How News Bias Apps Actually Rank Indian Media Houses

How News Bias Apps Actually Rank Indian Media Houses

TL;DR

Every bias-rating app you have ever used assigns labels using a mix of human panels, crowd surveys, and increasingly AI. None of them were built for India. The methodologies range from rigorous to "we just averaged what three other companies said," and understanding the difference is your best defense against trusting the wrong label.

The Label You Never Questioned

Open Ground News, AllSides, or any media bias app and you will see a clean, satisfying label next to every outlet: Left, Center, Right. Maybe a color-coded bar. Maybe a reliability score. It feels scientific. It feels settled.

It is neither.

Behind every "Left" or "Right" sticker on a news outlet sits a methodology. Sometimes it involves trained analysts reading hundreds of articles. Sometimes it involves an online survey where strangers rate headlines they skimmed for ten seconds. And sometimes, as the Columbia Journalism Review discovered, one company simply borrows another company's labels without formal permission and calls it a day.

The question is not whether bias ratings exist. It is whether you know what you are looking at when you see one.

India, where 43% of news consumers say they trust the news and 53% name WhatsApp as the biggest source of misinformation (the highest figure in any country surveyed by Reuters Institute), has a particular stake in getting this right. The country ranks 151st out of 180 on the RSF World Press Freedom Index. In a media ecosystem this contested, a misapplied label does not just mislead. It weaponizes.

The Big Three: How They Actually Work

Three organizations dominate the global media bias rating industry. Every app you use, from Ground News to SmartNews, either builds on their data or borrows it outright.

AllSides: The Crowd-Plus-Panel Approach

AllSides has been rating media bias since 2012, making it the oldest player in the space. Its approach blends multiple methods rather than relying on a single one.

The most distinctive tool is the Blind Bias Survey. Regular Americans read headlines and excerpts without knowing which outlet published them, then assign a political leaning. The idea is elegant: strip away brand loyalty and see what the text itself says. But as UCLA professor Tim Groeling told Poynter, expecting people to voluntarily consult bias charts and compare sources "would not actually be something people would do."

AllSides also conducts Editorial Reviews, where a panel of at least six reviewers (drawn from left, center, and right) reads six months of a source's output and looks for up to 16 distinct types of media bias, including spin, slant, sensationalism, bias by omission, and analysis presented as fact. They deliberate, vote, and arrive at a consensus rating on a five-point scale: Left, Lean Left, Center, Lean Right, or Right.

The result is a system that reflects "the average judgment of all Americans," as AllSides puts it. Note the word "Americans." That distinction matters enormously if you are trying to apply these ratings to an Indian outlet, where the political spectrum runs on entirely different fault lines.

AllSides has rated over 800 sources across its twelve-plus years. The staff composition, as reported by Poynter: 38% lean left or left, 29% center, 18% lean right or right.

Ad Fontes Media: The Two-Axis Chart

If you have seen a colorful diamond-shaped chart floating around social media, it was probably from Ad Fontes Media. Founded by Vanessa Otero, an intellectual property lawyer, Ad Fontes introduced a critical innovation: it rates outlets on two dimensions, not one. The horizontal axis measures political bias (from -42 on the extreme left to +42 on the extreme right). The vertical axis measures reliability (0 to 64).

This means an outlet can be biased but reliable, or centrist but unreliable. That nuance is genuinely useful.

The methodology involves over 40 trained analysts reviewing articles and episodes. Each piece of content gets rated by a "pod" of three analysts: one left-leaning, one centrist, one right-leaning. They score on seven categories of bias and eight of reliability, including expression, veracity, and political position.

The numbers are substantial. The January 2026 flagship chart features 137 sources (100 web/print, 19 podcasts, 18 TV/video programs). But the real database runs deeper: Ad Fontes has fully rated over 4,500 sources, including 2,840+ websites, 870+ podcasts, and 850+ TV/video programs.

In 2026, Ad Fontes began integrating AI alongside human analysts to rate content faster. The flagship chart now releases twice a year (January and August), with monthly updates for specific media categories in between.

But the chart has drawn fire. In 2021, Candice Benjes-Small of William & Mary and Nathan Elwood of the Missouri Legislative Library published a critique on the ACRL blog arguing the chart "promotes a false equivalency between left and right" and "lionizes a political 'center' as being without bias." Otero responded directly, acknowledging the chart condenses complex information but defending its methodology and inviting critics to explore the interactive version, which shows scatterplots of dozens of individually rated articles per source.

Media Bias/Fact Check: The Solo Reviewer

Media Bias/Fact Check (MBFC) is the largest database, with over 10,000 sources profiled. Founded in 2015 by Dave Van Zandt, it is also the most controversial methodologically.

