We Built Infrastructure to Move Information Faster. None of It Moves Trust.

Last month, I wrote that NPR Digital Services didn’t fail. The system it was built to serve was never designed to trust it. The argument was about trust inside an ecosystem. Who believes whom, who gets the benefit of the doubt, and what happens when that account runs dry.

I’ve been watching a larger version of that problem build for years. Not trust inside one organization. Trust attached to a piece of information as it moves through the world.

I saw the early shape of it from inside media. The signals that used to tell people what was credible were weakening, the places those signals lived were losing their audience, and little was being built to replace them. I’m surfacing it now because AI has turned a slow erosion into an urgent one.

The clearest example I have is also the closest to home. My sister and I don’t trust the same sources, and after years of going in circles, I stopped believing the gap was about her judgment or mine. The gap is structural. When a claim reaches either of us, through a feed, a search, or an AI answer, almost everything that would let us weigh it has been left behind. We each get a confident answer with its credibility missing, and then we argue about why we believe different ones.

That’s not a family problem. It’s the same problem I watched take shape in media a decade ago, arriving now at everyone’s kitchen table at once.

This didn’t start with AI. AI just sped it up.

The erosion has been underway for thirty years. I watched a good part of it from inside the industry, and the shape was clear long before it reached most people’s feeds.

A few shifts did the damage:

• Publishing stopped requiring a gatekeeper. The repeal of the Fairness Doctrine in the 1980s. Blogging's arrival in the late 1990s, anyone could publish anything to a wide audience with no editor, no fact-check, and no standard to clear first. That democratized voice was good. It also severed the old link between “this is published” and “this passed through someone accountable,” which was costly.

• The business model collapsed. Advertising that once funded reporting moved to platforms. Local newsrooms thinned or closed, and with them went the layer of coverage closest to the communities they served.

• Distribution moved to algorithmic feeds. Social platforms became the front door to news, and those feeds were tuned for engagement, not accuracy. The content that travels best is rarely the content that’s most true.

• The line between reporting and opinion blurred. Audiences increasingly couldn’t tell the difference between a sourced account and a confident take, partly because the formats started to look identical.

• News institutions came under direct political pressure. The most recent and most pointed example is the current administration’s attacks on news organizations and public broadcasting, including moves against the funding that sustains them. That pressure is real, and it has done real damage to institutional credibility. It is not the origin of the trust problem. It is the accelerant landing on top of three decades of erosion that was already well underway.

Then AI arrived and accelerated all of it. Every weakness above gets faster and harder to see when an assistant hands you a finished answer and skips the page entirely.

So the question I keep coming back to isn’t who broke trust in news and information. Plenty of forces share that bill. The question is harder and more useful than blame:

How do we rebuild trust in news and information in a world where AI is the thing intermediating it?

A change in administration won’t undo this. The damage is done, and most of it predates any single presidency. What’s left is to adapt to the landscape we actually have, and to help the news organizations and creators who want to be credible prove it in a form that survives the trip to the reader.

AI didn’t fix this. It moved the gap somewhere harder to see.

The common assumption is that AI assistants solved the sourcing problem. You ask a question, you get an answer, and the citations sit right there. It shows its work.

But showing a source isn’t the same as standing behind one. When an assistant pulls ten links to build an answer, a newsroom with a corrections policy and a content farm built to chase clicks arrive formatted the same way. Same little number. Same visual weight. Nothing in that footnote tells you who is accountable for the claim, whether it was ever corrected, or whether it earned its place on the merits or just ranked well, which is not the same thing.

To be fair, these systems aren't blind to quality. Some weigh the reputation of a domain. Some have arrangements with specific publishers. Developers can build guardrails that lean toward more reliable sources. But whatever weighting happens is inconsistent from one product to the next, invisible to the person reading the answer, and never shown in the answer itself. You can't see whether it happened, which means you can't rely on it.

The link is visible. The thing that would make the link mean something is not.

And here’s the shift that should concern anyone who cares about credible information: the destination is disappearing. People used to land on a website, where they could see the masthead, the byline, and the reputation on the line. Now the answer comes to them, and the page never loads. The one place trust signals used to live is the place fewer and fewer people visit.

What would actually have to travel

If trust is going to survive the trip from where content is made to where it’s consumed, a handful of things have to travel with it. Not opinions. Facts about the content that a machine and a person can both read.

• Where it came from

• Who is accountable for publishing it

• Whether it has been corrected, and when

• What evidence sits underneath the claim

• How it was produced, including what role AI played in making it

None of that is exotic. A serious newsroom already tracks most of it internally. The problem is that this information dies at the edge of the organization that holds it. It doesn’t get attached to the content in a form that anything downstream, a search engine, an assistant, or a reader can pick up and use.

I want to be careful here, because the next question is the one I’m not ready to answer in public yet. How exactly do you weigh those signals against each other? What counts as enough accountability? That’s the part that turns a concept into a system, and it’s the part that deserves more rigor than a confident paragraph. I’d rather name the signals honestly than pretend I’ve already built the scale.

This isn’t a brand-new idea, and pretending it is would be a mistake

Provenance standards already exist. The Coalition for Content Provenance and Authenticity has spent years building a technical standard for attaching origin data to media, and the Content Authenticity Initiative has been pushing adoption across major technology and media companies. The plumbing for “where did this come from” is further along than most people realize.

What’s missing isn’t the standard. It’s the operator. Someone has to run the layer that decides which signals matter, maintains the accountability records, keeps the corrections honest, and stays trusted enough that both the public and the AI systems treat its signals as meaningful. A standard tells you data can travel. It doesn’t tell you whom to trust to manage it.

That operator role is the open question. And it’s where one sector has an advantage almost nobody else does.

Why public media is the natural home for this

Public media has spent decades investing in exactly the things this layer requires. Editorial standards. Corrections practices. Accountability to a community rather than to advertisers. Governance that already assumes someone has to answer for what gets published.

Most organizations would have to build that credibility from scratch to operate a trust layer. Public media already paid for it. The asset is sitting there, underused, at precisely the moment the information ecosystem needs a neutral party that isn’t trying to win the argument, only to make the argument legible.

Localism has long been called public media’s moat. I’d add something to that. A moat you can’t see doesn’t protect anything. As content gets summarized, stripped, and answered before anyone reaches the source, the credibility that public media built has to become visible in the places people actually encounter information. Otherwise, it protects a destination no one is visiting.

I think public media is positioned to operate that layer better than anyone else in the ecosystem. That part I’ll say plainly.

• How it gets built is where the honesty has to hold

• Which signals matter most to real audiences 

• How governance works without sliding into deciding what people should believe 

• Who pays for it, and what stays open infrastructure versus what becomes a service

Those are genuine open questions, and I’m having them right now with people across journalism, technology, and public media.

The vision underneath this is simple, even if the build is not. We spent thirty years getting content to move faster. Not one of those advances did anything for the trust that’s supposed to move with it.

That’s the gap. I think public media is the one positioned to close it. The rest is a conversation I’d like to have in the open.

---

Do you have thoughts to share on this topic? Join the conversation!

The Connetion Lab meets every Wednesday at noon for entrepreneurs, leaders, and operators who want a place to think through real challenges, gain perspective, and leave with clearer next steps.