Frametra

Check news credibility by structure, not politics. Six-signal AI article analysis for journalists, students and researchers.

Frametra is a news credibility checker that reads how a story is built, not what it concludes. Most media bias checkers and fact-checking extensions hand you a verdict: a political label, a "credible" or "not credible" stamp, a color. Frametra shows you why. It measures the structural choices behind any article: how claims are attributed, how much evidence supports them, how language signals certainty or hedging, and how the piece is framed against the norms of its own genre. It is an AI-powered news analysis and media literacy tool for anyone who wants a clearer-eyed way to read the news, check a source before quoting it, screen coverage before a client sees it, or audit their own writing before it goes out. Click Analyze on any news article and get a real-time Structural Integrity Score in seconds. No account required. No political labels. No black box. WHAT MAKES FRAMETRA DIFFERENT FROM A MEDIA BIAS CHECKER Frametra is not a fact-checker and does not issue partisan verdicts. It will not tell you a story is left-leaning, right-leaning, true, or false. It tells you how the story is constructed, and gives you the structural evidence to judge for yourself. That distinction matters: a story can be structurally sound and still wrong on the facts, and a thinly sourced story can still turn out to be accurate. Frametra measures construction, not truth. Every finding comes with the exact phrases from the article that drove it, highlighted in the panel, so you can check the reasoning instead of taking a score on faith. THE METHODOLOGY: JERI SOURCE-DIVERSITY SCORING Frametra's source-diversity scoring is grounded in the JeRI framework, developed by Adamson and Malik at Toronto Metropolitan University. JeRI evaluates who gets to speak in a story, whose perspective is left out, and how claims are supported or left to stand alone. It is the same rigor an editor or fact-checker would apply by hand, made repeatable and consistent across every article you read. The same six-signal weighting applies on the free engine and on every paid version. GENRE-AWARE SCORING Before scoring, Frametra classifies the article's genre: News Report, Opinion or Op-Ed, Analysis or Explainer, Feature, Press Release, or Other. It then scores the piece against the structural norms of that genre. Assertive, first-person argument is expected in an Op-Ed. The identical language inside a News Report that presents itself as neutral reporting is flagged as a genuine structural integrity issue. This is the difference between Frametra and a generic bias detector: it does not hold an opinion column and a wire report to the same genre-blind standard, because that is not how journalism works. SIX FORENSIC SIGNALS ON EVERY ARTICLE Source Attribution. Named sources, direct quotes, and hyperlinked citations per claim. Detects when a story leans on unnamed or unverifiable sourcing, and credits sources properly when a claim links directly to its origin. Editorial Integrity. Loaded language and passive-voice framing versus neutral, direct reportage. This is the signal most people mean when they say "bias," measured structurally instead of politically. Claim Pressure. How much of the text pushes a conclusion versus letting evidence speak for itself. Flags unverified claim density relative to article length. Epistemic Hedging. Whether uncertainty is acknowledged and calibrated, or asserted with false confidence. Appropriate hedging on unconfirmed details is a sign of careful reporting, not weakness. Evidence Signals. Numeric references, data points, and documentary grounding behind the claims being made. Passive Voice Ratio. Whether responsibility for actions and claims is stated clearly or obscured through passive construction. HOW IT WORKS 1. Install the extension. It sits quietly in your browser until you need it. Available for Chrome, Edge, and Firefox. 2. Open any news article and click Analyze. The free heuristic engine runs entirely on your device, so no article text leaves your browser at the free tier. 3. Get an instant integrity score and a plain-language breakdown by signal, with the evidence phrases highlighted in each finding. 4. On the paid versions, run a Deep AI Analysis for richer findings, source-level detail, and exportable documents. Your heuristic score and your AI-verified score appear side by side, so you can see where the two methods agree and where they do not, instead of one number silently replacing the other. WHAT IS IN THE PANEL Deep AI Analysis. An AI-assisted pass on all six signals with specific, quoted findings, an article genre classification, and a structural certificate verdict. Draft Scanner. Check your own writing before you publish. It reads drafts directly inside WordPress, Substack, Ghost, Medium, and Notion editors and flags thin attribution and claim-heavy passages before an editor does. Paste a Draft lets you run the full analysis on any text with no live page, including Google Docs content, which browsers do not let extensions read directly. Emotive Tone Map. Loaded and emotionally charged language flagged down to the exact phrase. Framing Balance. A plain-language read on which perspectives