MBFC uses a composite scoring system that rates each source on a scale from -10 (Extreme Left) to +10 (Extreme Right). The score comes from four weighted categories: Economic Policy (35%), Social Values (35%), Straight News Balance (15%), and Editorial Bias (15%). Each evaluation requires reviewing a minimum of 10 headlines and 5 full news stories.

Here is the part that makes academics uncomfortable: MBFC describes its own methodology as "not a tested scientific method". A 2018 Poynter Institute review noted that MBFC is "a widely cited source for news stories and even studies about misinformation, despite the fact that its method is in no way scientific." The Columbia Journalism Review called it "an armchair media analysis" with "subjective assessments that leave room for human biases, or even simple inconsistencies, to creep in."

Yet MBFC remains influential. Scientific studies have found its ratings show high agreement with independent fact-checking datasets, NewsGuard, and BuzzFeed journalist evaluations. It also added an AI policy in January 2026, requiring that humans remain ultimately responsible for editorial decisions even when AI tools are used in content production.

The Aggregator Problem: Who Watches the Watchmen?

Ground News, the app that bills itself as a cure for media bubbles, does not actually rate anything. It aggregates ratings from AllSides, Ad Fontes, and MBFC and averages them into a single score.

That sounds reasonable until you learn how loose the arrangements are.

CJR reporter Martina Di Licosa investigated in September 2025 and found a revealing patchwork. Alice Griesemer Sheehan of AllSides told CJR that Ground News uses their ratings without "formal permission" and without compensation. Dave Van Zandt confirmed MBFC has no agreement with Ground News either. Only Vanessa Otero of Ad Fontes confirmed that Ground News pays for its ratings.

Ground News declined to comment, writing, "We don't disclose ongoing business relationships and arrangements."

Kathleen Hall Jamieson, co-founder of FactCheck.org, offered a pointed observation about the entire enterprise: "When people say they want unbiased news, they mean 'I want news I don't recognize as biased.'"

Why None of This Works for India

Every system described above was built for the American political spectrum. Left means Democratic-leaning. Right means Republican-leaning. Center means somewhere between the two.

Indian politics does not work that way. The BJP is economically liberal in some areas and socially conservative. The Congress party combines economic populism with institutional centrism. Regional parties defy any left-right axis entirely. A DMK editorial and a TMC press release are not "the same kind of left." A Republic TV debate and an OpIndia article are not "the same kind of right."

The problem gets worse at the article level. AllSides and MBFC rate outlets, not individual articles. An NDTV op-ed on Kashmir and an NDTV business report on Sensex valuations will carry the same bias label even though they occupy completely different editorial territories.

Ad Fontes does rate individual articles, which is better. But with over 4,500 sources and counting, the India-specific coverage is thin. The January 2026 flagship chart does not even include the Associated Press or Reuters on its visual (though they are rated in the database). Indian outlets are almost entirely absent from the flagship.

This is not a minor gap. India has the world's largest number of smartphone news consumers, with 68% accessing news primarily through mobile devices. The Reuters Institute's 2025 Digital News Report found that YouTube (55%), WhatsApp (46%), Instagram (37%), and Facebook (36%) are the primary gateways for news in India. When Indian users encounter bias labels, they are almost always labels designed for a foreign political context and applied to outlets that those rating organizations barely cover.

What Researchers Are Building Instead

Academic researchers have started building India-specific approaches, though none have reached consumer-app scale yet.

A 2026 paper from Ashoka University introduced MediaGraph, a network-theoretic framework that sidesteps the left-right problem entirely. Instead of labeling outlets, it maps how outlets cluster around entities. During the 2020-21 farmers' protests, the researchers analyzed 7,271 articles from Times of India, 3,454 from Indian Express, 538 from DNA, and 878 from Firstpost. They found that "mainstream outlets maintain a fundamentally political elite-centric network structure" while fringe outlets diversify coverage with non-political figures. Crucially, "farmer leaders remain covered insignificantly" across all outlets, despite being the central actors in the protest.

This kind of structural analysis reveals something a bias label never can: it is not just about left or right. It is about who gets covered and who gets erased.

Separately, researchers at ACM published work on political bias detection from Hindi-language news, using tools like MuRIL (a multilingual BERT model) and Extended Hindi SentiWordNet. The challenge is formidable. As earlier researchers noted on ResearchGate, "In case of Indian political news data, it is not available at all." Most global bias-detection models are trained on English-language American political text. Applying them to Hindi, Tamil, or Marathi news is not just inaccurate. It is meaningless.

The MediaGraph researchers identified a deeper methodological concern: "NLP-based approaches provide scalable estimates of media bias, [but] they often rely on annotated datasets and proxy signals (e.g., sentiment, lexical choice, or framing cues), which introduce subjectivity in the definition and measurement of bias."

