How ARLnow's AI Strategy Evolved From Automation to Augmentation
Sometimes there is lot to be learned from a small and contained local experiment about what works and what doesn't (yet?) with Gen AI in the newsroom
Left: A prompt on the ARLnow website to subscribe to the Premium-Newsletter ProUpdate. Right: Owner-operator Scott Brodbeck
In January 2025, ARLnow launched ProUpdate, an AI-powered newsletter with automated summaries of local news. The new membership product comes with human oversight and careful quality control. And with lessons learned from one of local journalism's most interesting pioneering Gen AI experiments.
ARLnow is the flagship site of Local News Now (LNN), a digital local news company, founded in 2010. It now reaches 2.4 million monthly readers viewing about 85 million pages per year across three Northern Virginia sites: Arlington (ARLnow), Alexandria (ALXnow) and Fairfax County (FFXnow). LNN’s measured approach to AI reflects hard-won experience about what works and what doesn't in automated journalism.
The Current AI Strategy
The main site ARLnow uses artificial intelligence across a dozen different functions. The AI toolkit includes interview transcription, grammar checking upon publication, automated event calendar evaluation, social media optimization, weather forecast writing based on National Weather Service data, emoji selection for posts, and article summaries for internal use.
ARLnow's newest product, the Pro Update membership newsletter was launched in January 2025. Using OpenAI's GPT-4o model, it generates both text and audio summaries of ARLnow's original reporting for paying subscribers. "The Pro Update newsletter has no display ads, larger photos and short AI-generated article summaries, so members can quickly catch up on the day's news without clicking into every story," the owner and operator Scott Brodbeck told me yesterday via email. The performance metrics are impressive: 75-80% open rates and 10-30% click rates, outperforming their standard daily newsletter by 20-50% in both categories.
Press Club is the newer paid membership tier of ARLnow. Revenue still mainly comes from advertising
Complete transparency is central to the approach. "We disclose that the summaries are generated with AI," Brodbeck explained. This frees up reporters to focus on news gathering rather than writing multiple versions of their own articles.
Additional planned AI applications include automated creation of weekly "things to do" posts based on event calendar data and optimizing sponsored content performance for local advertisers. The company has also developed a content classification system that can categorize articles with significant accuracy, though this remains primarily for internal organization.
What ARLnow explicitly avoids is telling. They tested AI tools for writing news articles based on press releases but found none produced publication-quality work. "Humans, frankly, are better and more interesting writers, who can add local context to stories in a way that AI struggles with,” Brodbeck pointed out in an AI policy post last year. The company also avoids using AI for email communications, preferring "a more personal touch."
Looking ahead, Brodbeck sees AI's potential expanding beyond content. "The thing I'm really excited about now is AI code generation and vibecoding — that's going to allow us to build a bunch of cool stuff for readers, ad clients and our own that was not financially or technically feasible before," he told me.
This measured approach partially reflects lessons learned from ARLnow's position as one of the early pioneers in fully automated AI newsletter production—an experiment that began with very specific constraints, initial hopes, but ultimately had to be abandoned.
The Business Foundation
To understand ARLnow's AI journey, it helps to understand the business constraints that have consistently shaped the company. Local News Now operates as a for-profit company in a challenging market, with memberships representing "less than 10% of our revenue, maybe even less than 5%" according to founder Scott Brodbeck in a November 2024 podcast interview. Instead, the company has achieved sustainability through advertising diversification—"no one advertiser has more than 1% of our revenue," allowing editorial independence.
This resource-constrained reality explains why AI automation seemed so appealing in 2023 when Brodbeck was running three sites with just ten employees. The company has since expanded by acquiring the GazetteLeader in September 2024, adding two journalists to a newsroom of now 9, but as Brodbeck noted in November 2024, "we're right about at our monthly average revenue in terms of our expenses."
The Automated Newsletter Experiment
In April 2023, three sites ARLnow, ALXnow and FFXnow already sent automated afternoon newsletters to a combined 36,000 subscribers, but Brodbeck wanted to add a morning edition to ARLnow with more personality. But he didn’t have enough staff to produce it.
His solution: a completely automated morning newsletter powered by ChatGPT. From concept to launch took just two weeks, with Brodbeck dedicating only a few hours daily to the project. The workflow: GPT-4 APIs integrated with Zapier, Airtable, and RSS feeds would identify the previous day's best-performing articles via Google Analytics, then generate three to four summaries with an introduction. GPT-4 even programmed the CSS formatting.
