Cato Institute Media Bias



Overall framing & worldview

  • Unifying ideology: Across unrelated domains, the coverage repeatedly centers a libertarian–conservative “small government + markets + constitutional/procedural limits” lens—skeptical of expanding executive/regulatory power and favoring market mechanisms or rights-protecting constraints.

    Examples include framing election-related secrecy/declassification as “misuse of the classification system” , opposing broad federal authority in tariffs by citing Article I constraints , and arguing for deregulation/market access in securities and crypto orientation: Government institutions are often treated as inherently manipulable or overreaching (e.g., intelligence narrative “politically motivated” manipulation , CFPB “overreach” and “substantial overreach” ).

Coverage patterns (topic selection vs. rhetorical emphasis)

  • Topic selection: The sample disproportionately targets regulatory agencies and administrative power (CFPB: ; FCC/licensing: ; SEC rule design: ; consumer/tech regulation: ), trade/tariffs (historical tariff effects: ; limiting presidential tariff authority: ), and constitutional/legal allocation of powers (tariff delegation: ; declassification/election interference: ).
  • Rhetorical emphasis: Many pieces use consistent valor words and governance motifs (“individual liberty, limited government, free markets, and peace” appears as a recurring framing device in multiple records) . This suggests a stable editorial identity; it is not sufficient to conclude AI authorship, but it is evidence of template-like positioning.
  • Credibility strategies (with limits): Several summaries highlight external evidence or data sources (e.g., ProPublica/Texas Tribune ; GAO and scholars ; MARAD data comparisons ).

    Yet the strongest normative conclusions often come with limited visible counterevidence in the provided excerpts—e.g., “colossal mistake” or “unmistakable verdict” from inferred preferences .

Observable framing techniques (possible propaganda)

  • Loaded verdict language: The writing leans toward decisive moral/policy condemnations (“colossal mistake” , “plain wrong” , “clear indicator” of misuse agenda setting: Complex policy domains are framed primarily as overreach/distraction (AI hearings “too much… artificial intelligence… too little… everything else” ; CFPB discretion as the “fundamental defect” ).
  • Selective caution vs. selective certainty: Some items explicitly stress data limits and urge measured reforms (wrong-residence voting: “data limitations” and avoiding “lawfare” ; tariff research “dearth of historical research” noted ).

    Counterexamples show less caution in the summary record (e.g., “unmistakable verdict” ; “clear indicator” ).

Is there evidence of AI authorship?

  • Probabilistic, not determinative: The recurring slogan-like phrasing and consistently structured bias patterns across topics could reflect human editorial templating rather than AI generation.

    However, without full texts, there’s insufficient direct stylometry (e.g., repetitive n-grams, hallucination markers) to credibly conclude AI writing.

What quantitative signals appear?

  • Figures include: ~6,600 noncitizens misregistered in NJ ; Do Not Pay blocked >4,900 payments totaling $99 million and helped prevent/detect/recover $11.7 billion in potential improper payments (FY2025) ; and a tariff study analyzing 21 tariff changes (1840–2024) .
Limits of this analysis: The assessment is based on bias summaries and select quoted snippets, not full articles; factual reliability cannot be audited beyond what the provided records explicitly show.



Helium Bias: I assumed the repeated ideological phrasing in summaries represents the source’s consistent framing, but I did not verify full article text, method, or internal counterarguments.

The dataset may be skewed toward high-bias or politically salient pieces.

I also treated “propaganda” as observable rhetorical techniques rather than inferred intent, because the evidence provided lacks direct stylistic analysis or complete context.

Automated source summary · Updated July 26, 2026 · Not human reviewed. Check recent article panels for claim-level evidence when available.




Use the Data in AI All Sources

Cato Institute Bias Profile

Weighted source-level patterns from recent analyzed coverage. Open recent articles below to inspect score-specific evidence and limitations when available.

🔵 Liberal <—> Conservative 🔴10

🧢 Populist <—> Elitist 🎩10

🗽 Libertarian <—> Authoritarian 🚔-29

🗞️ Objective <—> Subjective 👁️ 13

🚨 Sensational25

💡 Boring <—> Interesting21

📝 Prescriptive60

😨 Fearful16

📞 Begging the Question8

💭 Opinion100

🗳 Political62

Oversimplification34

🏛️ Appeal to Authority30

🍼 Immature8

🔄 Circular Reasoning6

👀 Covering Responses26

😢 Victimization12

😤 Overconfidence28

🔒 Ideological100

🏴 Anti-establishment <—> Pro-establishment 📺-15

📏📏 Double Standard24

❌ Low Credibility <—> High Credibility ✅27

🧠 Rational <—> Irrational 🤪-17

💔 Low Integrity <—> High Integrity ❤️23

🪨 Low Intelligence <—> High Intelligence 🦉56

🎭 Virtue Signaling48

🐐 Scapegoating8

🤡 Hypocrisy6

👤 Individualist <—> Collectivist 👥-20

🎲 Speculation32

🐍 Manipulative70

Subtle dimensions

📉 Bearish <—> Bullish 📈0

🗣️ Gossip0

🤑 Advertising3

💣 Terrorism0

✊ Woke5

🔪 Cruel2

🔍 Truth-seeking <—> Delusion 🌀-2

🔺 Conspiracy5

⛓️ Anti-enlightenment0

🦊 Anti-Corporate <—> Pro-Corporate 👔2

🔬 Scientific <—> Superstitious 🔮-4

How to interpret source scores →

Average social shares per article 0



Cato Institute Political Bias (?)





Cato Institute Subjective Bias (?)





Cato Institute Opinion Bias (?)





Cato Institute Oversimplification Bias (?)








Click points to explore news by date. News sentiment ranges from -10 (very negative) to +10 (very positive) where 0 is neutral.





Cato Institute Recent Articles



Sort By: