Local news is disappearing fastest in mid-size cities. Would AI audio actually help?

Local news is not vanishing evenly. It is disappearing fastest in exactly the places that used to have enough of it to take for granted: mid-size cities and the counties around them, large enough to support a daily paper and a City Hall reporter for decades, not large enough to matter as a national media market once the advertising money that paid that reporter moved online. The scale of that collapse is documented in real detail. What is not documented, at least not yet, is whether any of the newer answers being proposed for it, including AI-generated audio, actually reach the people losing coverage, or whether that is still a hope sitting on top of the data rather than a tested response to it.

This piece starts with the numbers, because they are the part of the story that is settled. It ends with a question that is not settled: whether AI-personalized audio narrows the local news gap or simply narrates it from a distance.

The collapse, measured

Northwestern University's Local News Initiative has tracked this decline since 2005, and its October 2025 "State of Local News" report is the most detailed public accounting of it available. As of that report, 212 U.S. counties have no locally based news source at all, up from 206 the year before. Two decades earlier, in 2005, that number was around 150. Another 1,525 counties are down to exactly one remaining local outlet, usually a weekly newspaper covering a wide area on a small staff. Northwestern's tally puts the population affected by these gaps, counties with zero or only one local source, at roughly 50 million Americans.

The newspaper industry carries most of the loss. Close to 3,500 newspapers have stopped printing since 2005. In the single year covered by the 2025 report, 148 more closed or merged, a little over two a week. Fewer than 1,000 daily print papers remain nationwide. In 39 states, the total number of working journalists of any kind has fallen below 1,000. Public broadcasting has partly filled the space: the report counts 342 public media stations reaching into roughly 90 percent of U.S. counties, covering 82 percent of news desert counties, and standing as the only local news source at all in nine of them.

The report is not uniformly bleak. Northwestern also counts 695 stand-alone digital-native news outlets, up from 662 the year before, evidence that some of the reporting capacity lost from print is being replaced in a different format rather than simply disappearing. That growth is real, but it does not cancel out the county-level numbers above: new digital outlets concentrate disproportionately in already-served metro areas, and a handful of them opening nationally does not put a reporter back in a county that just lost its only paper.

None of this describes a niche problem. It describes what used to be an ordinary level of civic coverage in an average American county, now absent or thinned to a single outlet across a large and growing share of the country.

What it looks like in a mid-size city

"News desert" tends to conjure rural counties, and many of the worst cases are rural. But the erosion also runs through cities that are not small at all. Take Boise, Tulsa, and Chattanooga: places with six-figure populations, real local government, and the school boards, zoning disputes, and courts that used to get a beat reporter apiece. Cities that size generally still have at least one daily paper and a TV newsroom or two. What has usually gone missing is the mid-tier: the alt-weekly, the second paper, the wire-service stringer, the reporter who checked in on the county commission every week rather than only when something went wrong.

It is worth being precise about where audio already exists in these markets, because it complicates any simple claim that nobody is producing local audio there. Boise has an NPR member station, Boise State Public Radio, running news and talk programming across southern and central Idaho. Tulsa has Public Radio Tulsa, home to KWGS, on air since 1947 and still producing local news alongside NPR's national feed. Chattanooga has WUTC, operated by the University of Tennessee at Chattanooga, reaching 35 counties across the Tennessee Valley. All three are real, currently operating stations, not names kept alive only on a masthead. This is the general pattern across most mid-size metros: a public radio station, usually university- or member-supported, still producing some local coverage and increasingly offering it as on-demand audio rather than only a fixed broadcast schedule.

That existing coverage is real, but it is not the same as a city that size having what it used to have. One station covering a metro area is not a substitute for the several competing newsrooms that used to check each other's reporting. And "some local audio coverage already exists here" is a different claim from "the audio-shaped part of this gap has been identified and is being addressed," which is the claim this piece is actually trying to test.

The demand question the data does not answer

Here is the honest gap in this argument, and it is worth naming directly rather than gliding past it: there is no strong public evidence that people are actively searching for, or asking for, something like "local news podcast [city name]" at any meaningful scale. Northwestern's report documents the supply side of the crisis in real detail: outlets, counties, jobs, closures. It does not, and is not designed to, measure listener demand for an audio substitute.

