You have seen the chart. A big triangle, or a grid, with news logos scattered across it: CNN a little left of the middle, Fox News out on the right, the wire services stacked up the center like a spine. Somebody posts it in an argument roughly every eleven minutes. It looks like a periodic table for journalism — settled, measured, done. * It is not a periodic table. It is three private companies, with three different methods, three different business models, and three different definitions of the word "bias," each publishing a picture that happens to look identical in shape and is not identical in meaning. Knowing how each one is actually made changes what you can honestly do with it. * ### The three charts, and how each one is built * **AllSides** rates outlets on a five-point left-to-right scale — Left, Lean Left, Center, Lean Right, Right — and its headline method is the Blind Bias Survey. Ordinary Americans across the political spectrum are shown headlines and article text with the outlet's name stripped out, and asked to rate the slant of what they just read. They do not know who published it. Those results are then normalized to better reflect the political makeup of the country, and combined with Editorial Reviews conducted by a multipartisan panel trained to spot common bias techniques. Readers can also vote that they agree or disagree with a rating, and enough disagreement can trigger a re-review. * **Ad Fontes Media** produces the chart with two axes instead of one. The horizontal axis is political position; the vertical axis is reliability. It was created in 2016 by Vanessa Otero, a former patent attorney who rated the first batch of sources herself, then rebuilt the process specifically to dilute her own judgment. Today each article, video or episode is rated by a team of at least three trained analysts — one who leans left, one who leans right, one in the center — working independently, and the scores are averaged. Ad Fontes says analysts complete at least 30 hours of training before they rate anything, plus 40 more hours every year they stay. Reliability is scored on expression, veracity and headline/graphic accuracy; bias on political position, language and comparison. * **NewsGuard** is the odd one out, because it refuses to place anyone on a left-right axis at all. Founded in 2018 by journalists Steven Brill and Gordon Crovitz, it scores sites 0 to 100 against nine criteria it calls apolitical: does the site repeatedly publish false content, does it disclose who owns and finances it, does it correct errors, does it separate news from opinion, and so on. Every criterion is pass-fail — you get all its points or none — and the weighted points add to 100. The result is published as what NewsGuard calls a Nutrition Label: the score plus a written explanation of the outlet's ownership, financing and record. NewsGuard's pitch is that it is rating process, not politics. * ### The thing all three have in common * They rate the brand. Not the story in front of you. * That distinction sounds academic until you look at what it did to NPR. In February 2022, AllSides ran a Blind Bias Survey on NPR and got an average score of -2.77, landing it in Lean Left territory and close to Left; normalized, 71 percent of respondents placed NPR left of center and 9 percent placed it right of center. AllSides moved the rating from Center to Lean Left. But read the fine print AllSides itself publishes: the rating applies only to NPR's online, written news content. It does not describe NPR's radio content, which is a different newsroom output entirely. * So the most-shared media bias rating of NPR in existence is a rating of one half of NPR, and the chart has no room to say so. The dot is a dot. It cannot hold an asterisk. That is not a scandal — AllSides disclosed it plainly. It is a limit built into the format, and it applies to every logo on every one of these charts. An outlet that runs excellent investigative work and a garbage opinion section gets one position. A wire service that files 800 stories a day gets one position. You get an average of things you were not planning to read, applied to the one thing you were. * ### What the critics say, and they are not who you think * The loudest objections to bias charts do not come from the outlets that scored badly. They come from journalism educators who think the whole axis is the wrong question. * Kelly McBride of the Poynter Institute has argued the charts hand consumers "a false level of faith" in the outlets that land in comfortable positions — pointing out that she can name a serious journalistic failure at just about every organization on the chart. Her preferred metric is not slant at all but the quality of the reporting, the writing and the editing. Tim Groeling, a communication professor at UCLA, has raised the more deflating problem: he doubts most people will actually do the work of consulting a chart and running the comparison, which is the only way the chart does anything. * Poynter's own examination noted the mechanical cost of a five-point scale — it forces a source into a box when the reality is a continuous spectrum — and that some nuance simply has to die the moment the answer becomes a graphic. * ### And then the government showed up * If you want proof that rating the news is a political act no matter how apolitical your criteria are, the last eighteen months supplied it. * In May 2025, NewsGuard received a civil investigative demand from the Federal Trade Commission — a sweeping legal order for records covering its rating system, its methodologies, its internal communications, and the identities of the entities and employees involved since 2018. FTC Chairman Andrew Ferguson had publicly suggested NewsGuard was biased against conservative publications and had raised antitrust theories, pointing among other things to the low score NewsGuard gave Newsmax. * On February 6, 2026, NewsGuard sued the FTC in federal court in Washington, D.C., arguing the demand was retaliation against speech the chairman disliked. On April 16, 2026, the FTC withdrew the demand, telling the court that its settlement with three advertising companies had eliminated its need for the records. * Read that sequence slowly. A company whose entire product is the claim that its criteria are apolitical spent eleven months defending itself from a federal regulator on the theory that its ratings were political. Both things can be true at once: the criteria can be genuinely mechanical, and the act of publishing a number next to a news brand can still be a fight about power. That is not NewsGuard's failure. It is the job description. * ### How to actually use a bias chart * Not as a verdict. As a starting position. * A chart can tell you the direction an outlet usually leans, which is genuinely useful when you are reading something from a source you have never heard of. It cannot tell you whether the specific article in front of you is accurate, whether the reporter did original work or rewrote a press release, or who benefits from the framing. For that you need the things the chart cannot plot: who owns the outlet, who funds it, whether the piece is news or opinion, and whether the sourcing is primary or secondhand. * Here is the test that has never failed us. Take any outlet on any of these charts and ask a question the chart has no axis for: who would have to be embarrassed for this newsroom to kill a story? A left-right position cannot answer that. An ownership record can, in about two minutes. * The chart is a map of where the logos sit. The interesting part is who is holding the map.