Fighting the notification deluge
Why the future of journalism belongs to minimal, trusted digital utilities—not the endless scroll.
A Quick Disclaimer: Backstory & Strategy is a personal, independent publication. The views, analysis, and commentary expressed here are strictly my own and do not represent the official position, strategy, or endorsement of the American Press Institute, its leadership, or its board. This is my personal space for analyzing the media landscape, testing new frameworks, and thinking out loud.
There is no digital channel with a higher open rate than SMS. Not email. Not push notifications. Not anything that depends on an algorithm deciding whether your content deserves to be seen. For most of the last decade, an estimated 98 percent of text messages sent to American adults were opened, most within three minutes of delivery. Publishers looked at that number and drew what seemed like an obvious conclusion: SMS was the most powerful direct-to-reader channel in existence, and journalism should be using it more.
They were right about the number. They were wrong about what it meant.
The trust SMS carried wasn’t journalism’s trust. It was the medium’s trust, built by decades of carrier compliance regimes, consumer habits, and the simple fact that for most people, a text message means someone who knows them is trying to reach them. Journalism didn’t earn that trust. It inherited it. And it never asked what would happen if someone spent it.
The 2024 election cycle answered that question.
What Campaigns Did
In 2022, Americans received 15 billion political texts. That represented a 158 percent increase from the year before, while political phone calls dropped by more than half. Campaigns had figured out that the marginal cost of a text was nearly zero, that phone numbers were freely available through voter registration records treated as public information, and that the regulatory framework governing SMS made their conduct not just legal but essentially unrestricted.
That last part is the piece most coverage missed. Traditional businesses are bound by the Telephone Consumer Protection Act, which requires affirmative consent before you can send a commercial text to someone. Political organizations texting from personal phones are largely exempt from those requirements. The consent architecture that every legitimate commercial sender had to build and maintain simply did not apply to campaigns. They could acquire your number without your knowledge, text you without your permission, and keep texting from new numbers when you blocked the old ones.
The volume in 2024 made 2022 look restrained. Security researchers documented a threefold jump in political spam volume — consumer-reported unwanted political messages — compared to the same period in the prior midterm cycle. A fifteen-year-old in Boston who had never voted told a local news station that her phone was “blowing up” with requests for donations and turnout. Campaigns were texting people who had never interacted with them, from numbers those people did not recognize, at frequencies that trained an entire generation of phone users to treat an incoming text from an unknown number as a threat to be dismissed or blocked.
The damage to the channel was not incidental. It was structural.
The Collateral Damage Problem
Campaigns got what they wanted. Fundraising conversion rates on SMS outperformed email. Volunteer recruitment worked. Voter contact moved numbers in close races. The tool worked for the people using it, in the narrow terms they were measuring.
Everyone else absorbed the cost.
When a local journalist sends a breaking news alert to subscribers who opted in specifically to receive it, those subscribers now arrive at the alert primed by months of conditioning to treat unexpected texts as something to ignore. The opt-in relationship and the campaign blast look identical at the moment of delivery. Both arrive as an SMS from a number the recipient may or may not have saved. The reader’s pattern recognition does not distinguish between them. It fires the same dismissal reflex it has learned to apply to everything coming through that channel.
This is the collateral damage problem. A nonprofit newsroom covering a city council meeting cannot exempt itself from the psychological environment campaigns created. A public radio station sending a membership reminder cannot buy its way out of a conditioned avoidance response. The degradation is ambient. It affects every sender, regardless of how responsibly they built their list.
Cybersecurity researchers have begun documenting the downstream effect directly, noting that excessive political messaging and the attendant rise in SMS-based impersonation and fraud are putting users’ longstanding trust in mobile messaging at risk. The warning is not hypothetical. The erosion is accelerating.
How the Field Responded
Journalism’s response to the degraded SMS environment has been, as it often is with channel disruption, fragmented and mostly reactive.
Some publishers retreated to browser-based push notifications, which have their own collapsing engagement problem and none of the intimacy that made SMS valuable. Some moved to WhatsApp Channels, which launched in September 2023and attracted dozens of publishers almost immediately. WhatsApp offered the feel of messaging without the carrier environment, no algorithm between publisher and reader, and a chronological feed that put content in front of followers without requiring them to seek it out. The New York Times, Vox, and major international publishers were early adopters.
The WhatsApp move solved the algorithm problem. It did not solve the platform dependency problem. WhatsApp is a Meta product. Publishers who build audiences there do not own the data, cannot port their subscriber relationships to another system, and remain subject to whatever decisions Meta makes about the product. It is a more intimate version of the same bargain journalism has been making with social platforms for twenty years. The terms look better today than they probably will in five years.
A smaller group of publishers went a different direction. They found purpose-built tools that treated SMS not as a broadcast channel but as a consent-based relationship, built strict opt-in architecture from the ground up, and made two-way conversation a feature rather than an afterthought. The most established of those tools in the journalism space is Subtext, a platform built specifically for media companies, journalists, and creators that has accumulated a publisher roster running from the Washington Post and Axios to Morning Brew and PBS. The Washington Post’s Texts with Tumulty program, in which then-opinion columnist Karen Tumulty sent direct updates to her most engaged readers, is among the cleaner examples of what journalist-driven SMS looks like when the consent model is working. Subtext also partnered with News Revenue Hub in 2024 to bring the platform specifically to nonprofit newsrooms, which is the corner of the market that has the most to gain from a high-trust direct channel and the least margin for wasted outreach.
