Template for a robust newsroom generative AI policy

Every news organization needs an artificial intelligence policy. The following is designed to be a starting point for newsrooms as they explore various uses for AI in their work.

This policy was developed by Kelly McBride, Alex Mahadevan and Tony Elkins, based on input from more than 75 news leaders during AI, Ethics and Journalism summits in 2024 and 2025. It will be updated often. It was last updated on June 1, 2025  (updates listed at end)

Please send us your edited version so we can see how you are proceeding. (ethics@poynter.org).

___(Your newsroom here)_______ guidance and policies on using AI in our work.

Last updated: _______________

Generative artificial intelligence is the use of large language models to create something new, such as text, images, graphics and interactive media. These terms will be referenced throughout this policy:

Generative AI: A type of artificial intelligence that creates new content, such as text, images, or media, by interpreting and generating based on input data.

Large language models (LLMs): AI systems trained on vast datasets of text to understand and generate human-like language, and is the information backbone that powers Generative AI.

AI prompt: A specific input or instruction provided to an AI tool to generate a desired output.

Hallucination: The phenomenon where AI generates information or responses that are fabricated, inaccurate, or not grounded in fact.

Training data: The dataset — articles, research papers or social media posts — used to teach an AI model patterns, relationships and knowledge for making predictions or generating content.

Although generative AI has the potential to improve newsgathering, it also has the potential to harm journalists’ credibility and our unique relationship with our audience.

As we proceed, the following five core values will guide our work. These principles apply explicitly to the newsroom and throughout other non-news departments including advertising, events, marketing and development.

Transparency

When we use generative AI in a significant way in our journalism, we will document and describe to our audience the tools with specificity in a way that discloses and educates. This may be a short tagline, a caption or credit, or for something more substantial, like an editor’s note. When appropriate, we will include the prompts that are fed into the model to generate the material.

Accuracy and human oversight

All information generated by AI requires human verification. Everything we publish will live up to our long-time standards of verification. For example, an editor will review prompts, and any other inputs used to generate substantial content, including data analysis, in addition to the editing process in place for all of our content.

We will actively monitor and address biases in AI-generated content, ensuring fairness and equity in our journalism. Our AI steering committee, made up of a diverse group across the Poynter Institute, will regularly evaluate and update our standards to ensure uses and tools are equitable and minimize bias.

Privacy and security 

Our relationship with our audience is rooted in trust and respect. To that end, we will protect our audience’s data in accordance with our newsroom’s privacy policies. We will never enter sensitive or identifying information about our audience members, sources or our own staff into any generative AI tools.

As technology advances and opportunities to customize content for our audience arise, we will be explicit about how your data is collected — in accordance with our organization’s privacy policy — and how it was used to personalize your experience.

We will disclose any editorial content that has been created and distributed based on that personalization.

Accountability 

We take responsibility for all content generated or informed by AI tools. Any errors or inaccuracies resulting from the use of these tools will be transparently addressed and corrected. We will regularly audit feedback forms and incorporate audience feedback into policy updates. And violations of this policy will require retraining and possible disciplinary action.

Exploration

With the five previous principles as our foundation, we will embrace exploration and experimentation. We will strive to invest in newsroom training so every staff member is knowledgeable about the responsible and ethical use of generative AI tools.

Logistics

The point person/team on generative AI in our newsroom is ___________, who is supported by the following AI committee members: _______. Coordinate all use of AI here. This team will also be the source of frequent interim guidance distributed throughout our organization.

The team will seek input from a variety of roles, particularly those who are directly reporting the news.

You should expect to hear at least monthly communication from this team with updates on what we are doing and guidance on what activities are generally approved.

In addition, members of this team will:

• Monitor your content management systems, photo editing software and business software for updates that may include AI tools. Because software changes quickly and AI is being added to nearly every technology product, it’s important to delegate appropriate team members to stay knowledgeable of updates.
• Write clear guidance about how you will or will not use AI in content generation.
• Edit and finalize our AI policy and ensure that it is both internally available and where appropriate, publicly available (with our other standards and ethics guidelines).
• Seek input from our audience, through surveys, focus groups and other feedback mechanisms.
• Manage all disclosures about partnerships, grant funding or licensing from AI companies.
• Understand our privacy policies and explain how they apply to AI and other product development. This includes regularly consulting with editors, lawyers or other privacy experts that influence newsroom policies.
• Innovate ways to communicate with the audience to both educate them and gather data about their needs and concerns.
• Outline a clear process on how the policy will be updated, as specific as the number of meetings per month of your committee and who is on the committee, etc.

