The nine criteria, and what each is worth
NewsGuard assesses news and information websites against nine criteria and publishes the weight attached to each. The total is 100 points, and the distribution is public, which makes it one of the few credibility systems whose scoring can be examined rather than inferred.
The five credibility criteria carry the bulk of the score. Not publishing false or egregiously misleading content at a significant level is worth 22 points. Gathering and presenting information responsibly is worth 18. Having effective practices for correcting errors is 12.5, as is handling the difference between news and opinion responsibly. Avoiding deceptive headlines is worth 10.
The four transparency criteria carry the remaining 25 points. Disclosing ownership and financing is 7.5. Clearly labelling advertising is 7.5. Revealing who is in charge, including possible conflicts of interest, is 5. Providing the names of content creators along with contact or biographical information is another 5.
Points out of 100 assigned to each pass-fail criterion, as published by NewsGuard
Source: NewsGuard, Website Rating Process and Criteria, rechecked October 10, 2026.
The first five are credibility criteria (75 points), the last four transparency criteria (25 points). NewsGuard says each criterion is pass-fail: a site gets all of its points or none.
The weighting is doing something specific and worth reading closely. Accuracy is the largest single line, but it is nowhere near a majority on its own: the top criterion is 22 of 100, and four of the nine criteria have nothing whatever to do with whether any individual claim on the site is true or false.
That design reflects a view about how credibility is assessed at scale. You cannot fact-check every article on every site, but you can check in an afternoon whether a masthead exists, whether advertising is labelled, and whether corrections are made.
Why corrections are worth more than a byline
Effective practices for correcting errors carries 12.5 points — more than twice the weight of naming authors, and a quarter of the weight of accuracy itself.
That is a deliberate statement about what distinguishes a news organisation from a publisher of content. Every outlet gets things wrong. What separates them is whether being wrong produces a visible, dated correction or silent editing.
It also aligns with what platforms and search engines say they look for, which we examined in our report on how publications handle corrections. A corrections policy is cheap to write, hard to fake over time, and legible to an outside assessor.
The same logic explains why news-versus-opinion handling also carries 12.5. A site that runs advocacy as reporting is not necessarily publishing falsehoods, but it is misrepresenting what kind of claim the reader is being offered.
For a company buying coverage, these two criteria are the most directly relevant in the whole framework, because they are the ones a placement can affect. Coverage that appears without clear labelling on a site with no corrections practice is being graded against both.
The transparency quarter
The four transparency criteria are the cheapest points on the board and the ones most often left on the table. None requires editorial investment; all four require disclosure.
Ownership and financing at 7.5 points asks who owns the site and how it is funded. Advertising labelling at 7.5 asks whether commercial content is distinguishable from editorial. Leadership disclosure at 5 asks who is responsible and whether conflicts exist. Creator identification at 5 asks for names and either contact details or biographies.
Together that is a quarter of the total score available for publishing facts about yourself that you already know and that cost nothing to disclose. No reporting is required, no sources are needed, and the information does not change from week to week.
The advertising criterion is the one that intersects directly with paid placement, and it is not about whether sponsored content exists. It is about whether a reader can tell. We have set out what regulators require of that labelling in a separate report, and the standards are broadly aligned.
There is an uncomfortable implication for this industry. A site that accepts paid articles and labels them properly loses nothing on this criterion; a site that accepts them and disguises them loses 7.5 points before anyone assesses a word of the copy.
Who does the rating, and can you answer back
NewsGuard states that a site is assessed by an analyst against the nine criteria, and describes its analysts as trained journalists rather than automated systems.
It also commits to a right of reply before publication. Where an analyst believes a site may fail one or more criteria, NewsGuard says it is its practice to contact the site's proprietor to seek comment before the rating is published.
That is a meaningful procedural safeguard and an unusual one in this field. Most systems that grade publishers — platform quality signals, ad-tech inclusion and exclusion lists, search ranking systems — provide no notice that an assessment is happening, no explanation of the outcome and no route to contest it.
It also means a rating is partly a test of responsiveness. A site that ignores the approach is assessed on whatever the public record shows, which for a thinly documented publication is frequently less than the proprietor would want.
The output is not only a score. Each rated site receives a written report describing ownership, financing, content, credibility practices, transparency practices and history, which is a considerably more useful artefact than a number.
The score bands, and the line at 60
The weights are not the only thing NewsGuard publishes. The same rating process and criteria page, last updated on 14 April 2026, also sets out what each score means, which settles a question secondary sources have long disagreed on.
A score of 100 is labelled High Credibility. 75 to 99 is Generally Credible, 60 to 74 is Credible with Exceptions, 40 to 59 is Proceed with Caution and 0 to 39 is Proceed with Maximum Caution. NewsGuard describes sites in both of the bottom bands as unreliable, so 60 is the line between a site it calls credible, with or without exceptions, and one it calls unreliable.
Lowest score in each band, out of 100, with NewsGuard's label; a classification, not a measurement
Source: NewsGuard, Website Rating Process and Criteria (see sources), page last modified April 14, 2026; rechecked October 10, 2026.
