How Newsrooms Can Verify AI-Generated News With Confidence Scoring
AI can help a newsroom research, summarize, classify, draft, translate, and repurpose news content. But one question remains difficult: How Newsrooms Can Verify AI-Generated news before it is published?
A practical answer is to create a newsroom confidence score that summarizes the strength of the evidence behind a story. The score should not mean that an article is “90% true” or that an AI model is 90% accurate. Instead, it should show editors how much confidence the newsroom has in the sources, verification, attribution, completeness, currency, and human review behind the story.
The strongest approach is to treat the score as an editorial decision-support tool, not an automated publication rule.
That distinction matters because current newsroom guidance continues to place verification and editorial accountability with journalists. The Associated Press's July 2026 AI standards, for example, permit AI assistance for tasks including early research, document summarization, transcription, translation, headlines, summaries, grammar, and search optimization, while stating that AI does not replace reporting, sourcing, editorial judgment, or verification. AP also says AI-generated output is reviewed and edited by journalists before publication.
For digital publishers building a human-governed AI newsroom, confidence scoring can provide a practical bridge between automated assistance and human editorial control.
What Is A Confidence Score For An AI-Assisted Story?
A newsroom confidence score is an internal assessment of how strongly the available evidence supports the important claims in an AI-assisted story.
It can consider factors such as:
Source quality
Source independence
Primary-source availability
Fact verification
Attribution
Recency
Evidence completeness
Conflicting information
AI intervention
Human editorial review
Story risk
A newsroom might use a 0–100 scale, but the number itself is not the important part.
For example, a publisher could internally classify stories as:
Score Range | Suggested Status | Editorial Meaning |
90–100 | High confidence | Strong evidence and completed review |
75–89 | Good confidence | Most important claims supported; minor gaps may remain |
60–74 | Review required | Evidence is incomplete or some claims need additional work |
40–59 | Low confidence | Significant uncertainty or unresolved conflicts |
Below 40 | Hold | Insufficient evidence for normal publication |
These ranges are a proposed newsroom framework, not an industry-standard measurement system.
The exact thresholds should be determined by each publisher.
A high-risk investigation might require a much higher internal threshold than a low-risk entertainment update.
Why Newsrooms Need Confidence Scoring
AI-assisted journalism creates a new editorial problem.
Traditional journalism already requires editors to evaluate sources and evidence. AI adds another layer because the system may produce fluent text that appears authoritative even when some underlying claims are incomplete, outdated, misunderstood, or unsupported.
The Reuters Institute's 2026 Journalism, Media, and Technology Trends report found that AI use by news organizations continues to expand. Its survey of 280 digital leaders across 51 countries found that 97% considered back-end automation important and 82% considered newsgathering applications important. Yet only 44% described their newsroom AI initiatives as promising, while 42% described results as limited.
This suggests that publishers need more than AI adoption.
They need AI quality controls.
A confidence score can help answer:
What do we actually know?
How strong is the evidence?
What remains uncertain?
Has a human checked the important claims?
Is this story ready for publication?
A Confidence Score Should Measure The Story, Not The AI Model
One of the biggest mistakes would be to assign a confidence score based primarily on which AI model produced the article.
A powerful model does not automatically produce a high-confidence story.
A smaller model working from excellent primary sources and carefully reviewed evidence may produce a safer result than a more capable model working from weak or conflicting information.
The score should therefore evaluate the editorial evidence chain.
Think of it this way:
Model quality ≠ story confidence
Instead:
Story confidence = strength of evidence + verification + editorial review + appropriate context
The model is only one component of the workflow.
The Six Core Factors Behind A Newsroom Confidence Score
A practical system can begin with six factors.
1. Source Quality
Ask:
How strong are the sources supporting the story?
A primary government document, court filing, regulatory record, direct interview, original dataset, or firsthand evidence may provide stronger support for a particular claim than an anonymous social-media post.
But source quality is claim-specific.
An official organization may be the best source for what it announced.
It may not be the best source for whether its own claims are independently accurate.
The scoring system should therefore evaluate both authority and relevance.
2. Source Independence
A newsroom should determine whether supposedly separate sources are actually independent.
Ten websites repeating one press release do not create ten independent confirmations.
Similarly, multiple social posts may simply originate from one account.
A confidence system should therefore ask:
How many genuinely independent evidence paths support the claim?
