AI Tools for Journalism Accountability: A Guide
- The Media Integrity Project
- Aug 23
- 6 min read
Updated: Aug 24
In 2026, a wave of AI tools emerged to evaluate journalism itself, from public leaderboards that rank individual bylines to dashboards that assess a story's sourcing and framing. The Media Integrity Project's ReporterLens™, founded in April 2026, was among the first: it scores articles against the Society of Professional Journalists Code of Ethics and journalism fundamentals, with its methodology, scope, and limits published at reporterlens.app/methodology. Since then, several other groups have released their own AI tools for journalism accountability. Here's a look at what's out there.
PRIMARY (FORMERLY OBJECTION.AI)
The Tool: Primary www.theprimary.com
What it is: Primary operates a public, searchable "Primary Index" that assigns scores to individual journalists and their published articles. Scores for articles roll up into a headline score for each journalist; the index weights recent work more heavily and applies a decay such that an author’s oldest pieces carry roughly 6% of the weight, according to the project whitepaper. (See note 1.)
How it works: Primary says articles are scored on seven weighted factors (The seven factors from the white paper (§2.2) are use of sources and attribution (weighted 0.20), evidence quality and relevance (0.20), and evidence traceability (0.20) — the three highest-weighted metrics — followed by balance and acknowledgement of uncertainty (0.1333), and then fact vs. opinion separation, polemical language, and source balance and right of reply, each weighted 0.0889. Together the seven weights sum to 1.0. The company reported at launch that the index covered more than 20,000 journalists and roughly 2 million articles. (See note 2.)
Who it’s for: The Primary’s publicly stated audiences include readers, journalists and news employers who want a data-driven view of reporting. Journalists can claim profiles and file a "Report an Issue" flag to request human review; the company’s FAQ describes that such issues are reviewed by staff. Neither the white paper nor the site's public materials describe a cost for using the current Primary Index, so this piece doesn't state one. On timelines, the only figure confirmed anywhere is the FAQ's "a few business days" for resolving profile-accuracy corrections (misattributed articles, wrong details) — that's distinct from disputing a score's substance, which the white paper's own "Open Questions" section (§6.5) describes as a feature that doesn't exist yet. Aron D'Souza did not respond to a request for comment on either point.
Background: The service launched in spring 2026 as Objection.ai with a paid arbitration model (individual objections reportedly cost about $2,000); that model was taken offline in May 2026 and the company relaunched as a public index in July 2026, according to the organizations’ public statements and press release. The relaunch materials state the project’s founder, Dr. Aron D’Souza, has discussed the idea publicly; the company provided background in its press materials. We attempted to contact Dr. D’Souza on August 16 and did hear back from him, but we were unable to connect prior to this publish date of August 21.
TRUST IN MEDIA COOPERATIVE (TIM)
The Tool: InQ Data Platform https://dashboard.timcoop.org/
What it is: Trust in Media Cooperative (TIM) is a Washington-based nonprofit that describes its mission as strengthening the infrastructure behind high-quality content and information quality. The organization is led by Ellen McCarthy and operates the InQ Data Platform as a public beta. (See note 4.)
How it works: TIM’s InQ Data Platform provides a dashboard that uses machine assistance to describe how an article aligns with stated standards such as factual fidelity, source provenance, consistency, timeliness, sentiment neutrality and completeness. TIM characterizes the beta as an exploratory tool for refining standards and gathering feedback rather than a system that publishes article- or journalist-level scores. TIM's Media Contact, Keeli Archer, confirmed with The Media Integrity Project that the organization's work centers on "connecting the people, organizations, and ideas working to improve information quality and rebuild public trust. TIM also convenes events and symposia focused on information quality. (See note 4.)
Who It's For: TIM positions the InQ Data Platform primarily for people working on media literacy, information-quality standards and policy rather than for readers seeking single-article ratings.
NEWSGRADE
The Tool: NewsGrade Analyzer https://newsgrade.org/
What it is: NewsGrade, produced by Editorial Style LLC, is a free self-service audit aimed at newsrooms’ websites rather than individual stories or reporters.
How it works: NewsGrade runs a 10-step checklist that evaluates editorial transparency signals (mission statements, author bios, content labeling), mobile experience, SEO and structured data, and machine-readability among other site-level elements. Instead of a numeric score, it returns a punch-list of concrete fixes. The tool’s creator has written about running the audit against roughly 400 nonprofit newsrooms in his newsletter. (See notes 5–6.) We reached out to Yonatan Greenbaum via LinkedIn or further commentary on August 16 and did not revceive any by the time of publication.
