Our Methodology
Every BiasChecker.ai analysis is an AI-assisted close reading of an article, run under a fixed set of instructions we have refined over thousands of analyses. The instructions are the same whichever AI model you choose — a different model changes who does the reading, not the instructions.
This page introduces the thirteen analysis types — we call each one a lens — and how to read the verdict each one returns. For the exact questions behind every analysis — and the ground rules they all share — see the questions behind every analysis.
We read the page, not the person
Every analysis describes the writing — the words on the page and the patterns in them. We tell you what a piece is built to make you see, feel, and conclude, and how it goes about it. That is a property of the text, and you can check it against the quotes we cite.
What we deliberately do not do is claim to know the author’s private intentions or state of mind. Naming who is biased, or asserting why someone wrote something, would be mind-reading — and unfair to the writer. So every lens is written about “the article” or “the text,” never the journalist or publisher: “the piece steers you toward…”, not “the author wants to hide…”. The same rule is why a finding is always anchored to a quote you can read for yourself, rather than to a motive we have guessed.
How to read a verdict
Most lenses lead with a verdict — a quick, at-a-glance read of the result. Like every rating on the site it is computed by our code from what the lens found, never asked of the AI model. It comes in two forms; each lens entry below says which form it uses and what its levels mean for that lens.
Concern & Evidence chips
Persuasion, Bias Analysis, Omissions, Scientific Assessment and Synthesis lead with two chips:
- Concern — None, Low, Medium or High — how much the lens found worth flagging. For Bias and Persuasion every finding is weighted by the fixed seriousness of its category or technique; for the other lenses it is simply how many findings there are (each lens says which findings count). Broadly, None means nothing notable, Low one isolated point, Medium a handful, and High a piece dense with them.
- Evidence — Limited, Moderate or Strong — how much verified, quoted support backs those findings.
Verdict headlines
Critique, Legal Risk and Moral Lens lead with one of a few fixed phrases instead of the chips: for these lenses one decisive problem matters more than several background notes, so a count would mislead.
The remaining lenses (Intent Analysis, Background, Historical Parallels, Neutral Rewrite and Roast) carry no verdict at all — they exist to inform, reframe, or entertain, not to flag concerns.
Indicators, not accusations. These are automated indicators from an AI reading of the text — not facts, accusations, or professional legal, medical, financial or scientific advice. A higher level reflects how much a lens flagged in the writing, which you can always inspect through the cited evidence and judge for yourself; it is never a statement that any person did something wrong. Different AI models may surface slightly different findings on the same article. See our Terms for the full disclaimer.
The thirteen analyses
Each analysis type gives the model a different expert role and a different question to answer about the same article. They fall into four families: Slant (how the piece leans and steers), Substance (whether what it argues holds up), Context & Extras (background, historical parallels, a neutral rewrite and a roast), and Synthesis (an analyst's report written from the rest). For a closer look at the exact questions behind each one, see the questions behind every analysis.
Slant— Spot the slant
Intent Analysis
Identifies what the text says versus what it steers readers to believe or do. Intent is treated as a property of the writing — the direction the text pushes — never as a claim about the author's private motives. The result names the text's primary purpose and direction, and gives you a quick verification question you can check for yourself in seconds.
Verdict: none; this lens informs rather than flags.
Persuasion
First asks whether the writing pushes you further than its facts support; straight reporting usually doesn't, and then nothing is listed. Where it does, each passage that pushes is named as one of eight persuasion techniques, with the exact quote and how it works on the reader. A passage that does not clearly fit a technique is left out.
Verdict: Concern & Evidence chips (see above). A rough measure of how hard the piece works to move you. Each technique found carries a fixed weight from our catalogue, so the level reflects how serious the techniques are, not just how many. Low: a minor technique or two. High: several serious persuasion techniques stacked together — often a sign that the conclusion the piece steers you toward isn't the one its own facts (or its headline) actually support. It flags that techniques are present, not that anyone is being dishonest.
Bias Analysis
Describes how the text leans, with reference to specific words and structural choices, across our catalogue of bias categories — framing, omission, loaded language, source imbalance, political slant and more. Each finding cites the passages that drive it.
Verdict: Concern & Evidence chips (see above). A rough measure of how much the framing — rather than the bare facts — shapes the impression you come away with. Each finding is weighted by how impactful its bias category is. Low: a few minor or easy-to-spot leanings. High: slant runs through the piece across several high-impact techniques (framing, omission, loaded language), so it reads more like a particular take than neutral reporting. A high level is about emphasis and selection — it does not mean any individual fact is false.
