ChatGPT for Journalism: Research, Drafts, and Verification

ChatGPT can help journalists research background, prepare for interviews, organize notes, and improve the clarity of drafts. Every claim that matters still needs a primary source, and every fact that goes to print needs a person responsible for checking it. A tool like fenbs keeps your verified source links, editorial decisions, and reporting tasks in a shared record that is available across sessions — without any confidential source information going into an AI tool.

Journalism's core obligations — accuracy, independence, source protection, and editorial accountability — apply to how AI tools are used. ChatGPT produces plausible text quickly. That speed is useful for background research and drafting; it is a liability if it leads to publishing claims that were not independently verified.

Where ChatGPT helps in a reporting workflow

The most defensible uses for ChatGPT in journalism are tasks that support a reporter's own research and verification work. ChatGPT helps you prepare and organize; primary reporting and fact-checking remain with the reporter and editor.

• Background research on a topic, institution, or issue to orient before deeper primary research.
• Generating interview question lists organized by the main areas to explore.
• Summarizing and organizing notes from a reporting session.
• Improving clarity, sentence structure, and flow in a draft after the reporting is complete.
• Checking whether a draft's structure logically presents the most important information first.
• Flagging passages that make claims that should have source attribution.

Background research and orientation

ChatGPT can give you a quick orientation on an unfamiliar topic: the key actors, the main debates, the regulatory landscape, or the historical context. That background helps you ask better questions and recognize gaps in the answers you get.

Do not treat ChatGPT's background summary as verified information. Its training data has a cutoff and may not reflect recent developments. It may oversimplify, conflate positions, or describe situations that have changed. Use the background as a starting map, not as reportable fact.

Every factual claim that will appear in a published story needs a primary source you have checked yourself: a document, a statement from someone in a position to know, an official record, or a direct observation. ChatGPT's orientation research points you toward where to look; it does not supply the primary sources.

Preparing for interviews

Interview preparation is one of the strongest uses for ChatGPT in a journalism workflow. Give it the subject's role and background, the topics you plan to cover, and the angle of the story. Ask for a question guide organized by area, with follow-up probes for each main question.

Review the question guide and edit it based on what you know from your research. ChatGPT produces a useful structure; the specific questions that will get the most informative answers depend on what you already know about the subject and what you have already confirmed in your research.

For adversarial interviews — confronting a subject with difficult findings — prepare separately. Know what you have documented and what you still need to give the subject a fair opportunity to respond to. ChatGPT can help you organize those questions by topic, but the substance of what you are confronting them with comes from your verified reporting.

Organizing notes and drafting

After interviews and document review, organizing a large body of notes into a coherent story structure is real work. ChatGPT can help by identifying the main themes in your notes, suggesting a story structure based on the key findings, or flagging the information that seems most important given the stated angle.

For drafting, the safest approach is to write the story yourself with the sources in hand, then use ChatGPT to improve clarity and flow in the revision stage. Ask it to flag sentences that are unclear, passages where the logic does not follow, or sections where a reader would need more context to understand the significance.

Accuracy checking is your responsibility, not ChatGPT's. The model does not know whether a claim is true; it only knows whether the text reads clearly. Do not ask ChatGPT to fact-check your story. Fact-checking requires checking each claim against a primary source — something a language model cannot do reliably.

Source protection and confidential information

Never share a confidential source's identity, identifying details, or the specific information that would allow someone to identify them in a public AI tool. That applies regardless of how the conversation is framed. The journalist's duty to protect sources is not waivable by the source's apparent consent to share information with a reporter in confidence.

For reporting that involves confidential documents, embargoed information, or materials that could be sensitive if exposed, keep those materials in your own secure systems. Do not paste them into an AI tool. ChatGPT can help you think through your approach and structure your questions without seeing the confidential material.

Working with public documents and data

Public records, financial filings, court documents, and government data are common research materials in journalism. When you have a large publicly available document that is not confidential, ChatGPT can help you identify the most relevant sections, understand the structure of financial filings, or explain terms in a regulatory document. Give it the text and ask specific questions about what you are trying to understand.

For data-heavy stories, ChatGPT can help you think through what the numbers mean, what additional data you would need to support or complicate a finding, and what the standard caveats are for the type of data you are using. That analytical support is useful for reporters who are not data specialists. The actual data verification and the reporting of findings still require you to check the numbers against their original source.

Government financial disclosures, lobbying records, and regulatory filings are all public documents where ChatGPT can help you understand the structure and locate relevant details. For filings that use specialized financial or legal terminology, ChatGPT can explain what the terms mean and what the relevant sections typically contain. That speeds the review without replacing your need to read and understand the actual document.

For any documents or data you use in a story, keep a clear record of the source, where you accessed it, and when you accessed it. If the document changes or is removed after publication, your record of what it said at the time is important for accuracy and for any corrections or follow-up coverage.

Editorial accountability and corrections

If an AI-assisted draft contains an error that is published, the responsibility lies with the reporter and editor, not with the AI tool. Being open with your editor about how AI tools were used in the reporting and drafting process is part of editorial accountability. Most newsrooms are developing guidance on AI use; follow your publication's policies.

For corrections, record what was published, what was wrong, and what the correct information is. That record is useful for the editorial process and for the next similar story. Lessons learned from corrections — a type of source that tends to be unreliable, a category of claim that needs extra verification — are worth recording and applying to future work.

How to do this with fenbs

Use fenbs to keep the reporting tasks and verified source library for each story in a shared record that is accessible across sessions and to editors.

1. Create a free account and set up a project for each major story or investigation.
2. Record the key editorial decisions as Decisions: the story angle, the claims to verify, the standards for what constitutes sufficient documentation.
3. As you verify sources, add each one as a link with a brief note on what it confirms. Build the source library as you report.
4. Connect your preferred AI app from Integrations and ask it to read the project context — especially the verified sources and open questions — before working on research or drafting.
5. After each reporting or editing session, update the project with what was confirmed, what remains to be verified, and what the next reporting step is.

fenbs does not hold confidential source information or unpublished materials. It keeps the editorial decisions, verified source links, and reporting tasks that make each session build on confirmed work. Source protection and editorial judgment stay with the journalists.