A New York Times/Siena poll released this week found that 35 percent of Americans think artificial intelligence is “mostly bad.” Only 16 percent think it’s “mostly good.”

Sit with that for a second. The technology that every media company is rushing to implement — the thing your leadership is mandating, your vendors are pitching and your competitors are announcing — is viewed unfavorably by more than a third of the country.

When the Hard Fork hosts put that number to Google CEO Sundar Pichai this week, he didn’t spin it. He said people’s anxiety is legitimate, that humans aren’t built to process change at this pace, and that the industry has more work to do.

He’s right. But he’s also the CEO of a company that sells AI products. His problem is winning back public trust. Your problem, if you run a newsroom, is more complicated than that.

News organizations are caught in a specific bind that Pichai doesn’t have to navigate.

You are simultaneously trying to cover AI fairly — to report on job displacement, on bias, on regulatory battles, on the legitimate concerns that 35 percent of your audience holds — while also implementing the technology inside your own walls. You are asking your staff to use tools that many of them distrust, to produce work for an audience that’s increasingly skeptical of it, and to do it without losing the credibility that makes your journalism worth anything in the first place.

That is a tension most AI coverage glosses over. The trade publications write about the tools. The think pieces write about the ethics. Almost nobody writes about what it actually feels like to be a news director in 2026 trying to thread that needle every day.

I’ve been in those rooms. Here’s what I’ve seen work and what hasn’t.

The newsrooms that are getting this wrong treat AI implementation as a technology project. They bring in a vendor, run a pilot, announce the tool in a staff meeting and wonder why adoption is slow and morale is worse. They skip the conversation that actually needs to happen first: what are we afraid of, and what are we trying to protect?

The newsrooms getting it right treat it as a journalism question. What does this tool let us do that serves our audience better? What does it not touch — and why? Where is the line between AI-assisted work and AI-generated work, and how do we explain that line to our audience clearly?

The second approach takes longer up front. It involves harder conversations. But it produces something the first approach never does: staff who are invested in the outcome rather than waiting to be replaced by it.

Pichai used an example in the interview that stuck with me. He described asking an AI agent to look ahead at his calendar and color-code meetings by category. It took seconds. It probably would have taken him or an assistant 20 minutes to do manually.

That is the version of AI that most newsroom staff can accept — a tool that handles the administrative noise so they can do more actual journalism. The version they can’t accept, and won’t, is one that feels like it’s being used to justify cutting the people doing the journalism.

The distinction isn’t about the technology. It’s about the intent, and whether leadership communicates it clearly.

The 35 percent number isn’t going to fix itself. Public trust in AI will shift based on whether the technology demonstrably improves people’s lives — and whether institutions like news organizations handle it in ways their audiences can respect.

That means being transparent about where and how you’re using it. It means not hiding AI-assisted work inside a byline that implies otherwise. It means giving your staff a voice in how it gets implemented, not just a training session after the decision is already made.

Pichai said anxiety is healthy; in a democracy, you want citizens engaged and making their preferences known. That’s true. It’s also true inside your newsroom. The journalists who are worried about what AI means for their jobs — they’re not wrong to be worried. The news directors who listen to that and respond thoughtfully will build something durable. The ones who don’t will have a different set of problems in about 18 months.