NewsGuard describes sites in the bottom two bands, below 60, as unreliable.
Secondary accounts have put the key line at either 60 or 75. The published bands show why both figures circulate: 75 is where Generally Credible begins, but 60 is where unreliable ends.
The rated-universe figures commonly cited — more than 6,000 sites accounting for 95 per cent of online engagement across five countries — trace to a company statement from January 2021 and are now several years old.
For a publisher the practical consequence is concrete. The weights and the bands are both published, so the distance between a site's score and 60 can be worked out from the criteria it fails, although only NewsGuard's analysts decide which criteria those are.
What a publisher can actually act on
The framework rewards process over prose, which is unusual and useful to anyone deciding where to spend effort. Four of the nine criteria can be satisfied inside a week by publishing pages that describe the organisation, without changing a single line of reporting or hiring anyone.
An ownership and financing page. A masthead naming who is in charge, with conflicts disclosed. Bylines with contact details or biographies. Advertising and sponsored content labelled so a reader can tell at a glance. That is 25 points of a 100-point score.
The corrections criterion is the next cheapest at 12.5, and it requires a published policy plus the discipline to use it. A corrections page with no corrections on it after a year is not evidence of accuracy.
None of this is gaming a system. Every one of those criteria is a thing readers have a legitimate interest in knowing, which is why an outside assessor uses them as a proxy in the first place.
And it explains why credibility assessment converges on the same handful of signals across very different institutions, from journalism trust indicators to search quality guidance: the things that can be verified from outside are ownership, authorship and labelling. Whether errors actually get corrected is NewsGuard's own addition, and at 12.5 points it is the heaviest criterion that is not about accuracy itself.
Methodology and limitations
The nine criteria and their point weights are quoted from NewsGuard's published rating process and criteria page, retrieved 26 September 2026, as are the statements that ratings are performed by analysts described as trained journalists and that NewsGuard's practice is to seek comment from a site's proprietor before publishing a rating where a criterion may be failed. The named score bands, and NewsGuard's description of sites scoring below 60 as unreliable, are quoted from the same page as rechecked on 10 October 2026; the page was last modified on 14 April 2026. The count of rated sites is taken from secondary summaries, and the rated-site and engagement figures trace to a company statement dated January 2021. We have not been rated by NewsGuard, have no commercial relationship with it, and have not independently verified any individual site rating.
Conclusion
NewsGuard's framework is unusually legible: nine criteria, published weights and score bands, a written report, and a right of reply before publication. Read against the weights, the lesson for publishers is that a quarter of the available score is pure disclosure — ownership, financing, leadership, bylines and advertising labelling — none of which requires changing a word of editorial. The largest single line remains accuracy at 22 points, but accuracy alone cannot carry a site: a publication that reports carefully and tells readers nothing about who owns it, who wrote it or which pages were paid for is leaving 25 points on the table and inviting exactly the scepticism the framework is designed to measure.
Frequently Asked Questions
How many points is accuracy worth?
Not publishing false or egregiously misleading content at a significant level carries 22 points, the heaviest single criterion. Gathering and presenting information responsibly adds 18. Together they are 40 of 100.
What score does NewsGuard treat as credible?
NewsGuard's criteria page publishes the bands: 100 is High Credibility, 75 to 99 Generally Credible and 60 to 74 Credible with Exceptions. Sites scoring 40 to 59 (Proceed with Caution) or 0 to 39 (Proceed with Maximum Caution) are described as unreliable, so 60 is the line to stay above.
Do sponsored articles hurt a rating?
Running them is not itself a failure. Clearly labelling advertising is worth 7.5 points, so the criterion turns on whether a reader can distinguish commercial content from editorial, not on whether commercial content exists.
Can a site respond before being rated?
NewsGuard states that where an analyst believes a site may fail one or more criteria, its practice is to contact the site's proprietor to seek comment before publishing the rating.
Is the rating automated?
NewsGuard describes ratings as produced by analysts it characterises as trained journalists, assessing a site against the nine criteria, rather than by an automated scoring system.
Which criteria are cheapest to satisfy?
The four transparency criteria, worth 25 points in total, ask only for disclosure: ownership and financing, advertising labelling, who is in charge including conflicts, and named content creators with contact or biographical detail.
Sources
- vendor research NewsGuard, Website Rating Process and Criteria, retrieved 26 September 2026 and rechecked 10 October 2026 (page last modified 14 April 2026). Source of the nine criteria, their point weights, the named score bands and the description of sites below 60 as unreliable, the description of analysts and the right-of-reply practice.
- trade press Secondary summaries of NewsGuard's rated universe, consulted 26 September 2026. Some also give the key score line, as either 60 or 75; NewsGuard's own published bands, above, are used instead. The rated-site figures trace to a company statement of January 2021.
- trade body The Trust Project, 8 Trust Indicators, retrieved 26 September 2026. Consulted for the overlap between externally verifiable credibility signals: Best Practices asks who funds a site, Journalist Expertise covers author credentials, and Labels covers the news-versus-opinion distinction.