Ratings Criteria

Learn more about how we rate bias and bring clarity to the news. Find out how our platform works and why understanding media bias matters for informed decision-making.

MediaPedia Bias Ratings Criteria & Measurement Framework

The following criteria, created by a panel of media experts, determine how our AI model rates the bias of each journalist’s story. A journalist’s overall rating is determined by the ratings average of the stories they’ve covered and that we’ve analyzed, going back as far as ten years. Outlets are rated using an average of the reporters ratings at the outlet.

Definition: Measures whether the story presents both sides of an issue fairly, with comparable space, placement, and depth.

Indicators:

  • Number of quotes from each side.
  • Placement of quotes (early/middle/late in story).
  • Depth/detail given to each side's arguments.

Measurement:

  • Compare total words/seconds devoted to each side.
  • Compare number of sources from left/right perspectives.
  • Adjust score toward the side that gets more favorable coverage.
  • Scale: -50 (all left perspective) -- 0 (balanced) -- +50 (all right perspective)

Definition: Measures the presence of loaded, partisan, or politically motivated terms.

Indicators:

  • Use of ideological labels ("far-left," "woke").
  • Politically charged euphemisms ("undocumented" vs. "illegal").
  • Identity-based language shifts (e.g., "pregnant people" vs. "pregnant women").

Measurement:

  • Automated detection of partisan terminology lists (customizable for left/right terms).
  • Context analysis for whether terms are used descriptively or pejoratively.
  • Scale: -50 (strong left-coded terms) -- 0 (neutral/precise) -- +50 (strong right-coded terms)

Definition: How the story frames the issue—who is portrayed as victim, villain, or hero—and whether it assumes a conclusion.

Indicators:

  • Cause-effect implications favoring one side.
  • Selection of "villains" and "heroes" by political identity.
  • Victimization framing of certain groups.

Measurement:

  • NLP sentiment analysis applied to political actors/groups.
  • Comparison of sentiment scores for left/right subjects.
  • Scale: -50 (framing favors left narrative) -- 0 (neutral framing) -- +50 (framing favors right narrative)

Definition: Whether one side's claims are scrutinized more heavily than the other's.

Indicators:

  • Fact-checks applied only to one side.
  • Counterarguments presented only against one perspective.
  • Use of "experts" to rebut only one side.

Measurement:

  • Tag and count rebuttals per political side.
  • Track frequency and tone of fact-check statements.
  • Scale: -50 (only right is fact-checked) -- 0 (balanced scrutiny) -- +50 (only left is fact-checked)

Definition: Which sources are chosen, their ideological lean, and how they're presented.

Indicators:

  • Overreliance on partisan or ideologically aligned think tanks.
  • Use of anonymous sources without balance.
  • Labeling or omitting partisan identity of sources.

Measurement:

  • Classify sources by political leaning (left/center/right) via known-source database.
  • Track proportion of quotes by side.
  • Scale: -50 (sources overwhelmingly left-leaning) -- 0 (balanced sourcing) -- +50 (sources overwhelmingly right-leaning)

Definition: Key facts and important context is either ignored or emphasized, depending on how the reporter wants to portray the story.

Indicators:

  • Lack of context or missing facts that would alter perception.
  • Highlight missing perspectives or omissions.

Measurement:

  • Detect omission of relevant, documented facts present in other credible reports.
  • Scale: -50 (coverage patterns favor left agenda) -- 0 (balanced selection) -- +50 (coverage patterns favor right agenda)

Definition: Whether the headline's tone, sentiment, and framing favor one side or exaggerate.

Indicators:

  • Emotional or sensational language.
  • Headline sentiment vs. article content mismatch.

Measurement:

  • Sentiment scoring of headline vs. body.
  • Detection of loaded language in headline.
  • Scale: -50 (headline slants left) -- 0 (neutral headline) -- +50 (headline slants right)

Definition: Overall positive or negative tone toward political figures, parties, or movements.

Indicators:

  • Positive descriptors for one side vs. negative for the other.
  • Use of pejoratives or accolades.