The results, for publishers who committed to the approach, were measurable. Reply rates and click-through rates on permission-based SMS programs consistently outperformed comparable email programs. Readers who actively chose to receive texts from a specific journalist or outlet engaged with those texts at rates that would be recognizable to anyone who remembers what email felt like before the inbox became a landfill.
The proof of concept exists. The consent model works. The problem is that it never reached the infrastructure layer where most journalism actually happens.
The CMS Gap Nobody Fixed
Here is the most revealing thing about journalism’s relationship with SMS: not a single major newsroom content management system has integrated it in any meaningful way.
Not WordPress, which powers more than 40 percent of all websites and has an ecosystem of 60,000 plugins. Not Arc XP, the enterprise CMS that the Washington Post built for itself and licensed to publishers nationwide. Not Newspack, the purpose-built platform funded by Google, Knight, Lenfest, and ConsenSys specifically for independent and nonprofit newsrooms, which integrates Mailchimp, ActiveCampaign, and three other email and CRM tools natively but has no SMS layer at all. Not Ghost, not Drupal, not any of the major platforms purpose-built for journalism.
WordPress plugins exist that can send SMS notifications. They were built for e-commerce. Their documented use cases are order confirmations, abandoned cart recovery, appointment reminders, and WooCommerce shipping updates. A newsroom could theoretically wire one of them up to send breaking news alerts, but there is no editorial workflow built around it, no story-level trigger, no subscriber segmentation by beat or geography, no two-way conversation architecture. It is a retail tool. Bolting it onto a journalism workflow requires technical resources most newsrooms do not have, and produces a result that does not serve readers well.
The publishers doing SMS well are running two entirely separate systems. The CMS handles content creation and publication. A dedicated SMS platform handles the reader relationship in that channel. Every send requires a journalist to log into a second tool, manually compose a message that mirrors or supplements what they just published in their primary workflow, and hope the two systems stay synchronized. The Washington Post, which built Arc XP and is among the most technically sophisticated newsrooms in the world, runs its journalist-driven text programs on Subtext, entirely outside its own CMS. The separation is the norm, not an exception for under-resourced organizations.
This is not primarily a technical failure. It is a priority failure. CMS platforms have historically understood their job as content creation and distribution, leaving audience relationship channels to the marketing stack. SMS sits in that gap. The result is that every newsroom that wants to use the channel responsibly has to build its own integration, maintain it indefinitely, and train journalists to work across tools in ways that add friction to an already demanding workflow.
The channel that campaigns exploited required no integration, no training, and no respect for the reader on the other end. The channel journalism needs requires all three, and the infrastructure that would make it manageable at scale does not yet exist inside the tools journalists actually use every day.
The Consent Architecture Is the Whole Argument
There is a version of the SMS conversation that treats this as a gear problem, a question of which platform to use and whether the economics pencil out. That version misses the point.
The publishers demonstrating that SMS still works as a journalism channel have one thing in common. They treated permission as the product. They built subscriber relationships through active opt-in, communicated clearly about what readers would receive and how often, and respected the intimacy of the channel rather than exploiting it. The open rates and engagement metrics that followed were not a function of SMS being a powerful broadcast medium. They were a function of readers who had specifically asked to be reached choosing to engage.
This is the reframe the field needs. SMS is not a broadcast channel that happens to have high open rates. It is a high-trust channel that produces high open rates specifically because it is not treated as a broadcast medium. The campaigns that spent down that trust were doing it precisely because they were treating it as broadcast. The publishers getting results are doing it because they are not.
I ran a text program at Lehigh Valley Public Media before I had a name for this argument, and the lesson I took from it was the same one any publisher running a serious SMS program eventually learns: the constraint is the feature. The discipline of earning each subscriber, of sending only what you promised, of respecting the channel enough to use it sparingly, is not a limitation on the tool’s power. It is the source of the tool’s power.
The regulatory asymmetry that let campaigns bypass consent requirements is a policy problem and a carrier enforcement problem, not something journalism can fix unilaterally. The CMS gap is a product investment problem, and the Knight Foundation, the Lenfest Institute, and the Google News Initiative — the funders who underwrote Newspack’s development — are the most logical place to push on it. They have already made the argument that journalism infrastructure is worth subsidizing, and SMS integration is an infrastructure problem. The operational and regulatory complexity that makes it hard to build is exactly why market incentives haven’t solved it, and exactly why it belongs in the same philanthropic infrastructure conversation that produced Newspack in the first place. What journalism can control right now is the consent architecture it builds. The publishers who build it correctly are proving, with actual data, that the channel is not broken. It was never broken. It was always possible to get this right. It just required treating readers as people who had given you something valuable when they handed over their phone number, rather than as targets whose coordinates happened to be available.
That distinction sounds obvious. In practice, it has been the difference between the publishers who are doing SMS well and the ones who decided the channel was compromised and moved on.
The channel is not compromised. The field’s relationship to permission is.
Disclosure: The American Press Institute has a professional relationship with Subtext. I used Subtext’s platform during my time at Lehigh Valley Public Media and have professional friendships with members of their team. Subtext appears in this piece as an example of consent-based SMS architecture — one of several approaches worth examining — and nothing here constitutes an endorsement of any specific platform or vendor.
If this piece made you think differently about something, the best thing you can do is share it. Forward it to a colleague, drop it in a Slack channel, or send it to someone building audience infrastructure who needs the frame. These arguments only travel if readers carry them.
If you want to push back, add context, or tell me about work I should know about — hit reply. I read everything. If you’re running an SMS program, thinking about the consent architecture problem from a different angle, or if I simply got something wrong, I want to hear it. Corrections that hold up will be noted in the piece.
If someone forwarded this to you and you want it in your inbox, you can subscribe here: