How newsrooms can build an AI policy that builds reader trust

Here's an uncomfortable truth for any newsroom: your readers probably already think you're using AI. Whether you are or not.

‍

That was the starting point for a recent Indiegraf webinar with Laura Davis, Indiegraf's Entrepreneur in Residence for AI, and Lynn Walsh, Assistant Director at Trusting News. Their argument was simple. If your community assumes AI is in your workflow, staying quiet doesn't protect you. A clear policy does.

‍

The good news is that getting one published is far easier than people think. You don't need a big team or a six-month committee process. You need to answer a few honest questions and put the answers somewhere people can find them.

‍

This piece walks through what the research says, what real newsroom policies look like, and how to draft your own.

‍

Why your readers already assume you use AI

The assumption is now baked in. A Reuters survey across six countries found that most people believe journalists use AI always or often. A University of Minnesota study found that more than a third of Americans believe journalists use AI to write stories at least sometimes, and only 8% think it never happens.

‍

As Davis put it, the perception exists in the aggregate no matter what any single newsroom is actually doing. That's the problem a policy solves. You're not introducing the idea of AI to your audience. They already have one. You're correcting it with the facts.

‍

"Whether you're using AI or not, whatever you're doing with it, your audience probably assumes that you are."

‍

What audiences actually want from newsrooms

Trusting News has spent three years asking news consumers a single question: how can journalists use AI without losing trust, and maybe even build it?

‍

The answer comes back consistently. People want transparency, and they want it strongly.

‍

• In one Trusting News survey, 94% of people wanted disclosure. A follow-up with the Local Media Association pushed that to 98%.
• The share of people who expect journalists to have clear ethical guidelines around AI rose to 72% this spring, up 10 percentage points from the previous fall.
• Over the same period, the share saying journalists should never use AI dropped to 20%.

‍

In other words, outright opposition is shrinking, but the demand for guardrails is growing. Most of the audience is landing in the same place: it's fine to use AI, as long as you're honest about it and you have rules.

‍

A separate study of around 2,000 people in Chile reinforced the point. Researchers asked participants to compare how different newsrooms handled and disclosed AI. Two things consistently raised perceived credibility: human oversight and disclosure of use.

‍

The two things every newsroom needs

Davis and Walsh break the work into two pieces. 

‍

1. A general AI policy on your website. One findable page explaining how you approach AI overall.
2. Story-level disclosures. A short note on the specific stories where AI played a role.

‍

You need both. Most people will never stumble onto your policy page, so the disclosure on the story is what reaches them in the moment. 

‍

When researchers tested in-story disclosures, the ones that linked back to a fuller AI policy performed better. People reacted more positively when "we used AI here" came with "and here's our thinking on all of it."

‍

The three questions your AI policy has to answer

A useful AI policy answers three questions your audience is actually asking:

• Are you using AI, and if so, how?
• How will you never use AI?
• What can you promise us?
  
  

Underneath those, readers want to know why you're using it, how you're keeping the work accurate, and whether a human is still involved. That last point matters most. Across the research, people are simply not comfortable with content created solely by AI with no human touching it before it goes live. The "human in the loop" is the thing they're checking for.

‍

What good newsroom AI policies look like

The webinar walked through several real examples worth borrowing from.

‍

Energetic City: easy to find, honest about risk

Energeticcity.ca, an Indiegraf publisher, keeps its AI policy linked right from the About Us menu, clearly labeled. It opens by acknowledging the risks of AI, lays out acceptable and unacceptable uses, and states that the newsroom retains ultimate accountability for anything it publishes.

‍

That accountability line does a lot of quiet work. It tells readers a person stands behind the output.

‍

‍

Nucleo: the FAQ approach

Nucleo, a Brazilian newsroom that has leaned into AI early, including some image generation, structures its policy as a simple FAQ. A short intro explains why they use AI and invites feedback, then three questions:

‍

• What might we do with AI?
• What will we never do?
• What do we promise you?

‍

The FAQ format is easy to read, answers the questions people care about most, and stays flexible. If readers start asking about, say, environmental impact, you add a question. It becomes a living document rather than a frozen legal statement.

‍

‍

USA Today: the story-level disclosure done right

At the bottom of relevant stories, USA Today notes that the piece was created by a named reporter with the assistance of AI, that journalists were involved in every step of gathering, reviewing, editing, and publishing, and links to the full policy. That single block hits every point the research says audiences want: who, what, human oversight, and where to learn more.

‍

‍

VNN: a policy for not using AI

This is the part many newsrooms miss. If you don't use AI, say so, loudly.