Measurement:

  • Apply sentiment analysis to all named political actors.
  • Calculate net sentiment difference between left and right actors.
  • Scale: -50 (positive toward left, negative toward right) → 0 (equal tone) → +50 (positive toward right, negative toward left)

  • Each criterion is rated -50 to +50.
  • The Overall Bias Score = average of all category scores (weighted equally by default).
Interpretation:
  • -50 to -30 → Strong left bias
  • -29 to -10 → Moderate left bias
  • -9 to +9 → Center / minimal bias
  • +10 to +29 → Moderate right bias
  • +30 to +50 → Strong right bias

CriterionDefinitionMeasurement MethodBias DirectionScales
1. Balance of PerspectivesDoes the story present both sides fairly, with comparable space and depth?Count quotes & word counts per side; compare placement and depth of coverage.-50 = overwhelmingly left perspective, +50 = overwhelmingly right perspective-50 ← 0 → +50
2. Language & Terminology BiasUse of partisan, politically coded, or euphemistic terms.Detect partisan terms from lexicons; assess context (descriptive vs. pejorative).-50 = heavy left-coded terms, +50 = heavy right-coded terms-50 ← 0 → +50
3. Framing & Narrative ConstructionWho is portrayed as victim, villain, or hero; implied cause/effect.NLP sentiment toward actors; narrative role classification.-50 = framing favors left narrative, +50 = framing favors right narrative-50 ← 0 → +50
4. Fact-Checking & Scrutiny ImbalanceIs one side's claim scrutinized more?Tag rebuttals and fact-checks per side; measure frequency and intensity.-50 = only right fact-checked, +50 = only left fact-checked-50 ← 0 → +50
5. Sourcing BiasWho is quoted, ideological lean, and labeling of sources.Identify source political lean; count per side; track anonymous vs. named.-50 = sources mostly left-leaning, +50 = sources mostly right-leaning-50 ← 0 → +50
6. Selection & Omission BiasWhat topics/facts are included or ignored.Compare to peer coverage; detect missing documented facts.-50 = context favors left agenda, +50 = favors right agenda-50 ← 0 → +50
7. Headline Framing & SensationalismTone and sentiment of the headline vs. the article.Sentiment analysis; detect loaded terms; compare headline/body alignment.-50 = headline slants left, +50 = headline slants right-50 ← 0 → +50
8. Sentiment Toward Political ActorsTone toward political figures, parties, movements.Sentiment scoring by named entity; net difference between left/right actors.-50 = positive toward left, negative toward right; +50 = reverse-50 ← 0 → +50

1. Balance of Perspectives

Definition:

Does the story present both sides fairly, with comparable space and depth?

Measurement Method:

Count quotes & word counts per side; compare placement and depth of coverage.

Bias Direction:

-50 = overwhelmingly left perspective, +50 = overwhelmingly right perspective

Scales:

-50 ← 0 → +50

2. Language & Terminology Bias

Definition:

Use of partisan, politically coded, or euphemistic terms.

Measurement Method:

Detect partisan terms from lexicons; assess context (descriptive vs. pejorative).

Bias Direction:

-50 = heavy left-coded terms, +50 = heavy right-coded terms

Scales:

-50 ← 0 → +50

3. Framing & Narrative Construction

Definition:

Who is portrayed as victim, villain, or hero; implied cause/effect.

Measurement Method:

NLP sentiment toward actors; narrative role classification.

Bias Direction:

-50 = framing favors left narrative, +50 = framing favors right narrative

Scales:

-50 ← 0 → +50

4. Fact-Checking & Scrutiny Imbalance

Definition:

Is one side's claim scrutinized more?

Measurement Method:

Tag rebuttals and fact-checks per side; measure frequency and intensity.

Bias Direction:

-50 = only right fact-checked, +50 = only left fact-checked

Scales:

-50 ← 0 → +50

5. Sourcing Bias

Definition:

Who is quoted, ideological lean, and labeling of sources.

Measurement Method:

Identify source political lean; count per side; track anonymous vs. named.

Bias Direction:

-50 = sources mostly left-leaning, +50 = sources mostly right-leaning

Scales:

-50 ← 0 → +50

6. Selection & Omission Bias

Definition:

What topics/facts are included or ignored.

Measurement Method:

Compare to peer coverage; detect missing documented facts.

Bias Direction:

-50 = context favors left agenda, +50 = favors right agenda

Scales:

-50 ← 0 → +50

7. Headline Framing & Sensationalism

Definition:

Tone and sentiment of the headline vs. the article.

Measurement Method:

Sentiment analysis; detect loaded terms; compare headline/body alignment.