This is particularly important for breaking news.
3. Verification Coverage
The newsroom should identify the major factual claims in the story and determine how many have been verified.
For example, an article might contain 20 important factual claims.
If 18 are verified but two remain uncertain, the story should not simply receive a generic “verified” label.
The system should record the remaining uncertainty.
This can be especially useful when AI generates long articles containing many individual claims.
4. Evidence Completeness
A story may contain accurate individual facts but still provide an incomplete picture.
For example, a company announcement might accurately describe a new policy but omit criticism or relevant context.
Confidence scoring should therefore consider whether the article contains enough evidence and context to support the central conclusion.
This is different from checking whether every sentence has a source.
5. Recency
Some facts remain stable for years.
Others can change within minutes.
A confidence score should account for the time sensitivity of the subject.
For example:
Historical fact: relatively stable
Company executive: can change
Election result: changes during counting
Market price: highly dynamic
Emergency situation: changes rapidly
Government policy: may change after publication
A confidence score should therefore record when important information was last verified.
6. Human Review
Human review should be a major part of the system.
An AI-assisted article that has been thoroughly checked by an experienced editor should not receive the same internal status as AI-generated copy that has received only a superficial review.
The score should record:
Who reviewed the article
What was reviewed
Which claims were checked
Whether sources were inspected
Whether significant edits were made
Whether unresolved issues remain
This aligns with NIST's AI Risk Management Framework, which identifies human oversight as part of responsible AI risk management and recommends clearly defining oversight roles and responsibilities.
A Practical Confidence-Scoring Framework
A publisher could create an internal weighted framework.
For example, an illustrative model might assign:
Factor | Illustrative Weight |
Source quality | 20% |
Independent corroboration | 15% |
Claim verification | 25% |
Evidence completeness | 15% |
Recency | 10% |
Human editorial review | 15% |
These weights are not a scientifically validated formula. They are a starting point that publishers should test against their own workflows.
The most important factor may vary by story type.
For an investigative article, source verification and corroboration may deserve greater weight.
For a breaking-news article, recency and update status may matter more.
For a data-driven article, evidence quality and numerical verification may dominate.
The purpose of weighting is not to create false precision.
It is to force the newsroom to examine the components of confidence systematically.
Do Not Treat 82/100 As “82% True”
This is critical.
A confidence score should not be presented to readers as a probability that the article is true.
A score of 82 does not mean:
“There is an 82% chance this story is accurate.”
Unless a publisher has conducted a rigorous statistical validation process, such a statement would be unjustified.
Instead, the internal meaning should be closer to:
“This story has passed the newsroom's defined confidence checks and is currently classified as high confidence.”
The number is a workflow indicator.
It is not a scientific probability.
Confidence Should Be Calculated At The Claim Level
One of the strongest improvements a newsroom can make is to avoid scoring only the whole article.
A story can contain:
Very high-confidence facts
Moderately supported claims
Unverified allegations
Developing information
Expert interpretation
Predictions
A single article score can hide these differences.
A better system can assign confidence to individual claims.
For example:
Claim | Evidence | Status |
Government published the regulation | Official document | High |
Regulation takes effect on a stated date | Official document | High |
Industry group criticized the policy | Direct statement | High |
Policy will reduce costs by a specific amount | Industry estimate | Medium |
Policy will create a certain number of jobs | Forecast | Low / attributed |
Long-term economic impact | Not yet established | Uncertain |
This creates a much richer editorial picture.
The article-level score can then summarize the state of the underlying claims.
Fact Packs Make Confidence Scoring Easier
This is where a Fact Pack can become useful.
A Fact Pack is a structured evidence layer containing verified facts, source references, quotations, dates, figures, context, unresolved questions and verification status.
If each important claim is already connected to its evidence, the newsroom can calculate confidence more systematically.
The workflow becomes:
Source Collection → Fact Pack → Claim Verification → Confidence Assessment → AI Draft → Human Review → Publication
Instead of:
Prompt → AI Draft → Editor Reads
The first workflow gives editors much greater visibility into why a story received its confidence status.
Confidence Scoring Should Begin Before AI Drafting
A common mistake is to score the article only after the AI has generated it.
By then, unsupported assumptions may already be embedded in the draft.
A better process begins during research.
