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.
I designed NewsGrade because the people who work in newsrooms, lead them, or support them deserved a fast, honest assessment of what their website actually does. Not what the CMS vendor claims. Not what feels true if you’re running it every day. What someone looking at the site from the outside would actually see, in the first ten seconds and the first ten pages. I liked the tool I built. I believed in it. But it wasn’t until I saw people besides me actually use it, watching them run their own sites through NewsGrade, reading reports on unfamiliar newsrooms, arguing over a couple of scores here and there, that I understood how good it could really be.
NewsGrade isn’t just a grading tool. It’s a benchmarking tool. And journalism, nonprofit journalism in particular, doesn’t get benchmarked very often. We get case studies. Annual surveys that ask newsrooms to self-report their metrics. Almost never do we get an outside look at how a whole sector’s websites are actually performing, measured the same way on the same day by someone whose paycheck doesn’t come from their side of the industry.
So I ran NewsGrade on roughly 400 nonprofit newsrooms across the country. Nonprofit news is having its collective existential conversation these days, and it's huge and sprawling and filled with good, important questions no one can answer. By my best estimate, there are only around 500 active nonprofit news orgs in the whole country, which means this isn't really a sample or subset of the field the way most statistical analysis is. If this were a normal survey of that same population, standard practice says you'd only need about 217 responses to call it statistically sound. I got almost double that. Run the actual numbers and testing 400 out of 500 gets you a margin of error of roughly ±2.2% at a 95% confidence level, tighter than just about anything else published about this sector, survey or otherwise. I hit basically everybody who has any kind of formal organizational presence. Lots of dollars moving through Press Forward right now. Lots of newsrooms launching every month. Hard questions about sustainability and reach coming from funders at a level we haven't really seen in years. Everyone has an opinion about the state of nonprofit journalism right now. Almost no one has data at this scope, compared this consistently.
Now we do. And the loudest thing this data says about the state of nonprofit journalism is something none of us needs me to tell: this sector has gotten really good at the technology problems of yesterday. Mobile-first website design. Site navigation. What happens when someone curls up your RSS feed. We mastered those problems years ago. That doesn’t mean our websites are perfect, but if there was ever a sector that was waving its arms and demanding “fund digital better!”, this was it. We got good at those problems, fast. Now we haven’t.
What NewsGrade Actually Tells You
Every newsroom has a website. Almost none of them know what that website looks like to the systems now deciding whether it gets found at all.
Yesterday’s problem, solved
Mobile score and site navigation both average about 8 out of 10 for newsrooms overall. That is really good. And it is consistent. If you’ve been running a nonprofit news site longer than six months, you know the awful luck leading Google and Facebook referral traffic numbers threw at us over the last decade. You built your site to work on phones and not frustrate your readers because of that pain. Almost every site got there. It shows in the scores. Getting here wasn’t special. It’s expected.
And there’s one last thing about these websites that really tells me how far we’ve come. It’s what separates merely functional newsrooms from the ones I actually click around on after they send me a link: it’s craft. Presentation of headlines and content. Typography. Article layout. Those are the scoring categories most strongly correlated to the overall score. High-scoring sites aren’t getting there because they cracked some coding_challenge.js that rearranges headlines based on viewport size. (Not that I’m opposed to that kind of thing, personally.) High-scoring sites are winning because someone cared how the actual words looked and how they read on a screen. Editorial skill and product design aren’t dollars. Not unless you’re paying someone who doesn’t actually care about your site that much. And if that’s the line separating “this newsletter is cool” from “I want to check out everything this newsletter links to,” it’s a gap we can close without founding an entire philanthropic program around it.
Step back again. Now we’re looking at an entire sector of websites that are almost nowhere near their best at distinguishing themselves and everything incredible they publish. Newsrooms across the country really got the basics down. Solidly. Consistently. Maybe too consistently. Few standout success stories or total failures. Just... not terrible.
Today’s problem: no one is winning
Which brings us to AI Readiness. That’s how prepared websites are for a world where not only social platforms and search engines but increasingly automated AI systems are going to control how new information reaches people who don’t know to come looking for it.
This isn’t sniffing around for vibes. Four concrete, objectively checkable things: does a newsroom’s robots.txt file properly manage AI web crawler access? Do they have an llms.txt file? Do they have an AI-related policy someone could actually find linked from their website? Do they disclose their own AI use on their site? Each one is simple, direct. Does the file exist? Does it do what it needs to do? Either yes or no. Opinion doesn’t come into these scores at all.
AI Readiness is NewsGrade’s lowest-scoring category by a country mile. Newsrooms average a 3.3 out of 10. The best performer in this entire group checked in at 6.7. Not one site hit 7.