Bias Direction:

-50 = headline slants left, +50 = headline slants right

Scales:

-50 ← 0 → +50

8. Sentiment Toward Political Actors

Definition:

Tone toward political figures, parties, movements.

Measurement Method:

Sentiment scoring by named entity; net difference between left/right actors.

Bias Direction:

-50 = positive toward left, negative toward right; +50 = reverse

Scales:

-50 ← 0 → +50

Example Rating 1

Washington Post: “Trump’s Oval Office ‘session’ on the economy? Charts about Biden.

https://www.washingtonpost.com/business/2025/08/07/trump-oval-office-economy/
CriterionScoreReasoning
1. Balance of Perspectives-25The story quotes Trump, Moore, and a WH official, but quickly pivots to challenges and criticisms. No economists or analysts are quoted supporting Trump’s claims; multiple references to “experts” counter Trump’s narrative. Balance skews toward skepticism of Trump.
2. Language & Terminology Bias-15Subtle language choices (“pep rally,” “delighted,” “wanted to show off”) carry a mocking or dismissive tone. “Falsely accusing” is stylistically charged. Labels are minimal, but the choice of descriptors leans left-coded skepticism.
3. Framing & Narrative Construction-30Structure frames Trump’s presentation as unserious, data as unverified, and his policies as harmful. Biden-era complexity is mentioned, but largely to contrast with Trump’s claims. Victim/villain dynamic: Trump as showman, experts as truth-tellers.
4. Fact-Checking & Scrutiny Imbalance-40Trump’s claims are directly fact-checked (“no evidence to support his claim”), data verification questioned. Biden’s economic record is only briefly addressed with “far more complex story” but not scrutinized for accuracy.
5. Sourcing Bias-25Sources: Trump, Moore, WH official (anonymous), “experts say,” CBO, Commerce Secretary Lutnick. No pro-Trump economists or outside right-leaning analysts quoted besides Moore. Heavy reliance on institutional/governmental sources typically coded center-left.
6. Selection & Omission Bias-20The article emphasizes potential harms of tariffs, deficits, and Trump’s firing of a BLS official, but omits any supportive economic data from Trump’s second term outside Moore’s charts.
7. Headline Framing & Sensationalism-15Headline uses scare quotes around “session,” framing it as unserious. Focus on Biden in headline primes the reader to expect contrast unfavorable to Trump.
8. Sentiment Toward Political Actors-30Overall tone: skeptical/negative toward Trump, mildly defensive or neutral toward Biden. Biden’s economic performance is described as “complex,” not “bad,” avoiding strong negatives.

1. Balance of Perspectives

Score: -25

The story quotes Trump, Moore, and a WH official, but quickly pivots to challenges and criticisms. No economists or analysts are quoted supporting Trump’s claims; multiple references to “experts” counter Trump’s narrative. Balance skews toward skepticism of Trump.

2. Language & Terminology Bias

Score: -15

Subtle language choices (“pep rally,” “delighted,” “wanted to show off”) carry a mocking or dismissive tone. “Falsely accusing” is stylistically charged. Labels are minimal, but the choice of descriptors leans left-coded skepticism.

3. Framing & Narrative Construction

Score: -30

Structure frames Trump’s presentation as unserious, data as unverified, and his policies as harmful. Biden-era complexity is mentioned, but largely to contrast with Trump’s claims. Victim/villain dynamic: Trump as showman, experts as truth-tellers.

4. Fact-Checking & Scrutiny Imbalance

Score: -40

Trump’s claims are directly fact-checked (“no evidence to support his claim”), data verification questioned. Biden’s economic record is only briefly addressed with “far more complex story” but not scrutinized for accuracy.

5. Sourcing Bias

Score: -25

Sources: Trump, Moore, WH official (anonymous), “experts say,” CBO, Commerce Secretary Lutnick. No pro-Trump economists or outside right-leaning analysts quoted besides Moore. Heavy reliance on institutional/governmental sources typically coded center-left.

6. Selection & Omission Bias

Score: -20

The article emphasizes potential harms of tariffs, deficits, and Trump’s firing of a BLS official, but omits any supportive economic data from Trump’s second term outside Moore’s charts.

