger alerts.

Next Steps