Omissions
The "what's missing" lens: what a well-informed reader would expect to see but the article leaves out, why each gap changes the impression the article leaves, and what kind of source would fill it. Only well-established context that would change your reading counts, so a thorough, balanced article comes back with an empty list. Where the article may have a serious flaw that cannot be stated as fact, the lens may add one or two neutral "questions worth asking" so you can weigh it yourself. Before you see the result, a second model checks each claimed gap against the article and removes any the article does in fact cover.
Verdict: Concern & Evidence chips (see above). A rough measure of how complete the picture is. Low: one notable gap. High: several consequential things a well-informed reader would expect are missing — so treat the article as a partial view and seek out the missing context. Only gaps that distort the picture raise the level; gaps that are merely useful background are listed but do not.
Substance— Check the substance
Critique
The "does it hold up?" lens. Instead of cataloguing rhetorical tricks, it evaluates the substance of the approach, argument or plan: are the goals right, is the approach sound, is it feasible, and what is being missed — taking on the strongest version of the argument before criticising it.
Verdict: Leads with one of holds up, holds up partly, or doesn't hold up as presented. Each problem it raises is marked as either fatal to the conclusion or a way to strengthen an argument that already stands; only the fatal ones set the headline, so a piece that holds up with a few polish suggestions still reads as holding up.
Legal Risk
Assesses whether actions or proposals described in the text raise questions under applicable law, working out the relevant jurisdiction from the text itself. About people and unverified facts it stays conditional: it says what would need checking.
Verdict: Leads with one of no red flags, context noted, or legal questions raised. Each finding is either a red flag — a specific party's conduct of the kind courts or regulators act on — or background legal context, and any red flag sets the headline. It is not a finding that anyone broke the law, only that, if the article is accurate, these are points a lawyer might want to examine.
Moral Lens
Judges two things separately: how the writing treats its subjects and its readers, and how the actions it describes measure against widely shared moral principles.
Verdict: Leads with a two-part sentence, for example “the reporting is fair; the events it describes raise ethical concerns”, so neutral coverage of grave events is never presented as a flawed article. The judgements are conditional and aimed at the text and the actions described — never at the character or motives of a named person.
Scientific Assessment
Checks the scientific rigor of the text: whether claims are supported by the evidence offered, how methodology and statistics are used, what is settled science and what is genuinely disputed, and where the text claims more than its own sources support.
Verdict: Concern & Evidence chips (see above). A rough measure of how much to trust the science as presented. The level follows the number of rigor problems found. Low: a minor issue such as overstated certainty. High: several rigor problems — unsupported causal claims, misused or cherry-picked evidence — so the scientific claims warrant independent checking before you rely on them.
Context & Extras— Get the context & extras
Background
A neutral briefing for someone landing mid-story: what this topic is, how it developed, and where it stands now. Pure orientation — factual, concise, and without taking sides — rather than a critique of the article. Where the public record allows, it adds a "what was promised versus what happened" note: early forecasts or pledges set against how they turned out.
Verdict: none; this lens informs rather than flags.
Historical Parallels
A historian's lens: finds past events and patterns that genuinely parallel the situation described — analogies, not a timeline of the topic itself — and explains how each parallel resembles the present case and how it played out, ranked by relevance.
Verdict: none; this lens informs rather than flags.
Neutral Rewrite
The article rewritten neutrally and more briefly: the loaded language and persuasion stripped out, the padding cut, and every fact, quote, and attribution kept exactly as the original had them. It only removes — it never adds a fact, a figure, or context that wasn’t in the article, so you can read it as a shorter, spin-free version and check it against the original.
Verdict: none; this lens informs rather than flags.
Roast
A warm-hearted comedian doing a stand-up bit about the article — a punchy tagline, a playful retelling, and a few standalone jokes about its most roastable details. The bar is whether it’s genuinely funny, not whether there’s a flaw: a dry, clean piece gets nothing to roast, and tragedies or sensitive human stories are never joked about.
Verdict: none; this lens informs rather than flags.
Synthesis— Then let it all come together
Synthesis
An analyst's report on the article, written by combining the other analyses' stored results — never a fresh, independent read of the text. It always rests on the four core slant lenses, plus any further lenses you add on the setup card; analyses you already have are reused, and nothing runs until you confirm. The report leads with a one-sentence "what this article is doing", connects the findings into one story about how the piece works on its reader, credits what checks out, and closes with a note on how to read it. Any passage our defamation check rejects is left out, and the report says how many were. How it chooses its inputs and works together with Consensus is covered under Synthesis below.