Brodbeck was upfront about limitations from the start, telling Nieman Lab at the time that a human journalist "would absolutely outperform the AI model in its current iteration." But for a small newsroom, the question was whether readers would find it "good enough for their local reading needs."
The newsletter was clearly labeled as experimental and required active opt-in. A few months after start, about 150 people had subscribed, with open rates between 65-70% - higher than the afternoon edition. Click rates were only 25% lower despite providing full summaries in the newsletter.
But the AI struggled with time context, writing about old stories as current news. The tone was frequently "cringe-y or corny," and there were occasional factual inaccuracies.
The AI experiment coincided with another controversy. In early 2024, the site faced criticism on Instagram for using AI-generated images to illustrate real estate and business stories. When ARLnow surveyed readers, 48% said they opposed AI images "under any circumstances."
ARLnow responded by suspending AI-generated images and commissioned human-created illustrations instead.
By June 2024, despite solid engagement metrics, subscriber growth of the automated newsletter remained minimal even with active promotion, and the system still required monitoring, maintenance, and API fees.
By June 2024, ARLnow announced its conclusion that "most readers are more interested in simple headlines from stories that happened today rather than voice-y summaries of things that happened yesterday."
Strategic Pivot to AI for Amplification
The failed newsletter experiment taught ARLnow that AI works best as amplification, not replacement. As Brodbeck told me yesterday, reflecting on the experiment: "It worked well from a technical standpoint but it just didn't have much user adoption as a standalone AI newsletter."
I asked Brodbeck about the lessons learned and he said: "AI helps us do some things, like article summaries and approval of events submitted to our event calendar, much more efficiently. These process improvements allow our staff to focus more of their attention on their core responsibilities, like news reporting or advertising sales."
His advice for other small publishers is direct: "Use generative AI primarily for summarization — it does much better with that than trying to generate complete articles."
The Broader Context
ARLnow's current use of AI for augmentation reflects a broader pattern in local media AI adoption: more intenal experimentation, more caution on the audience-facing side. The Minnesota Star Tribune uses AI to process decades of archives but maintains strict editorial control. The Baltimore Banner developed an AI taxonomy system that achieved 84% accuracy in content classification, but uses it for internal organization rather than public-facing content. Village Media in Canada operates AI-powered community engagement tools across 34 local sites, but focuses on engagement features rather than content generation.
Technology as Competitive Advantage
Interestingly, ARLnow has found success in leveraging its AI and technology expertise as a business line. The company provides technical and business support to three of its partner sites in the DC area - PoPville, MoCoShow, and Potomac Local - each handling 1-2 million page views per month.
"While local news doesn't scale from an editorial standpoint and sometimes a business standpoint, the technology side certainly scales," Brodbeck explained in November 2024. This technology licensing model provides additional revenue while maintaining editorial independence—the partnerships are "strictly on the business and technology side" with no editorial overlap.
Why is the ARLnow case study relevant?
The learnings from ARLnow's pioneering but ultimately failed automation experiment and their current successful AI integration offers a roadmap for other local news organizations:
Communicate transparently: ARLnow clearly labeled the newsletter as experimental and required opt-in rather than automatic enrollment. This transparency helped manage expectations while gathering genuine feedback.
Start small and contained: The experiment was limited to one additional newsletter, allowing for controlled testing without risking the main business.
Plan for success and failure: Brodbeck had expansion plans if the newsletter succeeded (rolling out to other sites, adding sponsorship) but also clear criteria for when to shut it down.
Technology as a business model: If in-house developed AI tools prove their value in your own local news organization they could be useful for other local news.
Most importantly: resist the temptation to use AI as a staffing solution. ARLnow's experience suggests that AI works best when it frees up human journalists to do more journalism, not when it tries to replace journalism itself.
The technology can accelerate workflows and handle routine tasks, but it can't replicate the local knowledge and editorial judgment that readers expect from their community news source. For ARLnow, that realization has led to a more sustainable and effective AI strategy - one that amplifies human capabilities.
Tools used for this newsletter:
Otter.ai: Podcast transcription
Claude 4.0 Sonnet Pro: Research and first draft