and stakeholders are present in a story's framing and which are absent. It is descriptive, never a left-right rating. Source Library. Every named source you encounter is tracked across every article you analyze and organized by JeRI source category. The library flags two descriptive patterns: a source that appears only in one outlet's coverage across three or more tracked articles, and a source classified differently from one article to the next. Sources can be renamed or merged by hand when two mentions of the same person do not match exactly. Frametra never assigns a source a trust score. Transparency Signals. Measured checks of what the page itself shows: author named, publication and update dates, corrections link, editorial standards link, ownership page, and dateline. They say nothing about accuracy or reputation. Narrative Timeline. A structured view of how a story's framing shifts within its own coverage, with flags where sources disagree or a claim goes uncontested. Compare Coverage. Add two to four articles about the same story and see how each is built, side by side, with a plain-language note on the widest structural gap. It runs on your device and is free. Batch Analysis. On paid versions, score up to 25 articles at once and download one spreadsheet. It runs on your device. Evidence Capture. On paid versions, preserve a page as a printable record with the full text, the capture time, SHA-256 fingerprints, and an independent time attestation signed by Frametra's server, which receives only a fingerprint, never the page. How This Score Is Calculated. Open any score to see each signal, its weight, and the points it contributed, plus what the tool cannot establish. Private Mode. One switch for confidential or unpublished material: nothing from the analysis is saved on your device. It does not change what Deep AI Analysis sends for review, which is described under Privacy below. Annotated Article export. A downloadable PDF of the article's own text with the exact evidence phrases behind each finding highlighted at their real location, color-coded by signal, with a legend. You see the evidence in context, not only in a list. Structural Certificate and Full Report. A one-page summary verdict, and a complete report of every finding across all six signals, both as PDF documents with a document ID and date. The Full Report includes a "What Frametra read" box showing how much of the article was analyzed. Translate to English. Translate any foreign-language article into a clean English PDF before you analyze or cite it. If the article is already in English, Frametra tells you instead of generating anything. FIVE VERSIONS, ONE ENGINE Every paid version runs the same six-signal Deep AI engine at the same price. What changes is the language of the tool. Each version names the features the way your field already talks, and the documents are labeled for your use. DEEP AI FOR JOURNALISTS (formerly Deep Analysis Pro). Run a structural check on a source before you quote it, or on your own draft before you file. Features keep their newsroom names: Scan My Draft, Narrative Timeline, Emotive Tone Map, Source Library, Transparency Signals, Structural Certificate, and Annotated Article. DEEP AI FOR EDUCATION. Built for journalism programs, media studies, and media literacy classrooms. Students use Check My Assignment to self-edit, Compare Sources to see how two outlets cover the same event, Language and Tone Map, Perspective Balance, My Sources, and Annotated Reading. Certificates and reports can be labeled with a student name or class code for classroom submission, with no accounts or roster setup. DEEP AI FOR LEGAL (formerly Counsel Pro). For lawyers, paralegals, and litigation support documenting how media coverage was constructed in a case, a defamation review, or a pretrial publicity question. Features are named for the work: Review a Draft Statement, Publicity Chronology, Charged Language Map, Source Register, Publisher Transparency, Structural Review Certificate, Evidence Capture, and Annotated Coverage. Legal Review Documents carry a matter reference field, a preparer field, and a built-in disclaimer stating plainly that this is structural analysis to support counsel's own judgment, not legal advice or expert testimony. DEEP AI FOR PODCASTERS AND RESEARCHERS. For podcast hosts, documentary producers, OSINT analysts, academics, and fact-checkers who work across many sources. Script Check, Paste a Script or Notes, Compare Reporting, Batch Analysis, Research Library, Source Transparency, Integrity Certificate, and Annotated Source. The Research Library builds a persistent picture of which sources and outlets recur across everything you research, and where real source diversity is present or missing. CONTRACTOR. For independent professionals who deliver structural reviews to paying clients. Every analysis becomes a client-ready PDF report under your own name: a Client Report, a Client Certificate, and an Annotated Client Article, each with your business name on the cover, a Client and Engagement line, and a document ID. Choose a standard or legal review layout, and keep a Client Report Log (CSV export). Plans include 50, 100, or 250 client analyses a month; making the documents costs no extra. To apply, email sales@frametra.com. WHO