The India-Specific Approach

The Balanced News takes a different path. Rather than importing Western bias labels, it analyzes over 50 Indian news sources directly, using AI trained on Indian political context. Each article gets a political bias score (Left %, Center %, Right %) derived from word choice, framing, entity coverage, and sentiment analysis calibrated to Indian politics, where "Right" means pro-BJP, not pro-Republican.

The app also introduces the Lens Score, measuring the gap between a story's real-world importance and its media coverage. A score of 70-100 flags critical stories being actively suppressed. This addresses something none of the Western tools even attempt: not just how a story is framed, but whether it is being covered at all.

With 10,000+ users rating it 4.9 stars on both app stores, it represents one of the few consumer-grade tools actually built for the Indian news ecosystem. The methodology, correction history, and editorial decisions are documented publicly, which is more transparency than most Western bias-rating companies offer.

How to Read Bias Ratings Without Getting Fooled

Kelly McBride, NPR's public editor and chair of Poynter's Craig Newmark Center for Ethics and Leadership, offered the sharpest advice on bias charts: "Overreliance on a chart like this is going to probably give some consumers a false level of faith" in knowing what is reliable.

She is right. But ignoring bias ratings entirely is not the answer either. Here is how to use them without being used by them.

Check the scope. If the app was built for American politics, its labels for Indian outlets are directional at best. Ask whether the rating organization has analysts who read Hindi, Tamil, or Telugu. (Spoiler: most do not.)

Look at the methodology, not just the label. A rating from Ad Fontes, where three analysts across the political spectrum scored an article on seven bias categories, carries more weight than a MBFC rating based on five articles reviewed by one person.

Distinguish outlet ratings from article ratings. An outlet rated "Center" can publish a wildly biased op-ed. An outlet rated "Right" can produce excellent investigative reporting. The label tells you about the average, not the specific piece you are reading.

Watch for the aggregator dodge. If an app averages three other companies' ratings without disclosing how, you are looking at a composite of methodologies that may contradict each other. When AllSides rates AP as "Lean Left" and Ad Fontes rates it near center, averaging those two numbers together does not produce accuracy. It produces mush.

Follow the money. As CJR's Martina Di Licosa wrote: "Truth is not a measure of how far a story strays from other coverage; it's a measure of how near it comes to reality." If a bias app paywalls its factuality ratings while giving away bias labels for free, ask yourself which metric the company actually values.

The Rating That Does Not Exist Yet

The honest truth is that nobody has solved media bias measurement. Not AllSides with its twelve years of surveys. Not Ad Fontes with its 4,500-source database. Not MBFC with its 10,000 profiles. And certainly not the aggregators who average other people's work and call it insight.

What these tools do is give you a starting point. The best ones give you methodology pages you can actually read, transparency about who does the rating and how, and honest acknowledgment of their limitations. The worst ones give you a clean label and hope you never look underneath.

India's media ecosystem, with its 50+ major outlets spanning a dozen languages and a political spectrum that refuses to map onto a Western axis, deserves tools built from the ground up for its own reality. The global bias-rating industry is not going to build them. That work will have to come from Indian researchers, Indian journalists, and Indian technologists who understand that the space between Times of India and The Wire is not a line. It is a landscape.

Until then, the most reliable bias detector remains the one between your ears. Use it.

Sources

• Ad Fontes Media - January 2026 Flagship Chart - details on 137-source chart, 4,500+ total rated, analyst methodology
• AllSides - Media Bias Rating Methods - methodology overview including blind surveys and editorial reviews
• AllSides - How to Spot 16 Types of Media Bias - list of bias types used in editorial reviews
• AllSides - News Aggregators Bias Analysis 2026 - Google News 73% left-leaning finding
• Media Bias/Fact Check - Methodology - composite scoring, categories, self-described non-scientific method
• Ground News - Rating System - how bias bar aggregates third-party ratings
• Columbia Journalism Review - The Business of Balance - investigation into Ground News business model and ratings usage
• Poynter - Should You Trust Media Bias Charts? - expert critiques from Kelly McBride and Tim Groeling
• ACRLog - Complex or Clickbait?: The Problematic Media Bias Chart - academic criticism of Ad Fontes chart
• Ad Fontes Media - Response to ACRL Critique - Otero's rebuttal
• Reuters Institute - Digital News Report 2025: India - trust (43%), platform usage, press freedom ranking
• MediaGraph - arXiv 2026 - network-theoretic analysis of Indian media bias
• ACM - Political Bias Detection from Hindi News - MuRIL-based Hindi bias detection
• ResearchGate - Media Bias Detection Using Sentiment Analysis - Indian political news data gap
• The Balanced News - Media Bias Checker - India-specific bias detection methodology
• The Balanced News - app overview, Lens Score, 50+ source coverage