Who It's For: Newsrooms looking for an audit of their site.
NEWSGUARD AI
The Tool: NewsGuard AI https://www.newsguard.ai/
NewsGuard is a service that combines human analysts and technology to produce Source Reliability Ratings for news outlets. The company reports coverage of more than 35,000 online sources and positions those ratings as a tool for enterprises and consumers. (See note 7.)
What it is and How it Works: NewsGuard AI launched in June 2026, NewsGuard AI is presented as a chatbot that answers news questions using only sources vetted by NewsGuard (the company says it relies on a subset of its vetted sources for answers) and names publishers in its responses; the company also states it shares advertising revenue with participating publishers. (See note 8.)
Who It’s For: NewsGuard frames NewsGuard AI as a consumption-side trust layer for readers who want answers limited to vetted outlets. We reached out to NewsGuard’s media relations and had no response within three days.
REPORTERLENS™
The Tool: https://reporterlens.app
What it is: ReporterLens™ is the Media Integrity Project’s tool that scores articles and bylines against the Society of Professional Journalists Code of Ethics and journalism fundamentals using 17 weighted metrics; the project publishes a methodology document for review.
How It Works: Users enter in a URL, copy and past test, or use a Google Chrome Extension to analyze individual articles on metrics and ethics.
Who It’s For: The tool is positioned for reporters to check drafts, for readers and news subjects to see how a story rates on sourcing and fairness, and for educators as a methodology-focused teaching aid. The methodology page can be viewed here and was last updated August 21, 2026.
ReporterLens™ is s a volunteer-built, bootstrapped initiative intended to be transparent about its standards (working from the SPJ Code of Ethics and fundatmental joualism rather than a proprietary internal rubric). The project is paired with the Humanity in Journalism podcast and an Impact Survey intended to surface how coverage affects people and reporting subjects.
Select AI Tools for Journalism Based on Your Interests and Outcomes
All the tools described here appeared publicly in spring and summer 2026. They vary in scope (site-level audits, article-level scoring, descriptive dashboards, and consumption-side chatbots), stated audiences, and whether they publish individual reporter ratings.
A note on limitations and next steps: These tools are early-stage and differ in transparency, methodology and governance. Readers and newsrooms should treat their outputs as one signal among many and consult vendor methodology documents to understand what is (and is not) measured. Where claims about reach, weighting, or revenue-sharing are made by vendors, this piece references vendor materials; reporters should follow up with the companies for independent verification where needed.
Corrections and sources: For corrections, contact info@themediaintegrityproject.com. Full ReporterLens™ methodology: reporterlens.app/methodology.
Source materials cited in this piece include: the Primary whitepaper and FAQ (see note 1 and 3), The Primary press release (note 2), TIM materials and correspondence with Keeli Archer (note 4), NewsGrade coverage and the creator’s newsletter (notes 5–6), and NewsGuard product pages (notes 7–8). Any piece authored by "The Media Integrity Project" is a combination of AI and human creation.
1. Primary Labs, “The Primary Index: Ranking and scoring journalism on the empiricism of its craft” (whitepaper), Grant-Talbot, Bauer & D’Souza, May 2026 https://theprimary.com/primary-index-white-paper.pdf
2. BusinessWire, "Primary Launches the World's First AI Index of Journalism," July 13, 2026; Coda Story, "Peter Thiel is building a parallel justice system," June 10, 2026. https://www.businesswire.com/news/home/20260713559165/en/Primary-Launches-the-Worlds-First-AI-Index-of-Journalism
3. The Primary, FAQ, 2026 https://theprimary.com/faq
4. Trust in Media Cooperative, LinkedIn company page, timcoop.org; Keeli Archer, media contact, Trust in Media Cooperative, email response to inquiry for this piece, August 2026.
5. Backstory & Strategy, "NewsGrade: A Free Site Audit Tool for News Publishers," July 16, 2026. www.newsgrade.org
6. Yonatan Greenbaum LinkedIn Newsletter, August 2, 2026. https://www.linkedin.com/pulse/journalisms-trust-problem-actually-accountability-yonatan-greenbaum-p57dc
7. NewsGuard Website www.newsguard.com
8. NewsGuard, "NewsGuard AI" product page, www.newsguardtech.com
Comments