The newsroom can score:
Source Confidence
How reliable is the source for this specific claim?
Claim Confidence
How strong is the evidence supporting the claim?
Story Confidence
How strong is the evidence supporting the overall story?
Publication Confidence
Has the story passed the editorial review required for its risk level?
This creates multiple checkpoints.
The AI can then be given only the evidence that has reached an acceptable status.
Confidence Scores Should Affect Editorial Workflow
The score becomes valuable when it changes what happens next.
A publisher could define actions such as:
High confidence: Eligible for normal editorial review.
Medium confidence: Additional source verification required.
Low confidence: Reporter or editor investigation required.
Unverified: Do not present as established fact.
Conflicting: Escalate to senior editorial review.
High-risk + medium confidence: Hold publication until enhanced verification is complete.
This turns the score from a decorative number into an operational control.
Risk Should Be Separate From Confidence
This is another critical distinction.
A story can have high confidence but high risk.
For example, a newsroom may have strong evidence for a serious allegation.
That does not mean the story should automatically be published.
The newsroom may need:
Legal review
Right of reply
Additional corroboration
Stronger attribution
Privacy assessment
Harm assessment
Senior editorial approval
Conversely, a low-risk story may be acceptable with a lower confidence threshold.
Therefore, publishers should use two dimensions:
Confidence: How strong is the evidence?
Risk: What could happen if we are wrong?
This produces a more useful editorial matrix.
Confidence | Risk | Recommended Action |
High | Low | Normal review |
High | High | Enhanced editorial review |
Medium | Low | Additional verification |
Medium | High | Hold or escalate |
Low | Low | Monitor or investigate |
Low | High | Do not publish as established fact |
This is more realistic than using one score for every editorial decision.
Build A Risk-Adjusted Confidence Threshold
Not every story needs the same publication threshold.
Consider three categories.
Low-Risk Stories
Examples might include:
Routine event coverage
Non-sensitive announcements
Basic cultural updates
The newsroom may use a standard verification threshold.
Medium-Risk Stories
Examples might include:
Corporate disputes
Political claims
Financial developments
Significant public-policy stories
Additional corroboration may be required.
High-Risk Stories
Examples include allegations involving:
Criminal conduct
Death
Public safety
Elections
Health
Sexual misconduct
Children
National security
Serious reputational harm
These stories should receive substantially stronger human verification.
A numerical score should never override the editorial judgment required for high-risk reporting.
Confidence Scores Can Help Editors Manage Breaking News
Breaking news creates a difficult tradeoff.
The newsroom wants to move quickly.
But information is often incomplete.
A confidence system can make uncertainty visible.
For example:
08:30 — Low confidence: Social reports indicate an incident.
08:42 — Medium confidence: Emergency service confirms response.
09:05 — Higher confidence: Official statement confirms the event.
09:30 — High confidence: Multiple independent sources confirm key details.
The story can evolve as the evidence improves.
This is better than treating publication as a binary decision:
Publish / Do not publish
A developing story can have an evolving confidence state.
Confidence Scores Should Be Dynamic
The score should not be permanent.
New evidence can:
Increase confidence
Decrease confidence
Resolve uncertainty
Introduce contradictions
Make existing information outdated
The newsroom should therefore recalculate or reassess confidence when meaningful changes occur.
This is especially important for live stories.
A confidence system should record:
Score
Timestamp
Reviewer
Evidence changes
Source changes
Major corrections
Publication updates
That creates an editorial audit trail.
NIST's AI RMF emphasizes continuous risk management, ongoing monitoring, periodic review, and clearly defined responsibilities.
The same principle is useful for AI-assisted publishing.
How AI Can Help Calculate Confidence
AI can assist with the mechanical components.
It can:
Extract claims
Identify source references
Compare documents
Detect conflicting numbers
Check whether claims have citations
Identify dates
Flag stale information
Detect unsupported statements
Compare drafts against approved Fact Packs
Identify claims requiring human review
But AI should not be the only authority determining the confidence score.
The system can make recommendations.
The editor approves the final status.
For example:
AI recommendation: Medium confidence.
Reason: Two sources support the central claim, but no primary document has been located.
Editor action: Approve medium confidence and assign reporter to locate primary evidence.
This is much more useful than:
AI confidence: 78%.
The explanation matters more than the number.
Every Score Should Have Reasons
A confidence score without explanation has limited editorial value.