7. Headline Framing & Sensationalism

Score: -15

Headline uses scare quotes around “session,” framing it as unserious. Focus on Biden in headline primes the reader to expect contrast unfavorable to Trump.

8. Sentiment Toward Political Actors

Score: -30

Overall tone: skeptical/negative toward Trump, mildly defensive or neutral toward Biden. Biden’s economic performance is described as “complex,” not “bad,” avoiding strong negatives.

Overall Bias Score Calculation

Sum of scores:
-25 + (-15) + (-30) + (-40) + (-25) + (-20) + (-15) + (-30) = -200
Average: -200 / 8 = -25

Final Rating

Score: -25
Bias Classification: Moderate Left Bias

Interpretation:
The article provides factual context but frames Trump’s actions as unserious and largely unsubstantiated, applies heavier scrutiny to his claims than Biden’s, and uses subtle word choices that convey skepticism. While multiple Trump-aligned voices appear, they are outweighed by critical framing and omission of supportive perspectives.

Example Rating 2

NYT: “Trump Escalates a Fight Over How to Measure Merit in American Education”

https://www.nytimes.com/2025/08/08/us/trump-merit-affirmative-action-colleges.html
CriterionScoreReasoning
1. Balance of Perspectives-30While the piece quotes some conservative voices (Edward Blum, Students for Fair Admissions), the majority of quoted sources are from left-leaning academics, think tanks, or advocacy groups critical of Trump’s policy. Quotes opposing the executive order are more numerous, longer, and placed more prominently than supportive quotes.
2. Language & Terminology Bias-20Terms like “resegregate,” “quota system for wealthy and white students,” and “dangerous game” carry strong negative connotations toward Trump’s policy. The phrase “critics argue” is used for opponents, while supporters are often labeled “conservative movement” or tied to lawsuits.
3. Framing & Narrative Construction-35The article frames the order primarily in terms of harm to diversity and historical discrimination, foregrounding risks and negative outcomes, while support for merit-based measures is presented briefly and often through a defensive lens. Historical context emphasizes exclusionary uses of standardized testing without similar depth on merit-based arguments.
4. Fact-Checking & Scrutiny Imbalance-25The article heavily scrutinizes Trump’s position with research on income-test score links and racial disparities. Supportive claims are not similarly tested or bolstered with independent validation.
5. Sourcing Bias-35Heavy reliance on left-leaning or progressive organizations (EdTrust, Progressive Policy Institute, Education Reform Now, FairTest) and critical academics. Only one main pro-policy source (Blum) plus indirect mention of administration statements.
6. Selection & Omission Bias-25Strong emphasis on negative impacts to diversity and equity; little discussion of possible benefits of more quantitative measures beyond fairness to high-scoring applicants. No data or quotes from institutions or analysts who favor test-based admissions apart from Blum.
7. Headline Framing & Sensationalism-15Headline positions Trump as “escalating” a “fight,” which primes readers for conflict framing.
8. Sentiment Toward Political Actors-30Tone toward Trump and his administration is skeptical/negative; toward critics of the policy it is generally positive or neutral. The only Trump-aligned voice (Blum) is presented with minimal elaboration compared to critical sources.

1. Balance of Perspectives

Score: -30

While the piece quotes some conservative voices (Edward Blum, Students for Fair Admissions), the majority of quoted sources are from left-leaning academics, think tanks, or advocacy groups critical of Trump’s policy. Quotes opposing the executive order are more numerous, longer, and placed more prominently than supportive quotes.

2. Language & Terminology Bias

Score: -20

Terms like “resegregate,” “quota system for wealthy and white students,” and “dangerous game” carry strong negative connotations toward Trump’s policy. The phrase “critics argue” is used for opponents, while supporters are often labeled “conservative movement” or tied to lawsuits.

3. Framing & Narrative Construction

Score: -35

The article frames the order primarily in terms of harm to diversity and historical discrimination, foregrounding risks and negative outcomes, while support for merit-based measures is presented briefly and often through a defensive lens. Historical context emphasizes exclusionary uses of standardized testing without similar depth on merit-based arguments.

4. Fact-Checking & Scrutiny Imbalance

Score: -25

The article heavily scrutinizes Trump’s position with research on income-test score links and racial disparities. Supportive claims are not similarly tested or bolstered with independent validation.

