Methodology
How we analyze content and calculate the Complixo Score.
How It Works
When you submit content to Complixo, our AI engine analyzes it in multiple rounds:
Round 1 — Rhetorical Analysis
Extracts individual claims and scores each across 10 credibility dimensions. Identifies red flags, context signals, and positive findings.
Round 2 — Fact-Checking
Cross-references claims against external sources. Adjusts scores based on factual verification. Available for Plus subscribers.
Scope
Complixo analyzes factual claims about health, products, science, and finance. We check whether claims are supported by available evidence. We do NOT rate political opinions, religious beliefs, personal values, or artistic judgments. A low Complixo Score means the content contains rhetorical patterns commonly associated with misleading claims — it does not mean the person is dishonest or that the product doesn't work.
The Complixo Score
The Complixo Score ranges from 0 to 100. A higher score means more credible content.
How it's calculated:
- Each claim gets a weighted score (primary claims count 2×)
- Max-claim anchoring prevents one critical issue from being hidden by averages
- Context signals (disclosures, caveats) adjust the score up to ±10 points
- Final score is clamped to 0–100
The 10 Dimensions
Every claim is analyzed across these credibility dimensions:
1. Verifiability of Claims
Can the claims be independently verified? Are sources cited? Are references traceable to real publications or institutions?
2. Source Credibility
Are the cited sources reliable and authoritative? Do they have a track record of accuracy? Are credentials verifiable?
3. Logical Fallacies
Does the content contain logical errors such as false dichotomies, straw man arguments, circular reasoning, or appeals to authority?
4. Language Manipulation
Does the content use loaded language, weasel words, absolutist claims ('always', 'never', 'proven'), or clickbait patterns?
5. Emotional Manipulation
Does the content use fear, urgency, guilt, outrage, or artificial scarcity to persuade rather than inform?
6. Statistical Misleading
Are statistics cherry-picked, presented without context, or used to create misleading impressions? Are sample sizes adequate?
7. Omission & Selectivity
Is important context missing? Are counterarguments acknowledged? Is the presentation one-sided?
8. Authenticity
Does the content appear genuine or manufactured? Are there signs of astroturfing, bot-generated text, or coordinated campaigns?
9. Platform & Medium Context
Is the medium appropriate for the claims being made? Are health claims made in a product review? Are financial promises in a social post?
10. Factual Accuracy
Based on available evidence, are the specific factual claims in the content accurate, inaccurate, or unverifiable?
EU AI Act Compliance
Under the EU AI Act (Article 50), AI systems that generate content must be transparent about their nature. Complixo complies by:
- Clearly labeling all analyses as AI-generated on every page
- Publishing this methodology page describing how scores are calculated
- Never attributing intent or making moral judgments about people
- Using language like "analysis" and "assessment" rather than "truth" or "verdict"
- Providing sources and references for verification
Limitations
- AI analysis can produce errors. Always verify important claims independently.
- Round 1 is a rhetorical pattern analysis, not a full fact-check.
- Analysis is optimized for English content. Other languages may be less accurate.
- Scores reflect our methodology at a point in time — they may change as we improve.
- We analyze content, never people. We do not attribute intent.
- Cached analyses use the same ruleset version — a new version triggers fresh analysis.
Open Questions
If you believe an analysis is incorrect, you can challenge specific findings on any analysis page. We review all challenges and adjust scores when warranted.