A better interface might show:
Confidence: 84 — High
Why:
Primary source available
Three independent confirmations
All major numerical claims checked
Quotes verified
Human editor reviewed
One minor contextual gap remains
This makes the score interpretable.
Editors can challenge individual components instead of trusting a black-box number.
Explainability Should Be Built Into The Workflow
NIST's AI RMF emphasizes documentation, accountability and human oversight as parts of responsible AI risk management.
For newsrooms, this means the system should preserve the reasoning behind a confidence recommendation.
A useful confidence record might include:
Score: 86
Status: High confidence
Primary source: Yes
Independent sources: 3
Claims verified: 17 of 18
Conflicting claims: 1
Human review: Completed
Last checked: Timestamp
Reviewer: Editor
Remaining issue: One secondary statistic requires additional context
This is much more valuable than a single badge.
Confidence Scores Can Help With AI-Assisted Drafting
Once the Fact Pack and confidence system exist, the AI drafting system can receive additional instructions.
For example:
Use high-confidence facts as established information.
Attribute medium-confidence claims.
Do not state low-confidence claims as facts.
Preserve uncertainty.
Do not invent missing information.
Flag contradictions for editorial review.
This changes how AI generates content.
Instead of treating every piece of retrieved information equally, the system receives an evidence hierarchy.
That is particularly useful for developing stories.
Confidence Scoring And Source Attribution
Attribution should influence confidence.
Suppose an article says:
“The company will cut 5,000 jobs.”
The Fact Pack might reveal that the only source is an anonymous social post.
The confidence should be low.
If the company's official filing later confirms the figure, confidence can increase.
If three independent publications report the same figure but all cite the same unnamed source, the confidence should not automatically increase substantially.
The system should track the source chain.
This makes source provenance a core part of confidence scoring.
Confidence Scores And AI Search
Confidence scoring can also support the quality of publisher content for search and AI-generated answers.
But publishers should not treat confidence scores as an SEO ranking factor.
There is no established Google ranking signal that says a page receives better visibility because a newsroom internally assigned it an “85 confidence score.”
Google's current guidance focuses on accuracy, quality and relevance when content is automatically generated, and its guidance for AI features continues to emphasize foundational SEO and helpful content rather than special AI-search markup.
The value of confidence scoring is therefore editorial quality, not a direct ranking shortcut.
Better editorial quality can support stronger content, but publishers should not confuse that with a guaranteed search benefit.
Confidence Scoring Can Strengthen Publisher Trust
The Reuters Institute's 2026 report says publishers expect to put more emphasis on distinctive reporting, analysis, human stories, and fact-checking and verification as AI-generated content becomes more widespread.
This creates an opportunity.
Publishers can make their editorial verification processes more systematic.
They may not need to show every internal score publicly.
In fact, displaying unexplained numerical confidence percentages to readers could create more confusion than clarity.
Instead, publishers could communicate process information such as:
Sources checked
Last updated
Reviewed by editor
Correction history
Methodology
AI assistance disclosure where appropriate
The internal score can power the workflow while the public-facing page provides meaningful transparency.
What Should Never Be Automated Solely From A Confidence Score?
A confidence score should never independently decide:
Whether an allegation is published
Whether a person is identified
Whether sensitive information is disclosed
Whether a story is legally safe
Whether a correction is required
Whether a source should be revealed
Whether an anonymous claim is credible
Whether an image is authentic
Whether a high-risk story should be published
Those decisions require human editorial judgment and, when appropriate, legal or specialist review.
The score is a tool.
It is not an editor.
Common Mistakes When Designing Confidence Scores
Mistake 1: Pretending The Score Is A Probability
An 85 score does not automatically mean an 85% probability of truth.
Mistake 2: Giving AI Full Control Of The Score
The AI can recommend a score, but humans should control consequential editorial decisions.
Mistake 3: Ignoring Source Independence
Repeated claims from the same original source should not be treated as independent confirmation.
Mistake 4: Using One Threshold For Every Story
High-risk journalism requires stronger controls.
Mistake 5: Ignoring Time
Breaking news can change rapidly.
Mistake 6: Scoring Only The Whole Article
Individual claims can have different evidence levels.
Mistake 7: Hiding The Reasons Behind The Score
Editors need explanations, not just numbers.