Curious where your own site lands? Grade it free at NewsGrade — takes about ten minutes.
Break that average out into its four components, and it gets even uglier. Two of the four tests, AI crawler access management via robots.txt and AI use disclosure, had zero sites scoring “Strong” across the entire pool. Zero out of 403. Neither of those items are “do you have a policy?” one-offs. They’re checkboxes in a text file anyone can review online.
llms.txt? Eighty-three percent of nonprofit news websites don’t have that file at all. Half of us can’t point you to any written AI policy if you ask. Of the roughly 46% that do have something in place, only 1.5% do it well enough to score above “Weak.” Compare that to overall site grade if you want to see how badly nobody’s cracked this. Newsrooms with strong, accessible AI policies aren’t excelling elsewhere on the test. Website craftsmanship and clarity don’t suddenly get awesome when you tack on a well-placed link to “Our AI use policy.” If your site isn’t making your AI policy available to your readers, and most sites aren’t, you’re not just missing internet points. You’re invisible on this one particular metric to the AI systems that are rapidly becoming a first stop for people seeking information online.
Look at how that breaks down another way, and it only gets harder to find winners. AI Readiness scored almost entirely independent of everything else I looked at. It didn’t correlate with site quality overall. High-traffic, shiny new websites from major outlets scored about the same as local reporting hubs with five employees and a Kickstarter. It didn’t track with a newsroom’s reach, either. National outlets, state outlets, and hyperlocal shops all landed in the same narrow band. It didn’t track with beat, either. Health desks, climate desks, government reporters, arts coverage: didn’t matter. There’s no hint here that AI Readiness is a sustainable-journalism version of rich-versus-poor internet counties, a telling “if you have the money, you’ve solved this” kind of gap. There’s just... nobody’s solved it. Not really. Not yet.
That same “nobody’s cracked this” pattern shows up somewhere else, too, and it’s worth sitting with for a second. The Institute for Nonprofit News runs its own annual Index, a straight-up survey where newsrooms self-report what they’re doing. Totally different method than what I’m doing here. And their 2026 Index found the exact same shape: AI adoption inside newsrooms is fairly consistent across market size, geographic scope, and coverage priorities. No segment ahead of any other segment. Two completely different instruments, pointed at two different questions, landing in the same place. That’s not the kind of agreement you get by accident.
Here's where it gets interesting, though. That same Index says 81% of nonprofit newsrooms are using AI now, up from 63% in 2024 and 34% in 2023. Eighty-one percent. That's a strikingly fast adoption curve for a famously resource-strapped sector. Read that number on its own, and you'd think nonprofit news is moving fast on AI. Read it next to my numbers, and you get a very different, much better story: we've adopted AI everywhere except the one place it actually costs us something to ignore. We're using AI to summarize board meetings and crunch data like it's going out of style, and our own websites are sitting there almost entirely unprepared for that exact same technology to even find them, let alone cite or represent them fairly once it does. This isn't a sector that missed the AI moment. It's a sector that had the moment and pointed it at itself instead of out at the world.
None of this promises fair treatment once you're found, either. A newsroom could nail all four checks here and still get summarized without a link back, or worse. I shared an excerpt of this piece with INN before publishing, and Andrew Haeg, their product manager, raised exactly this point: readiness doesn't guarantee correct treatment, and INN's members have real, live concerns about AI systems summarizing their work without attribution, or taking it outright. That's a separate, unresolved fight, one INN's own product and standards teams are actively working through, and these four criteria don't touch it. Readiness, as NewsGrade measures it, is about whether you can be seen and understood by these systems at all. What happens after that is a different problem entirely.
Worth being precise here, because these two datasets aren’t just answering two different questions. They’re two different kinds of evidence. INN’s numbers come from a survey: newsrooms reporting what they believe they’re doing, or intend to do. NewsGrade’s numbers come from checking the actual files on a live site. Does robots.txt really manage AI crawler access? Does llms.txt exist? Is there a policy a reader could actually find? Does the site disclose its own AI use? Nobody’s asked their opinion. So when the Index finds 19% of outlets say they block AI scraping, and NewsGrade finds crawler management scoring weak across nearly the entire sector, that’s not necessarily a contradiction. It may just be the gap between a policy a newsroom believes it has in place and a robots.txt file that was never actually configured to do it. Self-report and technical reality can diverge quietly. This may be exactly where.
The Fog of Good Intentions
I've made this same "someone needs to just fix this" argument before, about a different problem. In this piece, I argued that the journalism support world doesn't actually have too many organizations, it has too little coordination between them, and that the fix looks less like a directory and more like a 911 dispatcher. Same instinct, different symptom.