All uses of AI should start with journalism-centered intentions and cleared by the appointed AI committee. Human verification and supervision is essential. Here’s the form you should use:

[Create a form to collect information, suggested questions]

How do you want to use AI?

What is the journalistic purpose of this work?

How can you gather knowledge on audience needs and attitudes about your intended use?

How should the audience’s needs and attitudes inform your AI use?

How will you fact-check the results?

Will any material be published?

Which journalists will be responsible for overseeing this work and reporting out the results?

Which editors or managers will oversee the work?

What are the risks, the bad things that might happen? (ie. hallucinations, copyright or legal issues, data privacy violation) What safety nets can you devise to intervene before negative outcomes harm your newsroom’s reputation?

What are the privacy implications of this use, and how will we protect user data?

Have we run this by the experts who create and maintain our privacy policy?  

Editorial use

Approved generative AI tools

Here is a list of tools that are currently approved for use at [your newsroom] Please reach out to [AI point person]  with any new tools you’d like to start using, and we can update the list pending an audit.

[List any AI tools you are comfortable encouraging your colleagues to use. This may include those from companies with which you have a formal partnership. For example:

• ChatGPT
• Claude
• Perplexity
• NotebookLM
• Otter.ai
• Google Pinpoint
• Fireflies.ai
• Beautiful.ai
• Adobe Firefly
• Apple Intelligence

Existing tools (Zoom, Canva, etc.) that have added AI capabilities.]

Entering our content: [Your policy will either approve entering your content into a specific LLM that you have a partnership with, or forbid entering your content into any LLMs]

We encourage the use of generative AI to improve efficiency and automate routine tasks. In upholding the five principles of AI use in our organization, these caveats apply:

• [Begin with principles you have determined to be important based on the questionnaire from earlier, and a discussion of your newsroom’s values.]
• Preserve our editorial voice: We will be cautious when using AI tools to edit content, ensuring that any changes maintain Poynter's editorial voice and style guidelines.
• Avoid full writes and rewrites: Generative AI tools will not be used for wholesale writing or rewriting of content. We will use them for specific edits rather than rewriting entire paragraphs or articles.
• Proprietary content: We will not input any private or proprietary information, such as contracts, email lists or sensitive correspondence into generative AI tools.
• Verification: We will be mindful that generative AI tools may introduce errors, misinterpret context or suggest phrasing that unintentionally changes meaning, and will review all suggestions critically to ensure accuracy.
• Disclosure: In most cases, we will disclose the use of generative AI. Our goal is to be specific and highlight why we’re using the tool to better engage with readers.

Research

We may use generative AI to research a topic. This includes using chatbots to summarize academic papers and suggest others, surface historical information or data about the topic and suggest story angles. Generative AI tools may be used by fact-checkers to find checkable claims to pursue, or by journalists to sift through social media posts for article topics.  A reminder: These tools are prone to factual errors, so all outputs will be verified by researchers, reporters and editors.

Transcription

We may use generative AI to transcribe interviews and make our reporting more efficient. Our journalists will review transcriptions and cross-check with recordings for any material to be used in articles or other content.

Translation

We may use generative AI tools to translate material for article research. We may also use those tools to translate article content to reach new audiences, which will always be reviewed by an expert in the language and include the following disclosure: This article/audio/video was translated using generative AI to be able to reach new audiences. It has been reviewed by our editorial team to ensure accuracy. Read more about how and why we use AI in our reporting [here]. Send feedback [here].

Searching and assembling data

We may use AI to search for information, mine public databases or assemble and calculate statistics that would be useful to our reporting and in the service of our audience. Any data analysis and writing of code used on the website will be checked by an editor with relevant data skills.

Headlines or search engine optimization

Our journalists and editors may use generative AI tools to generate headlines or copy to help our content appear more prominently in search engines. We will put enough facts into the prompt that the headline is based on our journalism and not other reporting.

