The economics of fabrication have collapsed. A few years ago, producing a convincing fake news article took time, effort and some writing ability. Today it takes minutes, costs almost nothing, and produces output that passes casual inspection. Synthetic images illustrate events that never happened. Cloned voices put words in the mouths of officials. Entire websites — dozens of them, on the same template, with invented bylines — publish plausible articles around the clock.
None of this is hypothetical, and none of it is distant. Misinformation in this region spikes predictably around elections, protests, disasters and communal flashpoints — exactly the moments when accurate information matters most and when people are least inclined to check what they share.
This is an account of how Digital News Network verifies what it publishes in that environment: the actual working process, what we look for, and what readers can do themselves.
Why the Old Shortcuts No Longer Work
The traditional reader’s heuristics have all degraded. A professional-looking website proves nothing; one can be assembled in an afternoon. A confident, fluent writing style proves nothing; fluency is what generative tools do best. Even a named byline proves little when bylines can be invented at scale, complete with generated profile photos and fabricated biographies.
The deeper problem is volume. A newsroom’s verification capacity is finite; the supply of fabricated content is effectively unlimited. The asymmetry defines the era: the cost of making a fake is near zero, while the cost of disproving one remains high. Anyone who wants to flood the information environment can do so. Accuracy has become the expensive good.
That is the context in which verification has to work.
Source Verification: Where Every Story Starts
Everything we publish traces back to a source an editor can name. That sounds obvious; it is the rule that most fabricated content cannot survive, because fabricated content either has no source or has one that dissolves when contacted.
Named over anonymous. We prefer sources who speak on the record. Anonymity is granted only where there is genuine risk — professional retaliation, legal exposure, physical safety — and where the information cannot reasonably be obtained another way. At least one editor always knows an anonymous source’s identity, and the story explains why the person is not named. “Sources say” without any accountability attached is how unverifiable claims get laundered into print, and we do not use the formulation.
Two independent sources for contested claims. Any significant claim that cannot be checked against a document requires corroboration from at least two sources with independent knowledge of the matter. One source can be honestly mistaken, and in politically charged situations sources sometimes coordinate their accounts. Independence is the operative word — we ask how each source knows what they know, because two people repeating the same rumour are one source, not two.
Documents over recollection. Where a story concerns money, contracts, official decisions or court matters, we look for the paper trail: circulars, filings, judgments, audit reports, official datasets. A document is not automatically truthful, but it is dated, it is quotable, and it does not adjust its account between phone calls.
Direct contact, always. Anyone criticised in a story is told what will be said about them and given fair time to respond. This functions as a verification step as much as an ethical one — responses regularly produce documents, corrections and context that change the story before it publishes rather than after.
Verifying Images and Video
Visuals carry more persuasive weight than text and are more easily abused. Old footage recirculated as new is the single most common form of visual misinformation in this region — a flood video from three years ago reposted during this year’s flood, protest footage from one country captioned as another. AI-generated imagery has added a second problem: scenes that never existed at all.
Our checks before any visual runs:
Provenance. Where did this come from, and can we trace it to its origin? Material that arrives through messaging apps stripped of context gets treated as unverified until it is not.
Reverse image search. A standard check that catches most recirculated material in seconds. If the image appeared online before the event it supposedly depicts, it is not what it claims to be.
Metadata and technical inspection. Where available, file metadata indicates when and where an image was created. Technical analysis can identify signs of manipulation or generation — though we treat these tools as indicators, not verdicts, because both generation and detection keep improving.
Cross-reference against verified accounts. Trusted wire services and confirmed eyewitness accounts establish what genuinely documented imagery of an event looks like. Visuals that match nothing else from a well-covered event are a warning sign.
Synthetic content is labelled. Where AI-generated or AI-altered visuals are ever used — an illustration, for example — they are marked as such, without exception.
What We Do During Fast-Moving Events
Misinformation travels fastest exactly when verification is hardest: the first hours of a major event. Our breaking news practice is built around that tension.
Publish only what is confirmed, attributed to its source. Early casualty figures, arrest counts and official claims are attributed to whoever provided them, not stated as settled fact. In this environment, early figures are routinely revised — sometimes by an order of magnitude — and the outlet that stated the first number as fact owns the error when it changes.
Say what is not known. “We have not been able to independently verify this” is accurate reporting. During a developing story, the honest description of uncertainty is often the most useful thing we can publish.
Refuse the plausible rumour. The hardest discipline is declining to publish something dramatic and widely shared that we cannot yet stand behind. Competitors will run it. Sometimes it will even turn out to be true. The rule does not change, because a newsroom’s reliability is the accumulated record of exactly these decisions.
Keep the story’s history visible. Developing stories are updated with timestamps as facts firm up. A reader arriving at any point can see what was known when — rather than a story silently rewritten until it says something different from what it said at noon.
Where AI Fits — and Where It Does Not
We use AI tools in the newsroom where they genuinely help: transcribing interviews, translating material, processing large datasets, monitoring public information flows. They save time on mechanical work so reporters can spend it on reporting.
Two rules are fixed. First, a named human editor is accountable for everything we publish — AI output is checked like any other unverified material, and it does not publish itself. Second, material AI involvement in producing content is disclosed. We do not publish machine-generated articles under human bylines, because a byline is a promise that a person stands behind the work.
We also use detection tools to flag potentially synthetic media submitted to us, while remaining clear-eyed about their limits: no detector is reliable enough to be the final word, and human judgement stays in the loop on every decision.
What Readers Can Do
Verification should not end at the newsroom door. A few habits catch the large majority of misinformation:
Check the outlet, not the article. Is this a site you recognise, on its genuine domain? Fake versions of real outlets and professional-looking sites with no traceable organisation behind them are both common. Look for named staff, contact details and a corrections history — an outlet with no visible corrections is not one that makes no mistakes, it is one you cannot catch.
Check whether anyone else is reporting it. A genuinely major event is covered by multiple independent outlets within minutes. A dramatic story that exists on only one site should wait.
Check the date and the image. Old stories recirculate as current, and old images illustrate new events. A reverse image search takes seconds and is the single most effective verification tool available to a non-journalist.
Be slowest when the story is most emotional. Content engineered to make you furious or triumphant is engineered to be shared without checking. The stronger the emotional pull, the more valuable the thirty-second pause before forwarding.
Report what you find. If you spot an error in our work, tell us. Every report is reviewed by an editor, and corrections that originate with readers are among the most useful checks we have.
The Standard We Are Holding
The AI era has not changed what verification is — sources, documents, corroboration, disclosure — but it has raised the cost of doing it and the cost of skipping it. Fabrication is now cheap, fast and fluent. Trust has to be expensive, slow and specific.
That is the trade we have chosen, and it is checkable. Our corrections are visible. Our sources are named or accounted for. Our developing stories show their own history. If we fall short, we would rather hear about it from you than not.