‍

VNN, a local newsroom serving Indigenous communities in Oklahoma, states plainly that it does not currently use AI in news production, has no current plans to, and will tell readers if that changes. As Walsh noted, that kind of statement shows the newsroom knows the technology exists and made a deliberate choice. You should get credit for that thoughtfulness, and the research suggests there's a real appetite for "by humans, for humans."

‍

"Don't let perfection stop you from getting something out there. You can always improve it and go deeper as time goes on."

‍

‍

A real example: building a policy live

During the session, we built an AI policy in real time. Katherine Stanley Obando of El Colectivo 506, a bilingual solutions-journalism organization in Costa Rica, volunteered to work through the Trusting News worksheet.

‍

Her newsroom's situation is a good stand-in for a lot of small, mission-driven outlets:

• Where they use AI: translation and transcription, since they publish in English and Spanish, plus research and summarizing long solutions-journalism pieces into shorter formats.
• Where they draw hard lines: no AI in editing, no AI-generated photographic or published imagery, and crucially, no AI writing under a byline. Every story on the site is written by a human.
• What they can promise: stories are human-written, and any exception gets flagged right where it happens.

‍

Stanley Obando also described a twist on AI use. Her team built a chatbot that helps freelancers strengthen solutions-journalism pitches against four pillars: response, evidence, insights, and limitations. Rather than receiving obviously ChatGPT-drafted pitches, they channel that energy into a tool they control and trust. The bot pushes journalists to think harder, not to write for them.

‍

Will disclosing AI use cost you trust?

This is the question that makes editors nervous, and the research is honest about it.

‍

When newsrooms added in-story disclosures, some did see a slight drop in trust once readers learned AI was involved. But the size of that effect varied a lot by audience. International newsrooms, including those in Switzerland and Brazil, saw little or no decrease compared to US newsrooms.

‍

The conclusion from Trusting News is still firm: disclose anyway. The backlash from getting caught not disclosing is worse than the modest cost of being upfront. Know your audience, but don't let a small risk talk you out of the honest move.

‍

And if you've publicly said you don't use AI and that changes? Don't just quietly add a disclosure. Make a moment of it. A letter from the editor, a newsletter note, a short story explaining that you're now experimenting and why. Framing it as an ongoing, transparent process keeps trust intact.

‍

How to actually get started

‍

If you want to leave with something usable, here's the short version:

‍

• Reflect on what you're hearing from your audience, reporters, and social channels.
• Define why you're considering or using AI at all.
• Get specific about how you use it, when you won't, and what you promise.
• Write it for readers, not in journalism jargon.
• Make it easy to find. An existing About page is fine. Findable beats fancy.
• Add story-level disclosures that link back to the full policy.
• Revisit it at least annually, or every six months if you're using AI heavily. Keep the "how we use it" and "how we don't" sections current above all else.

‍

If your newsroom is forming an AI committee, build it from a mix of perspectives, people who use AI and people who don't, and make it clear how the rest of the staff can reach them.

‍

And if you’re not using AI at all, we’ve built a “No-AI Policy Builder” that lets you draft your policy in less than five minutes. 

‍

The bottom line

The pressure on local newsrooms isn't whether to talk about AI. Your audience is already having that conversation with or without you. The opportunity is to lead it.

‍

A short, honest, findable policy and a few clear disclosures do more for trust than perfect prose ever will. Start with the three questions. Get something live. Improve it as you go.

‍

Frequently asked questions

1. Do newsrooms really need an AI use policy? Yes. Surveys show 94 to 98% of news consumers want transparency about AI, and a majority now expect newsrooms to have clear ethical guidelines. Because audiences tend to assume journalists use AI regardless, a policy lets you correct that assumption with facts.
   ‍

2. What should an AI policy include? At minimum, answer three questions: whether and how you use AI, how you will never use it, and what you promise your audience. Readers especially want to know that a human reviews the work before it's published. You can also access the Trusting News AI policy worksheet.
   ‍
3. Should we publish a policy even if we don't use AI? Yes. Stating clearly that your content is human-produced shows you've considered the technology and made a deliberate choice. There's a strong audience appetite for "by humans, for humans," so you should get credit for it. Create yours in less than five minutes using Indiegraf’s No-AI policy builder. 
   ‍‍
4. Does disclosing AI use hurt trust? It can slightly, especially with US audiences, but research consistently finds that not disclosing is worse. The reputational damage from being caught using undisclosed AI outweighs the modest cost of being upfront.