Mistake 8: Using Confidence As An SEO Signal
A confidence score is an internal editorial mechanism, not a known ranking shortcut.
Mistake 9: Treating Human Review As A Checkbox
A reviewer should actually inspect important evidence.
Mistake 10: Never Testing The Scoring Model
A publisher should compare scores against real correction rates and editorial outcomes.
How Publishers Should Build A Confidence Scoring System
Publishers can begin without building a complex AI platform.
Step 1: Define What “Confidence” Means
Write a newsroom policy.
For example:
Confidence measures the strength and completeness of evidence supporting the material claims in an AI-assisted story.
Step 2: Identify Claim Types
Separate:
Established facts
Direct statements
Allegations
Estimates
Predictions
Analysis
Opinions
These should not receive identical treatment.
Step 3: Define Source Categories
Create internal source classifications.
Step 4: Connect Claims To Evidence
Use a Fact Pack or similar evidence record.
Step 5: Create Initial Scoring Criteria
Start with simple factors rather than a complex mathematical system.
Step 6: Add Risk Levels
Define low-, medium- and high-risk story categories.
Step 7: Require Human Approval
Determine which scores require editor approval.
Step 8: Record The Reasons
Every score should have an explanation.
Step 9: Monitor Outcomes
Track:
Corrections
Retractions
Major edits
Unsupported claims
Reader complaints
Source conflicts
Editorial overrides
Step 10: Recalibrate
If stories receiving “high confidence” regularly require major corrections, the scoring system is not working well enough.
A Practical Newsroom Confidence Workflow
A publisher can integrate confidence scoring into its broader AI newsroom operating system.
The workflow can be:
Story Discovery
A trend, source, event, document, or reporter observation identifies a potential story.
Source Collection
The newsroom gathers primary and independent sources.
Fact Pack
Important claims are structured and linked to evidence.
Claim Verification
Editors or reporters check the key claims.
Risk Assessment
The newsroom determines the potential consequences of error.
Confidence Assessment
The system calculates or recommends an internal confidence level.
AI Drafting
The AI works from approved evidence.
Human Editorial Review
A journalist or editor checks the draft.
Publication
The story is published only when it meets the newsroom's requirements.
Monitoring
New evidence can change the confidence level.
This structure prevents AI from becoming the final authority.
NewsBolts Confidence Framework
For NewsBolts, a useful model would combine Confidence, Risk, and Evidence rather than relying on a single score.
Evidence Layer
Tracks:
Sources
Claims
Documents
Quotes
Dates
Data
Verification status
Confidence Layer
Tracks:
Source quality
Corroboration
Claim coverage
Evidence completeness
Recency
Human review
Risk Layer
Tracks:
Potential harm
Sensitivity
Legal exposure
Public impact
Reputational consequences
Editorial Layer
Tracks:
Reporter
Editor
Approval status
Corrections
Updates
This creates a more complete editorial control system.
The score becomes one output from a larger evidence and governance framework.
NewsBolts Research Opportunity
NewsBolts could turn this concept into a first-party research project: an AI-Assisted News Confidence Benchmark.
The goal would be to test whether a structured confidence system actually improves editorial outcomes.
A study could compare stories produced through:
Workflow A: AI-assisted drafting without structured confidence scoring.
Workflow B: AI-assisted drafting with claim-level evidence and confidence scoring.
The research could measure:
Factual errors
Unsupported claims
Correction rates
Editorial intervention
Verification time
Time to publication
Source attribution
Human override rates
Confidence-score calibration
The key metric would be calibration.
If stories classified as high confidence consistently have fewer significant errors than medium- or low-confidence stories, the scoring framework may be useful.
If the categories do not meaningfully predict editorial outcomes, the scoring system needs redesign.
Until such a benchmark is conducted, publishers should treat the framework as an operational proposal rather than a scientifically validated scoring standard.
What Publishers Should Do
Publishers should not begin by asking:
“What confidence percentage should our AI give every article?”
They should ask:
“What evidence must exist before we allow an AI-assisted story to move toward publication?”
That question leads to a better system.
Start with:
A standardized Fact Pack.
Claim-level source tracking.
Verification status.
Risk classification.
A simple internal confidence category.
Human approval requirements.
Reasons behind every confidence decision.
Continuous monitoring after publication.
The first version does not need sophisticated machine learning.
