How DW uses artificial intelligence in journalism

AI helps improve workflow, accessibility, translation, data and content analysis, and format adaptation.

However, the involvement of AI does not change our editorial standards. Humans remain responsible for every journalistic decision and every article we publish.

We use AI only when it serves clear editorial, public interest, or accessibility purposes. We don’t use AI just because it’s new or attention-grabbing.

Trust, accuracy and accountability remain the cornerstones of our journalism.

Principles for the use of AI in journalism at DW

1) Public value first

We only use AI when it provides a clear benefit. For example, we can improve accessibility, speed up daily production steps, help process large datasets, support translation, and make it easier for our viewers to access our journalism. AI should enhance journalism, not hinder it.

2) Editorial rights always remain with journalists

We do not publish AI-generated output without human review. Trained journalists must always make editorial decisions according to “human participation” principles, and DW reporters and editors are always responsible for the quality of content.

This means:

• AI may assist in drafting, translating, summarizing, generating titles and teasers, subtitles, moderating comments, or processing data. However, editorial rights still remain with journalists.
• Check facts, check sources, and assess context before publishing.
• Further uses of AI require additional editorial, legal, and technical reviews.

3) AI should support journalism, not replace it.

At DW, AI is not intended to replace journalism, but to support it. Quality journalism requires verification, editorial judgment, ethical considerations, understanding of context, and accountability. These responsibilities cannot be delegated to machines. While AI may be useful for repetitive or time-consuming tasks, it cannot replace reporting, research, editorial responsibility, or human relationships with sources and audiences.

4) Accuracy, Verification, and Source Integrity

We do not treat the output produced by AI as a source of information itself. AI systems can produce output that is inaccurate, incomplete, or misleading, even if it appears convincing at first glance. When AI supports research or production, we always independently verify the claims, confirm the original source, and assess whether the output meets DW’s editorial standards.

We pay particular attention to the potential involvement of AI in relation to breaking news, crisis and conflict reporting, investigative reporting, and reporting on sensitive topics where mistakes can harm people or undermine trust.

5) Responsible use of synthetic media

Synthetic media, or content that is generated or modified by AI, may be useful in limited situations. Examples include accessibility, translation, and format adaptation where editorially justifiable and ethically appropriate. However, it also comes with risks that affect audience trust. Therefore, we apply strict standards, including:

• Synthesized speech must be clearly labeled.
• We do not create photorealistic AI images or videos, especially if they could be mistaken for documentary evidence or real-life events.
• When we use synthetic media generated by third parties, such as coverage of AI and its capabilities or fact-checking, we always clearly label it.
• When we use synthetic media, whether generated by us or a third party, we identify it as “AI-generated” and are responsible for the editorial context in which it appears.

6) Transparent label

We will be transparent and clear when using AI in ways that our audience could reasonably consider important, particularly in the creation, presentation, and delivery of content. Clearly label AI-generated or AI-modified content when it could be mistaken for authentic, human-created or real-world material.

This is especially true if:

• synthetic voice
• Images and videos generated or modified by AI
• Chatbots, recommendation systems, and other algorithmic interfaces that help users understand the role of automation.

7) Approved, safe and legal use

We only use tools and workflows that meet DW’s standards for security, editorial integrity, legal compliance, and data protection. We do not upload highly sensitive information or editorial materials to unauthorized systems. Carefully evaluate the limitations, risks, and terms of use of any third-party tools you rely on.

8) Equity, diversity and inclusion

AI systems can reproduce or amplify biases that exist in training data and design choices. We therefore pay close attention to the inherent risks of stereotyping, exclusion and discrimination. At the same time, we also use AI that can improve accessibility and inclusion, such as through translation, subtitles, or more accessible formats.

9) Environmental responsibility

We recognize that training and running large-scale AI systems can consume large amounts of energy and resources. Environmental impact is a key factor when evaluating whether and how to use AI. We aim to choose proportionate solutions and use AI when the editorial or public interest warrants the cost.

10) Accountability and continuous review

As AI technology advances rapidly, we continually review our standards. We carefully test new applications, learn from pilot projects, and update our practices as tools, risks, regulations, and audience expectations evolve. Responsibility always lies with DW and its editorial leadership.

FAQ:

Why is DW using AI in journalism in the first place?

We use AI where it provides real-world benefits, such as improving accessibility, supporting translations and subtitles, creating titles and teasers, helping process large amounts of information, and streamlining repetitive production tasks. The purpose is to enhance journalism and service to audiences, not to use AI for its own sake.

Will DW journalists be replaced by AI?

No, while AI can support certain tasks, it cannot replace reporters, editors, or producers. We do not use AI to write entire articles or scripts, nor do we replace human journalists with audio or video content. Editors remain responsible for verification, judgment, ethics, context, and final publication decisions.

What does “human stakeholder” actually mean?

This means remaining accountable at critical points in the workflow, such as selecting use cases, checking output, fixing errors, determining suitability, and approving publication. AI output is never treated as independent editorial content.

How do I know if AI was used?

Clearly label when AI is used in ways related to understanding content, especially when content is synthesized, modified, or highly automated. Where useful, we also explain how and why AI was used.

Why does DW sometimes use synthetic speech?

Synthesized speech may be useful for accessibility, translation, multilingual services, or other clearly legitimate editorial purposes. However, we use audio sparingly and label it clearly because audio is closely tied to trustworthiness and trustworthiness.

Does DW use AI-generated images or videos?

Only in limited and clearly controlled circumstances. We pay particular attention to photorealistic composite visuals, as they may be interpreted by viewers as evidence. Where such materials are used, such as reports about AI and its capabilities, they will be clearly labeled.

Can AI write DW news articles?

No, while AI may assist with day-to-day production tasks such as summarizing, translating, and adjusting content formats, we do not delegate editorial responsibilities to AI. Publishing journalism requires human verification, editing, and approval.

How do you deal with mistakes, illusions, and prejudices?

We assume that AI systems can be wrong. That’s why we verify our output, check our sources, test the reliability of our systems, and stay on the lookout for stereotypes, discrimination, and contextual errors. AI can produce convincing but flawed results, so human oversight is essential.

What about privacy, copyright, and data security?

We use tools and workflows that meet strict legal, editorial, and security requirements. Sensitive materials should not be placed on unauthorized systems. We are also taking a critical look at the origins of training data and the obligations placed on external tools and providers.

What about the impact of AI on the environment?

AI can require significant computing power and energy. We take that into account when choosing tools and use cases, weighing the environmental costs against editorial and public value.

What if an external third party provides AI-generated content?

AI material generated by third parties should be carefully scrutinized to determine whether its use is editorially necessary. In some cases, such as reporting on AI, it is necessary to include AI images and videos generated by third parties. If such material is used, we will properly identify its source and label it as AI-generated. DW is responsible for the context in which it appears.