Summary paragraphs and repackaged content

If you want to add summary bullets at the top of an article — or in other formats across the website — you can use a generative AI tool to do so. For summaries, use the following disclosure: This summary was generated with artificial intelligence to give readers a quick overview, and help us save time to focus on deeper reporting. It has been reviewed by our editorial team to ensure accuracy. Read more about how and why we use AI in our reporting [here]. Send feedback [here].

Similarly, these tools may be used to summarize or repackage articles or other content for new editorial or teaching uses, such as in newsletters, quizzes or courses (see social media content for additional guidance.) For repackaged content, we will use the following disclosure: This [newsletter/lesson/course module] was generated with artificial intelligence based on [the original source] to [provide value to the reader]. It has been reviewed by our [editorial/teaching team] to ensure accuracy. Read more about how and why we use AI in our reporting [here]. Send feedback [here].

Copyediting

Generative AI may be used as a tool to assist with copyediting tasks, such as identifying grammar issues, suggesting style improvements or rephrasing sentences for clarity.

Social media content

Generative AI tools can be used to summarize articles to create social media posts. For infographics, use the above disclosure. To avoid label fatigue, we will not require labels for social media posts that include AI-generated summaries, as long as a producer and editor review content, and link to this policy in Linktree or other resource list on social platforms. Audience teams should do regular content audits to ensure social copy and posts meet ethical guidelines, including those around biases.

Visuals

[The use of visual generative AI tools exposes newsrooms to the most risk in regards to  audience trust, and should be discussed in depth at your newsroom. It’s recommended to have at least one dedicated meeting of the AI steering committee devoted to the topic, and seek feedback from diverse audiences  before any substantial experimentation — such as using AI to recreate realistic illustrations,  scenes or cloning your own reporters’ likeness or voice.]

[These guidelines are specific to a newsroom that has chosen to pursue AI use in visuals, and may vary widely on your internal discussions, along with answers to the questionnaire from the previous section of the template. ]

[Your newsroom name here] holds AI-generated visuals to the same rigorous ethical standards as all forms of journalism. Because images shape perception instantly and powerfully, our use of generative AI in visual storytelling is governed by principles of truth, transparency and audience trust.

These guidelines apply to all AI-generated or AI-assisted visual materials, including illustrations, composites, animations and enhanced photographs. Every visual must serve a clear editorial purpose and uphold our responsibility to inform, not mislead.

Humanity first

When a scene can be documented ethically and accurately by our journalists, human coverage is the preferred option.

AI-generated visuals may only be used when:

• They are essential to the audience’s understanding
• The image is impossible or inappropriate to obtain through traditional means
• Example: [Insert newsroom-approved use case.]

Accuracy over aesthetics

AI photo enhancement tools (e.g., sharpening, lighting correction, denoising) must reflect reality, not dramatize or distort it, see AP’s guidelines on page 11 for photos. Edits that exaggerate emotion, alter mood or misrepresent the scene violate visual ethics. For example, deepening shadows to heighten drama in disaster imagery is not permitted. All enhancements must be disclosed internally and reviewed against the original.

Review and verification
Given the rise of AI generation tools for the public, editors and journalists must be vigilant about analyzing reader-submitted content. Media verification must rely on multiple methods — metadata checks, source verification, AI-assisted forensics — and never on one tool. Verification decisions must be documented internally for future review and accountability. [Insert list of approved visual forensics tools, such as InVid, TinEye or AP Verify.]

No manipulation of real people or events

We do not use AI to create or alter depictions of real people or places unless clearly disclosed and editorially justified. This includes recreating faces, changing expressions, or adding or removing individuals from scenes. We [will not/may] use AI to simulate likenesses of staff or sources in news reporting. [This is done with the approval of said staffer, and only to reach new audiences, such as for multiple language translations.]

Disclosure

AI-generated illustrations or composites must be clearly labeled. Captions should disclose the method and source of generation. Here is an example, in which a reporter is working on a story about a rescue in a cave at a National Park: This image was generated using Adobe Firefly based on geological maps, 3D topographic data and field notes from our reporter’s visit to the site. It is a visual approximation of the cave’s interior and is intended to help readers understand the story’s context. No photographs were taken inside the cave due to access restrictions. This illustration has been reviewed by our editorial team for accuracy. Read more about how we use AI in our reporting [here]

[You may or may not choose to include the specific prompt used in the generation process. Prompt used: “Interior of a natural limestone cave system with uneven rock walls, narrow passageways, and underground pools, based on elevation maps and field notes. Dim natural lighting from overhead shafts, no artificial lighting, realistic texture, no people.”]