A structured editorial checklist can be more valuable than a complicated algorithm that nobody understands.
Over time, publishers can compare confidence scores against real editorial outcomes and improve the system.
The Future Of AI-Assisted News Needs Better Editorial Signals
AI will continue to increase the amount of information newsrooms can process.
That creates an opportunity.
It also creates a risk.
If the volume of AI-assisted content grows faster than the newsroom's ability to verify it, publishers can produce more stories without necessarily producing more trustworthy journalism.
Confidence scoring is one way to address that problem.
It creates an explicit signal between information processing and publication.
But the most important part is not the score.
It is the discipline behind the score.
A newsroom needs to know:
What the source is
What the source actually establishes
Which claims are verified
Which claims remain uncertain
How current the information is
What risks exist
Who reviewed the material
Why the story is ready to publish
The number simply summarizes that work.
For a modern AI newsroom, that is the real value of confidence scoring.
AI can help create the story faster. Confidence systems can help the newsroom decide whether the evidence is strong enough to publish it.
Frequently Asked Questions
What Is A Confidence Score In AI Journalism?
A confidence score is an internal newsroom assessment of the strength of the evidence supporting an AI-assisted story. It can consider source quality, independent corroboration, claim verification, evidence completeness, recency, risk, and human editorial review.
Is A Newsroom Confidence Score The Same As AI Accuracy?
No. An AI model's accuracy and a story's editorial confidence are different concepts. A story can receive high confidence when it is based on strong sources and human verification even if the AI model itself is not independently rated with a specific accuracy percentage.
Should Confidence Scores Be Shown To Readers?
Not necessarily. A raw numerical score can create false precision or confuse readers. Publishers may find it more useful to communicate meaningful information such as sources, update times, editorial review, methodology, and AI-use disclosures where appropriate.
Can AI Automatically Assign Confidence Scores?
AI can analyze sources, compare claims, identify missing evidence, and recommend a confidence level. However, consequential editorial decisions should remain under human oversight. NIST's AI RMF specifically emphasizes defined human oversight and accountability for AI systems.
What Factors Should Be Used In A Confidence Score?
Useful factors include source quality, source independence, verification coverage, evidence completeness, recency, attribution, human review, and story risk. Publishers should adjust the factors and weights for their own editorial requirements.
Does A High Confidence Score Mean A Story Is Definitely True?
No. A confidence score is not a guarantee of truth and should not be presented as one unless it has been statistically validated for that purpose. It is an editorial decision-support signal.
Should Every News Story Have The Same Confidence Threshold?
No. High-risk stories should generally require stronger evidence and deeper human review than routine, low-risk stories. Confidence should be considered together with potential harm and editorial sensitivity.
How Do Fact Packs Help Confidence Scoring?
Fact Packs connect individual claims to sources, verification status, dates, quotations, figures, and context. This provides the evidence structure needed to make confidence assessments more transparent and auditable.
Can Confidence Scoring Improve SEO?
There is no established Google ranking factor that rewards a publisher simply for assigning internal confidence scores. The benefit is editorial: better verification can help publishers produce accurate, useful and trustworthy content. Google emphasizes accuracy, quality and relevance for automatically generated content.
How Should A Newsroom Test Its Confidence Scoring System?
Compare confidence categories against real outcomes such as corrections, major editorial changes, unsupported claims, source conflicts and verification failures. If high-confidence stories do not consistently outperform lower-confidence categories, the scoring model needs recalibration.
Conclusion
AI-assisted journalism needs more than faster writing.
It needs better ways to understand how much confidence a newsroom should have in the information behind an AI-assisted story.
A confidence score can provide that signal.
But the score should never pretend to be a probability of truth. It should summarize the evidence and editorial work behind a story.
The strongest model combines:
Sources → Fact Pack → Claim Verification → Risk Assessment → Confidence → AI Assistance → Human Editorial Review
This creates a controlled workflow in which AI can accelerate research and production without becoming the final authority on what gets published.
For NewsBolts and other digital publishers, the opportunity is larger than adding a confidence percentage to an editorial dashboard.
The real opportunity is to build evidence-aware publishing infrastructure where every important claim can be traced to its source, every high-risk story receives appropriate review, and every AI-assisted output remains under human editorial control.
That is how confidence scoring can become useful in a modern newsroom.
Comments