AI for journalists: a practical guide to using it without losing trust

AI for journalists isn't one product, and AI in journalism isn't one decision. It's a set of assists spread across the reporting workflow, and many of them already live inside software you use: transcription apps, search, translation, your CMS. The useful question isn't whether to use AI, but which parts of the job are mechanical and which parts are judgment.

Mechanical work is turning a 40-minute interview into text, finding the three paragraphs in a 200-page council packet that mention the budget, shaping rough notes into a structured first draft, writing five headline options, alt text and a meta description. Judgment is deciding what's news, who to call, whether a source is credible, what to leave out and how to frame a story fairly.

AI is good, and getting better, at the first list. It's unreliable at the second, because a language model predicts plausible text instead of retrieving facts. Google's guidance on generative AI content says it plainly: fact-check and review everything AI produces before you publish it.

The most useful AI for journalism work today sits in five places, all of them before or after the reporting itself:

• Research and documents. Summarize long reports, search a folder of PDFs, pull every mention of a name or a contract. Open the original page before you quote it.
• Interviews. Automatic transcription turns hours of tape into searchable text. Check every quote you use against the recording.
• Drafting. Give the AI your verified notes, quotes and facts and ask for a first draft in your house style. The better the input, the less you rewrite.
• Editing and packaging. Headline options, summaries, social copy, SEO titles and meta descriptions, alt text, a second-language edition.
• Audience and search. Find the questions readers actually type into search engines about a topic, so your story answers them early.

For specific tools in each category, with what they do well and where they fail, see our guide to AI tools for journalists.

The safest workflow keeps the AI downstream of your reporting. It shapes and speeds up material you gathered; it never becomes the source.

1. 1

   Start from your own reporting

   Feed the AI your notes, transcripts, documents and data instead of asking it what happened. A model asked an open question fills the gaps with confident guesses.

2. 2

   Ask for structure, not facts

   A good prompt: “Turn these notes into a 600-word story in inverted pyramid, using only the facts below.” A risky one: “Write a story about the city budget.”

3. 3

   Verify every claim

   Check names, titles, figures, dates and quotes against the primary source. In 2023 the Associated Press told its staff to treat generative AI output as unvetted source material — a sound default for any newsroom.

4. 4

   Edit for voice and judgment

   Rewrite the lede, cut what doesn't matter and add the context only you know. This is where the draft becomes your story.

5. 5

   Keep a record

   Note which tools you used and what material you gave them. If a correction comes, you'll know where the error entered.

• Publishing lightly edited output. AI copy reads fluently, which makes errors harder to spot, not easier.
• Leaking sensitive material. A source's name, a leaked document or an embargoed release pasted into a consumer tool may leave your control. Read the data policy first, and keep sensitive work in tools your organization has approved.
• Letting AI flatten your voice. Generic copy is easy to produce and easy to ignore. Your style, your local knowledge and your sourcing are what readers come for.
• Scaling up thin content. The worst kind of “AI journalism” — dozens of rewritten stories published to chase search traffic — is exactly what Google's spam policies describe as scaled content abuse.
• Passing off AI images as photos. Never present a generated image as a photograph of a real event. If you run an AI illustration, label it in the caption.

Not for being AI-assisted. In its February 2023 guidance, Google said appropriate use of AI or automation isn't against its guidelines, and that its systems aim to reward original, high-quality content however it's produced. What it does act against is using generative AI to churn out many pages without adding value for users.

Google's guide to people-first content frames this as three questions: who created the content, how, and why. It encourages accurate bylines where readers expect them, and notes that giving AI an author byline is probably not the best way to disclose its role. In practice: a real journalist's name on the story, and a short note on how AI helped when readers would reasonably want to know.

AI is also changing how people reach your work. The Reuters Institute's Digital News Report 2026 found that 10% of people use AI chatbots for news each week, up from 7% a year earlier. For the rest of the search picture, read SEO for news publishers.

The relationship between AI and journalism is, above all, a question of trust, and readers are wary. The Digital News Report 2025 found audiences remain skeptical of news made mainly by AI, though more comfortable with journalists using it in a supporting role. And trust in news overall stands at 37% across the 48 markets of the 2026 report, the lowest since the Institute began measuring it in 2015.

• Always label AI-generated images and illustrations.
• Add a short note when AI did substantial work, for example: “Drafted with AI from the reporter's notes; reported, edited and fact-checked by the author.”
• Publish your AI policy and link it from your standards or about page.
• Skip the label for routine help such as spell-checking or transcription; it adds noise without informing anyone.

Keep it to one page. A policy people read beats a long one nobody opens. Cover these points:

• Which tasks AI may be used for, and which it may not.
• What must never go into an AI tool: confidential sources, unpublished documents, readers' personal data.
• The verification standard: every fact checked against a primary source, every quote against the recording or notes.
• Who is accountable: the person whose name is on the story.
• When and how you disclose AI use to readers.
• Who approves new tools, and when the policy gets reviewed.