False and misleading claims can spread around the world in a matter of minutes, making it increasingly difficult to separate fact from fiction.
uncertain
it → spread → fiction
At a time when false claims can spread faster than facts, independent verification is essential.
uncertain
verification → spread → facts
That's where DW Fact check comes in.
asserted
check → come → ?
Allow us to introduce ourselves and explain more about our work.
asserted
us → allow → work
What topics we choose
DW fact-checkers investigate a wide range of claims and stories.
uncertain
checkers → choose → claims
This can include statements made by politicians, government officials and political parties, as well as claims involving celebrities, athletes, business leaders and non-profit organizations.
uncertain
This → include → celebrities
We verify the authenticity of images, videos or audios that go viral on major social media platforms.
asserted
that → verify → platforms
Increasingly, we also investigate whether widely shared content is authentic or has been generated or manipulated using artificial intelligence.
asserted
content → investigate → intelligence
To decide which topics to fact-check, we assess claims based on four criteria:
- Virality: How quickly is a narrative spreading?
- Reach: Is the narrative being shared across multiple platforms and/or in multiple languages?
- Relevance: Why does the topic matter to our target audience?
uncertain
topic → decide → audience
How we verify information
asserted
we → verify → information
At DW Fact check, we investigate the origin of a claim, when and where it emerged, whether it has been manipulated, and who is behind it.
uncertain
who → investigate → it
Whenever possible, we trace claims back to their original source and examine the evidence supporting them.
uncertain
we → trace → them
We then compare the claim with reliable data, documents and expert assessments to determine its accuracy.
uncertain
We → compare → accuracy
To fact-check information, we use a range of open-source intelligence (OSINT) tools and verification techniques.
asserted
we → check → tools
DW has created a dedicated website where you can learn more about the verification workflows and tools we regularly use.
asserted
we → create → workflows
There, you can explore methods for verifying images, videos, audio recordings, sources and text-based claims.
uncertain
you → explore → images
Every fact check includes hyperlinks to the sources used in our reporting whenever possible, allowing readers to review the same evidence our journalists relied on.
asserted
journalists → include → evidence
Like all DW journalism, our fact checks are based on information that has been verified through reliable and independent sources.
asserted
that → base → sources
Whenever possible, we follow the two-source principle and seek confirmation from more than one source.
asserted
we → follow → source
While sourcing information from interview partners, we always select reliable and unbiased people.
asserted
we → source → people
All interviews are recorded, with the consent of the interview partner, and archived.
asserted
interviews → record → partner
We mention all sources by name and occupation and, where possible, link to their contact information so there is no lack of transparency.
asserted
We → mention → transparency
Anonymity is only granted to sources in rare cases when their security is at stake or the topic is sensitive.
asserted
topic → grant → stake
In these cases, we are transparent about the reasons for granting anonymity.
asserted
we → grant → anonymity
How we rate claims
DW uses a collaborative process to evaluate claims and determine verdicts.
uncertain
DW → rate → verdicts
Before publication, the lead reporter recommends a verdict to two editors.
asserted
reporter → recommend → editors
The rating system has eight categories:
- True: The statement is accurate.
- False: The statement is not accurate.
asserted
statement → have → categories
- Misleading: The statement is partially correct, but the information is portrayed in a way that creates a wrong impression.
- Unproven: The statement cannot be proven true or false.
asserted
statement → portray → impression
- Fake: The image, video or audio is manipulated.
asserted
image → manipulate → ?
- AI- generated: The image, video or audio was created using artificial intelligence.
asserted
image → generate → intelligence
- AI- manipulated: The image, video or audio was altered using artificial intelligence.
asserted
image → manipulate → intelligence
How we use artificial intelligence
DW has clear editorial guidelines when it comes to the use of artificial intelligence in journalism.
asserted
DW → use → journalism
AI can help us improve workflows, accessibility, translation, data and content analysis and format adaptation.
asserted
us → help → workflows
However, humans remain responsible for every journalistic decision and every item we publish.
asserted
we → remain → decision
You can read DW’s ten AI editorial guidelines here.
asserted
You → read → guidelines
At DW Fact check we clearly label any content that we verify as having been generated or manipulated using artificial intelligence.
asserted
we → label → intelligence
We also explicitly disclose when we have used AI tools to help us verify content.
asserted
us → disclose → content
As part of our verification workflow we may use AI detectors to help assess whether content has been created or manipulated with AI.
uncertain
content → use → AI
But because these detectors often produce inaccurate results, we use them with great caution.
asserted
we → produce → caution
We use multiple tools to reduce the risk of error.
asserted
We → use → error
…and 30 more, not listed.
The Aporia