Verdict: Concern & Evidence chips (see above). Taken from the inputs: the highest Concern any of them reached, except that High needs two inputs at High — so one alarmed lens raises the report to at most Medium, while its own finding stays on its own tab. Evidence counts the quotes carried over from the inputs.
Synthesis — one report, a panel of analyses
Every other analysis reads the article. Synthesis reads the analyses this article already has, and never re-reads the article to form opinions of its own.
- Built on the four core lenses. Synthesis always takes Intent Analysis, Persuasion, Bias Analysis and Omissions as inputs. The setup card runs any of them the article does not yet have, and lets you add the other lenses as optional inputs. If some inputs fail, the report is still written from the rest, but never from fewer than two: with fewer it declines, says so, and costs nothing.
- Which analysis of each lens it uses. For each lens it prefers the result from the model you chose for the synthesis; failing that, the best stored result from another model, and a Consensus result ahead of any single model's. So different sections of one report can rest on different models' work; each input's model is recorded in the result.
- Nothing invented. Every section cites the analyses it rests on, and its supporting quotes are reused from those analyses' own evidence, matched by our code: a quote that cannot be traced back to one of the inputs is dropped rather than printed.
Synthesis and Consensus do different things, and they can be combined. Consensus takes one analysis and runs it across several models, reporting where they agree. Synthesis takes one model and runs it across several analyses, reporting how they fit together. Run the lenses you care about as Consensus first and the two combine: Synthesis uses those panel results ahead of any single model's, so the report rests on what several models agreed.
Consensus — one analysis, a panel of models
Any single AI model has blind spots. For Bias, Persuasion, Legal Risk, Moral Lens and Scientific Assessment you can run Consensus instead: several models from different makers each analyse the article independently, and a separate model merges their findings into one result, under the same instructions as a single-model run.
- One seat per maker. Models are grouped by the company that built them (Anthropic, OpenAI, Google, DeepSeek, Z.ai, Moonshot AI, xAI) and the panel seats one model per group. Two models from the same maker share training data and house style, so their agreement is not a second opinion.
- Three makers or more; cached seats are free. The panel starts with one model from each of three makers, and the setup card lets you swap a seat's model or add makers. Any analysis the article already has from a maker's model fills that seat at no cost, and every maker the article already covers joins, so a well-analysed article can seat four or five without buying more. Seats you leave empty are filled by Claude Haiku 4.5 (Anthropic), GPT-6 Luna (OpenAI) and DeepSeek V3.2 (DeepSeek); Gemini 3 Flash performs the merge. We review this line-up as models change.
- Agreed, or raised by one model. A finding enters the consensus only when at least two panel members reported it independently; our code counts that, not the merging model. Each agreed finding says how much of the panel backed it ("3 of 4 models"). Findings only one model raised are kept under Raised by one model, with the model named and its evidence intact: it usually means the others were looking at other parts of the article, not that they disagreed. If the panel agreed on nothing, the result says so; if it split on the headline verdict, the result shows the split.
- Same scoring, full record. The merged result is scored exactly like a single-model result, quotes verified and verdict computed by our code, and records which models sat, which were reused from cache and which merged. A consensus needs at least three members to answer: if fewer do, the analysis reports the failure so you can run it again.
- Strengthen. A stored consensus is never changed behind your back. If the article later gains analyses that would make a stronger panel, from a new maker or a stronger model of one already seated, the result shows a Strengthen button that says which models it would add or upgrade and what the rebuild costs, usually only the merge pass.
Where the panel agrees is the strongest read an analysis can give. Pricing and when to choose Consensus are on the models page and in the FAQ.
Publisher content & community analyses
An analysis is read in your own browser, at your request, from the page you are looking at. What we keep and share afterwards is deliberately minimal:
- We store our analysis, not the article. Community-shared results contain our AI-generated commentary plus short positional markers that let your browser re-locate the quoted passages on the live page — not the article text itself. Short excerpts shown alongside findings are quoted for analysis and review, always attributed and linked to the source.
- The community feed is curated and signed-in only. Shared analyses surface only inside the product to signed-in users, and only for outlets on a hand-maintained list. Before an outlet goes on that list we check whether it has opted out of text-and-data mining (a TDM Reservation Protocol file,
tdmrep.json, an ai.txt file or anoaidirective); an outlet that has opted out stays off the shared feed. - Disputes are heard. Every shared analysis can be rated by readers, and analyses enough readers flag as inaccurate are hidden automatically. Publishers can reach us via the contact page to dispute or request removal of an analysis of their content.
Analyses are AI-interpreted patterns and opinions about how a text argues — never verified facts about the publisher, and never a fact-check. See the terms of use for the full framing.