USES FRAMETRA Journalists and Writers. Check a source's structure before you quote it, and audit your own draft before you file it. Journalism Educators and Students. Give students a hands-on, citable framework instead of an abstract lecture on media bias. Researchers, Fact-Checkers, and OSINT Analysts. Apply one repeatable methodology across many articles instead of a manual read of each one. Legal Professionals. Document the structural pattern of media coverage relevant to a case, with documents formatted to support counsel's own review. PR and Crisis Communications Professionals. Screen a release or a client's coverage before it goes further. CHOOSE YOUR PLAN FREE. No account required. The complete six-signal heuristic engine, unlimited use, no sign-up, no credit card. It is the same structural engine, running entirely on your device, with paid findings visible as a preview so you know what an upgrade would unlock. DEEP AI FOR JOURNALISTS, LEGAL, AND PODCASTERS AND RESEARCHERS. $9 per month or $79 per year for any one version, with 40 Deep AI analyses a month. Scoring, Compare Coverage, Batch Analysis, Evidence Capture and exports run on your device and never count toward the 40. AI-assisted analysis on every signal, the Draft Scanner, emotive tone mapping, the Framing Balance indicator, the persistent source library, the timeline view, transparency signals, the Annotated Article export, English translation, and the Structural Certificate and Full Report PDFs. Education is priced to your class size and needs, with customizable class outlines for a turn-key school program: sales@frametra.com. CONTRACTOR. Monthly plans with a defined analysis allowance, for independent professionals who deliver client reports. See the Contractor section above and apply at sales@frametra.com. PR AND CRISIS COMMUNICATIONS. Run it yourself with any Deep AI version, or have us run it: a $99 Pre-Publication Review per article, Coverage Analysis Reports from $300. Details at frametra.com. PRIVACY Free-tier analysis runs entirely in your browser. No article content or browsing history leaves your device at this tier, and nothing is logged or sold. On a paid version, the text of the article you choose to analyze is sent for AI-assisted review. It travels through Frametra's secure server to Anthropic's API, or directly to Anthropic if you supply your own key. Frametra's server passes the text along to get your result and does not store it. Anthropic processes the text under its own API data terms. Your analysis history, source library, and client log stay in your browser's local storage on your own computer, and Private Mode keeps an analysis out of them. Evidence Capture sends Frametra's server only a fingerprint of the record, never the page or its text. Translating an article works the same way as Deep AI Analysis. Article text is never used to train models and is never shared with advertisers. A paid version also checks your access key with Frametra's server so your plan and monthly allowance stay accurate. Frametra needs no account, and the extension does not collect your name, browsing history, or the pages you visit. RESEARCH AND RESOURCES Frametra maintains a research catalog of source-diversity audits, methodology guides, and curriculum packs, including JeRI Framework Applied: Canadian Media and How Disinformation Is Built. Available at frametra.com. FREQUENTLY ASKED QUESTIONS Is Frametra a fact-checker or a media bias checker? Neither, in the usual sense. A fact-checker rules on whether claims are true. A bias checker tells you which side a story favors. Frametra measures how a story is built: attribution, evidence, hedging, and framing. It leaves the judgment about accuracy and fairness to you. Is Frametra really free? Yes. The free tier is the full six-signal heuristic engine, unlimited use, with no account and no credit card. Paid versions add AI-assisted deep analysis and the features described above. Does a failed analysis use up my monthly allowance? No. Only analyses that complete count toward your 40. If the AI service can't be reached, Frametra retries once automatically, and a failed attempt is not counted. Can students use Frametra for assignments? Yes. Check My Assignment lets a student review their own writing before submitting, and certificates and reports can carry a student name or class code. Programs can ask about a group code at sales@frametra.com. Can I cancel anytime? Yes. Cancel from the Stripe billing link in your welcome email at any time, with no cancellation fees. Your access continues until the end of your current billing period. Payments are processed securely through Stripe. We offer a 7-day full refund on first-time subscriptions. Does it work in other languages? The free engine scores English text only. Deep AI reads other languages, and Translate to English converts a foreign-language article for review. Scoring has been tested most on English. ABOUT FRAMETRA Frametra Software Inc. is an independent Canadian company based in London, Ontario, applying the JeRI structural journalism framework from Toronto Metropolitan University in an everyday browser tool. We are small and still building. If a score looks wrong or something is confusing, we want to hear about it directly: sales@frametra.com.