We don’t know what comes next, but something’s coming
Which is exactly where the journalism industry was on mobile usability in 2011. Remember when mobile penetration hit about 35% nationwide, and every website started yelling about their responsive designs? That was when mobile-first actually became mobile-first. A problem big enough that the whole sector scrambled to fix itself at once. Something that massive doesn’t just happen.
We don’t know what happens when the vast majority of your potential audience can only find you through AI-powered discovery. Not yet. Maybe your continuing education team is already on it. Hell, maybe someone in your office is reading this newsletter right now and has the answer figured out. My guess is that’s unlikely, though. More likely: we don’t know what happens, but the ship is about to sail on this one regardless. Either we fix it ourselves, because someone will, and there will be massive incentives to leap cleanly from “existing websites suck at AI readiness” to setting the standard once we start paying attention to it, or it gets fixed for us. Mobile-first didn’t need drama to conquer the web. Its forcing event was a funding crunch and a referral traffic collapse that left newsrooms rebuilding websites across the country simultaneously. We don’t know exactly where AI and AI readiness are heading, but we know they’re here. My suspicion is that something’s coming for AI readiness that’s going to force us to pay attention to it. And again: whoever makes that leap first will have advantages that don’t go away. The question is whether we solve it before it solves us.
Bonus terrain: what this means if you support or fund nonprofits
Quick section here that’s less “what these numbers mean” and more “using these numbers to think about your own practice if you work at or on behalf of nonprofit news.”
“Digital capacity” is at least two different problems. Treating it as one is probably underserving everyone involved. AI readiness is not website craft. They’re not two halves of some cohesive whole that add up to “digital whatever.” One is an unsolved technical standard, industry-wide. The other is local editorial voice and skill, the thing differentiating between newsrooms that are fine and websites we actually want to click around on when we visit. Bundling both under “digital whatever” and treating them as commensurate misses the nuances of both.
AI readiness is, ideally, not a problem your organization should fund. It should solve it. AI readiness feels like a coordination problem more than a set of skills newsrooms lack. Almost every other technical dilemma facing small to mid-size newsrooms today isn’t that. If there’s a better way to write headlines that pops up, who does that training? Individual newsrooms. AI readiness isn’t something one newsroom nabs expertise on and then prices others out of for years. Your journalism school could teach every outlet in your region to fix this tomorrow if they wanted to. So could your local press club. Almost no other sustainable-journalism technique or trick is that easily mobilized to the entire sector’s benefit once you figure out the solution.
If most newsrooms look mostly competent in this space, that’s an opportunity for capacity-builders. Not newsrooms. Sorry, I digressed into this above, but if that little section clicked with you at all: “digital sustainability” is probably not a suite of problems your organization should be approaching by throwing dollars at newsrooms and hoping they find answers. Solutions to AI readiness go directly in journalists’ inboxes. So should programming that gives them something to do with that know-how. A full-stack competency like that doesn’t exist in small-to-midsize nonprofit news right now, because nobody outside of large organizations building custom systems has AI readiness nailed down either. There’s an opportunity here to help more than individual newsrooms, if your org’s looking for a place to focus on someone’s to-do list.
Information like this just doesn’t exist in journalism. Period. Journalism loves to say it wants evidence-based policy or wants to know what works. There’s never been much opportunity to get that data, except by individually pinging every other newsroom in the country about how they feel about X, Y, or Z. NewsGrade lets me collect that “how” data at scale. It’s strange that more funders and journalism orgs don’t want this same information about what’s actually on their grantees’ websites right now. Just how fast is AI spreading through newsrooms? How are our websites actually working?
Here’s the thing about this piece. I didn’t set out to write a love letter to NewsGrade. But it lets me say things like “here is what the median nonprofit news website does and doesn’t do when it comes to AI readiness, measured across 400-plus websites by the same criteria on the same day,” rather than “AI readiness is a problem I think exists and suspect we’re terrible at.” NewsGrade is how I know that. It’s easy to doubt one newsroom’s website diagnosis. It’s difficult to question hundreds of them. NewsGrade just ran those tests. And I’m going to keep running them.
That’s what the data tells you about where your site stands today. What it can’t tell you is why a sector that solved mobile-first so decisively still hasn’t touched AI readiness at all. That’s not a website problem. That’s an operations problem. More on that Thursday.
If you’re running a newsroom: grade your own site free at NewsGrade. Takes about ten minutes, comes back with a full report, not just a score.
If you’re on the funding or organizational side, and there’s any way you come away from this thinking “being able to answer questions like ‘where does my sector actually stand?’ would let me do my job better,” I want to hear how and where. Reply to this email or reach out directly.