5. Sourcing Bias

Score: -35

Heavy reliance on left-leaning or progressive organizations (EdTrust, Progressive Policy Institute, Education Reform Now, FairTest) and critical academics. Only one main pro-policy source (Blum) plus indirect mention of administration statements.

6. Selection & Omission Bias

Score: -25

Strong emphasis on negative impacts to diversity and equity; little discussion of possible benefits of more quantitative measures beyond fairness to high-scoring applicants. No data or quotes from institutions or analysts who favor test-based admissions apart from Blum.

7. Headline Framing & Sensationalism

Score: -15

Headline positions Trump as “escalating” a “fight,” which primes readers for conflict framing.

8. Sentiment Toward Political Actors

Score: -30

Tone toward Trump and his administration is skeptical/negative; toward critics of the policy it is generally positive or neutral. The only Trump-aligned voice (Blum) is presented with minimal elaboration compared to critical sources.

Overall Bias Score Calculation

Sum of scores:
-30 + (-20) + (-35) + (-25) + (-35) + (-25) + (-15) + (-30) = -215
Average: -215 / 8 = -26.9 → -27

Final Rating

Score: -26.9 → -27
Bias Classification: Moderate Left Bias

Interpretation:
The NYT article gives the appearance of thorough reporting with historical and statistical context, but the source selection, framing, and language tilt toward a critical view of Trump’s order. Pro-policy perspectives are present but sparse, less detailed, and often presented in rebuttal form, resulting in a moderate left-leaning bias in coverage.

Example Rating 3

AP: “Democratic Detroit lawmaker Joe Tate drops out of US Senate race”

https://apnews.com/article/joe-tate-exits-us-senate-race-michigan-c5849ac408b73d20da51a1c37560b751
CriterionScoreReasoning
1. Balance of Perspectives0Story is straightforward and reports facts about both Democratic and Republican sides without favoring one. Provides details about both parties’ candidates and campaign fundraising.
2. Language & Terminology Bias0No partisan or emotionally charged language. Terms are neutral and descriptive.
3. Framing & Narrative Construction0Framing is purely informational — no editorializing, no implied hero/villain roles. Both parties’ candidates are presented factually.
4. Fact-Checking & Scrutiny Imbalance0No claims in need of fact-checking; all details are numerical or directly sourced to candidates or public filings.
5. Sourcing Bias0Sources include Tate’s own statements and campaign finance reports. No reliance on ideologically skewed or partisan voices.
6. Selection & Omission Bias0Includes relevant information for both parties’ primaries. No notable omission of material facts relevant to understanding the race.
7. Headline Framing & Sensationalism0Headline is straightforward and factual, free from emotive or partisan framing.
8. Sentiment Toward Political Actors0Neutral tone toward all candidates. No adjectives or framing that convey approval or disapproval.

1. Balance of Perspectives

Score: 0

Story is straightforward and reports facts about both Democratic and Republican sides without favoring one. Provides details about both parties’ candidates and campaign fundraising.

2. Language & Terminology Bias

Score: 0

No partisan or emotionally charged language. Terms are neutral and descriptive.

3. Framing & Narrative Construction

Score: 0

Framing is purely informational — no editorializing, no implied hero/villain roles. Both parties’ candidates are presented factually.

4. Fact-Checking & Scrutiny Imbalance

Score: 0

No claims in need of fact-checking; all details are numerical or directly sourced to candidates or public filings.

5. Sourcing Bias

Score: 0

Sources include Tate’s own statements and campaign finance reports. No reliance on ideologically skewed or partisan voices.

6. Selection & Omission Bias

Score: 0

Includes relevant information for both parties’ primaries. No notable omission of material facts relevant to understanding the race.

7. Headline Framing & Sensationalism

Score: 0

Headline is straightforward and factual, free from emotive or partisan framing.

8. Sentiment Toward Political Actors

Score: 0

Neutral tone toward all candidates. No adjectives or framing that convey approval or disapproval.

Overall Bias Score Calculation

Sum of scores:
0 + 0 + 0 + 0 + 0 + 0 + 0 + 0 = 0
Average: 0 / 8 = 0

Final Rating

Score: 0
Bias Classification: Center / Minimal BiasModerate Left Bias

Interpretation:
This AP piece is a textbook example of neutral wire reporting: balanced treatment of both parties, factual data from official sources, and no framing that favors one side. It sits squarely in the center on the bias scale.