Product development

[Your newsroom name here] recognizes that AI-driven personalization and product tools shape how audiences discover, understand, and engage with journalism. We treat these systems with the same ethical rigor we apply to our reporting: prioritizing transparency, fairness, human oversight, and audience trust.

This section applies to all AI used in product design, including chatbots, recommendation engines, search assistants, and personalization algorithms. All tools must serve the public interest—not just engagement metrics.

Human-in-the-loop

AI tools will be reviewed during the development process by editors, product leaders and/or designated AI steering committee members. [Insert review or approval process here.]

Inclusive design and bias mitigation

All AI product tools must be tested for differential performance and exposure across topics and audience segments. [Insert who conducts testing and mitigation steps.]  [Describe how underserved communities or topics are monitored for visibility.]

Break the bubble

We will design personalization tools to expose audiences to a broad range of stories and viewpoints. [Insert method for “forced serendipity” or algorithm overrides.]

Transparency and user control

Audiences must be informed about AI-driven personalization. [Insert where disclosures appear — e.g., “About this summary,” tooltips, FAQ pages.] Include language such as:

“This feature uses generative AI to summarize content based on your reading history. It was reviewed by our editorial team. [View our AI policy.]” Users will be able to [enable/disable/edit personalization] via [location of controls].

Audience education

AI tools must help audience members understand complexity, not just consume more content. [Insert what kinds of tools you prioritize: timelines, explainers, layered context.]

Clear off-ramps and fallback options

Every AI-powered feature must include a way for users to access full content or request human help. [Insert what your product offers: “Read full article,” “Report this response,” etc.]

Comment moderation requires transparency

AI comment tools will follow the newsroom’s code of conduct and moderation guidelines. [Insert how rules are communicated and how flagged users are informed.]

Staff will review [daily/weekly] moderation logs. Appeals can be submitted via [method].

Privacy is non-negotiable

All data used for personalization must comply with [Your newsroom’s] privacy policy. [Insert data handling rules, what data is off-limits, and what tools are approved.]

Open development and accountability

All AI product features will be documented with:

• A named point person
• Development goals
• Stakeholder consultation (journalists, technologists, ethics leads, audience members)
• Correction and error response process

Public service over product novelty

AI tools must meet a specific user need that aligns with our mission. [Insert how tools are pitched, approved, or measured.]

Ongoing training

Regular training on AI tools and experiments will be available and at times even mandatory. This training will be delivered by a combination of members from the internal committee and outside experts.

Environmental impact

[Your newsroom name here] acknowledges the energy demands associated with training and deploying large-scale AI systems. As part of our commitment to sustainable journalism, we recognize that responsible AI use includes minimizing our environmental footprint.

We commit to:

• Prioritizing efficient tools
• Advocating transparency from vendors and AI companies
• Offsetting responsibly

Commitment to audience AI literacy

Along with this AI policy, we have developed an AI literacy [page/article] to help our audience understand how and why we’re using generative AI. This material will be regularly updated to reflect our most current experimentation. As our language evolves, we will be better able to describe specific AI applications and tools. On the page, we link to resources, articles and other materials to/this article will:

• Help our audience understand the basics of generative AI.
• Explain why newsrooms use AI in their work.
• Build a more robust vocabulary for describing AI.
• Avoid AI-generated misinformation.
• Use chatbots responsibly to seek out factual information.
• And create responsibly using new generative AI tools.

Updates

June 1, 2025

- Added visual and product guidelines and section on commitment to environmental sustainability

March 27, 2025

- Paragraph starting  “Use reliable sources…” was added under “Custom GPTs”

Aug. 8, 2024

- Added in fifth value of privacy and security

- Enhanced transparency values

- Expanded the role of the committee members to include expertise on privacy

- Added in material to encourage newsroom embrace AI literacy as part of news literacy responsibilities

- Added guidelines for product teams and technologists