The closest available data on the audio side comes from Pew Research Center, whose 2025 podcasts-and-news fact sheet found that 32 percent of U.S. adults say they get news from podcasts at least sometimes, up from 22 percent in 2020. That is a real and growing number. But Pew's fact sheet does not break that figure down by local versus national news, by market size, or by geography at all; it measures podcast news consumption in general, not anything specific to city or county coverage. The number that would actually answer "do people in news deserts want local audio" does not appear to exist yet in public research, at least not in a form this piece was able to locate and verify.

That absence matters more than it might seem. It is easy for anyone building or writing about audio products to assume that wherever journalism has a supply problem, audio has a demand-shaped solution waiting to meet it. That is a plausible hypothesis. It is not the same as a tested one, and treating it as settled would be exactly the kind of confident-sounding claim without a source behind it that this guide to verifying an AI-generated news summary warns against accepting at face value.

Would AI-personalized audio actually close the gap

Given that the supply crisis is real and the demand evidence is thin, the honest position is that nobody currently knows whether AI-personalized audio helps close the local news gap. The two most tempting answers, that more access obviously helps and that synthetic audio obviously is not real reporting, both skip past the actual mechanism in question.

The case for "maybe" is straightforward. Audio generation removes the cost of producing a spoken segment about a specific place or topic, a cost that used to require a studio, a producer, and an open broadcast slot. If a resident of a thinly covered county wants to follow city council decisions or a local ballot measure, personalized audio can turn a written record into something they will actually listen to on a commute, which is a real and different distribution problem than the one journalism has traditionally solved.

The case against is just as real. AI-personalized audio does not create reporting. It can narrate, summarize, or read aloud whatever source material already exists, and in a genuine news desert that source material is often thin or missing outright. If no reporter attended the county commission meeting, no amount of audio personalization produces an account of what happened there; the technology can only work with what somebody already wrote down. Earlier attempts to solve news fatigue and information overload with personalization technology have their own uneven record, documented in the graveyard of personalized news audio: good intentions and real engineering did not reliably translate into people getting more informed, let alone more locally informed, as a result.

Put together, the fair statement is that AI audio might make existing local coverage more accessible to people who would not otherwise read it, in places where that coverage still exists in some written or recorded form. It has no mechanism for producing new reporting where none exists. Whether that access effect is large enough to matter for the roughly 50 million Americans in a news desert or single-source county is an open, testable, and currently untested question, not a claim either side of this debate should make with confidence.

What would settle the question, in either direction, is data that does not appear to be published anywhere yet: whether people living in a news desert or single-source county actually use audio products, AI-personalized or otherwise, to follow local government and civic life, and whether doing so changes anything measurable about how informed or engaged they are. Search interest by market, podcast download data broken out by county, or a survey that asks news-desert residents directly what they already listen to would each be a start. Until something like that exists, the honest position holds — plausible, unproven, worth testing, not something to advertise as already solved.

What Lissin actually does here, and what it does not

It is worth being exact about where this leaves a product like Lissin, because the honest answer is narrower than "AI audio for local news."

Lissin turns a topic, question, mood, or source that a listener chooses into a personal audio show. If someone wants a daily update built around a specific local publication they already read and trust, a city paper's website, a regional outlet's feed, a local newsletter, they can point Lissin at that source themselves, the same way they would point it at any other topic or publication. That is a real capability today, and it depends entirely on the listener already knowing which local source to name and that source already existing and publishing.

That is a genuinely different thing from Lissin natively curating city-specific local news the way a wire service or a local newsroom does. Lissin does not currently maintain its own city-by-city local news coverage, assign anything resembling a beat, or promise that a listener in an under-covered county will get local reporting simply by opening the app. If the underlying local reporting for a place does not exist, personalized audio has nothing to draw from, a distinction worth keeping in mind whether you are evaluating this product or any other one making a similar-sounding claim. It is also the same distinction behind keeping up with the news without living in a feed: a briefing habit only works as well as the sources feeding it, and no amount of personalization changes what a source did or did not report. For now, the most honest use of AI audio in a thin local market is turning the reporting that survives into something a busy person will actually hear, not filling in the reporting that is already gone. Explore audio on Lissin if you want to build a show around local or national sources you already trust.

Sources

• The State of Local News 2025, Local News Initiative, Medill School of Journalism, Northwestern University.
• Podcasts and News Fact Sheet, Pew Research Center.