National Today ran a section that looked, at a glance, like local journalism: bylined stories about city council votes, school openings, and small-town events in markets across the country. SEO consultant Glenn Gabe, who has spent years documenting what he calls "Mt. AI," the rapid rise and inevitable collapse of sites that scale on AI-generated content, picked this one apart in a case study once the section had already crumbled.
The site that scaled past the point of no return
The /us/ directory of National Today's site had grown to more than 850,000 indexed URLs, and per Gabe's analysis, the section was 100% AI-generated. That's not a site that dabbled in AI assistance for a handful of posts. It's a purpose-built content operation, scaled to a size no human editorial team produces organically, covering hyperlocal news in enough markets to plausibly claim national reach.
Eight hundred fifty thousand URLs is worth sitting with for a second. A well-staffed digital newsroom covering a single metro area publishes a few thousand stories a year. National Today's local section was operating at a scale that would require thousands of full-time reporters working simultaneously across every market it claimed to cover, produced instead by a content pipeline with no reporters in it at all.
Investigative outlet Futurism dug into how the section actually got built, and found it wasn't just AI-generated, it was AI-generated from other people's reporting. The investigation describes the operation lifting original stories from outlets ranging from major newspapers to small local newsrooms, rewriting them without credit or a link back to the source, and publishing the result under invented bylines. When Futurism tried to count how many articles the section published in a single day, they lost count around 300.
What the investigation found
That distinction, AI-generated versus AI-generated-from-stolen-reporting, matters more than it might seem. A site publishing AI summaries of public information is operating in a gray area plenty of publishers occupy today. A site publishing AI rewrites of a competitor's original journalism, without attribution, at a scale of hundreds of pieces a day, is running a plagiarism operation with an AI content pipeline attached. Futurism's reporting treated it as the latter, and named the specific publications whose work had been lifted without credit.
The bylines are what make this case land differently than a typical thin-content penalty. A generic AI content farm at least has the decency, if that's the word, of not pretending a human reporter drove out to a school board meeting. National Today's section did exactly that: invented names, invented datelines, attached to stories that were themselves lifted from someone else's actual reporting. That's a layer of deception on top of the content-quality problem, and it's the layer that made the story worth a national investigative piece rather than just another SEO forum thread about thin content.
It's also worth asking why this particular pipeline scaled as far as it did before anyone caught it. Local news is a category with unusually low competitive scrutiny; most towns don't have more than one or two outlets covering them closely enough to notice a plagiarized rewrite of their own reporting showing up elsewhere, styled as an independent local outlet with national reach. That's precisely the gap an operation like this is built to exploit: enough real-sounding specificity to pass a casual read, spread across enough disconnected small markets that no single victim publication has the scale to notice the pattern on its own. It took a national investigative outlet connecting the dots across markets to see what individual local newsrooms couldn't.
“"Do not implement risky and spammy tactics just to rank in AI search." (Glenn Gabe, When 'Mt. AI' crumbles, ChatGPT can follow)”
The manual action
Google applied a manual action for "Scaled content abuse," its specific policy against publishing large volumes of unoriginal content with little value to searchers, and the /us/ section came out of the index relatively quickly after Futurism's article ran. That timing is itself instructive: this wasn't a broad algorithmic update sweeping through months later. It reads as a manual action triggered, at least in sequence, by public exposure. A broad spam update rolls out on its own schedule and catches sites algorithmically. A manual action is a human reviewer looking at a specific site, and public reporting is one of the more reliable ways to get one looked at.
Manual actions for scaled content abuse aren't new as a policy. What's changed is the scale the policy is now being tested against. Google's spam guidelines were written with content farms of a few thousand thin pages in mind. An operation running past 850,000 AI-generated URLs in a single directory is testing whether the enforcement mechanism built for the old scale still works at the new one, and in this case, once attention landed on it, it did.
| BROAD ALGORITHMIC SPAM UPDATE | MANUAL ACTION | |
|---|---|---|
| Trigger | Scheduled rollout, applied systematically | A reviewer looking at a specific site |
| Timing | Weeks to roll out fully across the index | Can land within days once flagged |
| What tends to draw it | Patterns detectable at scale across many sites | A single site's behavior, often after public exposure |
| Recovery path | Wait for the next update cycle, fix and hope | Reconsideration request after removing the violation |
That distinction is why the sequence in this case matters as much as the outcome. Nobody has confirmed the exact mechanism Google used to flag National Today's directory, but a public investigation naming the site, quantifying the scale, and detailing the plagiarism made it trivially easy for a human reviewer to find, evaluate, and act on. Sites relying on "no one's looking closely enough to catch this" as their actual risk model are betting against exactly the kind of scrutiny that took this one down.
It didn't stop at Google
The part of Gabe's case study that should worry anyone treating AI search as a separate acquisition channel from Google: National Today's citations in ChatGPT largely disappeared alongside its Google visibility, with only a few scattered mentions surviving. That's consistent with what we've found tracking how ChatGPT's citation sources shift: AI engines lean heavily on the same crawled, indexed web that Google ranks, and a chunk of that retrieval pipeline runs through infrastructure Google itself operates or influences. A site deindexed from Google doesn't automatically vanish from every AI engine's training or retrieval data, but the citation collapse here suggests the overlap is large enough that a Google-side penalty carries real cost on AI-answer visibility too.
That's the part of this story that should change how teams think about channel risk. A common mental model treats Google rankings and AI-engine citations as two separate bets, worth hedging independently, so that a problem in one doesn't necessarily sink the other. National Today's collapse argues the two are more correlated than that model assumes, at least when the underlying cause is a content-quality violation severe enough to trigger a manual action. The infrastructure AI engines lean on for retrieval overlaps enough with Google's own index that a penalty on one side doesn't stay contained to it.
The pattern behind it
This isn't the first site to walk this path and it won't be the last. Gabe has flagged the broader trend repeatedly: sites racing to publish "commodity content" generated at scale to catch every possible AI-search prompt, treating volume as a strategy in itself. He's been explicit that manual actions, not yet a fully automated detection system, are doing most of the enforcement work right now, which means the sites getting caught are disproportionately the ones that get noticed, by a competitor, a journalist, or a reviewer who happens to spot-check the right directory. That's a rougher, less predictable enforcement regime than an algorithmic update, and it's also a reason not to assume a scaled AI content operation is safe just because it hasn't been caught yet.
The uncomfortable middle ground is where most teams actually operate: not running a plagiarism farm, but leaning harder on AI-assisted content production than they'd want to defend in public. Our own read on that middle ground, laid out after SEO Week 2026, is that speed and credibility aren't opposites you have to trade off evenly. The trade gets punishing fast once volume outruns originality, and National Today's 850,000 URLs are what the far end of that curve looks like.
Few teams will ever approach that scale, which can make a case this extreme feel like it doesn't apply. That's the wrong takeaway. The mechanism that got National Today caught, public scrutiny surfacing a pattern a reviewer could then confirm and act on, doesn't require 850,000 URLs to trigger. A content directory a hundredth that size, built on the same premise of volume over originality, is vulnerable to the same review once something points a reviewer at it: a competitor complaint, a journalist's tip, or simply a spot-check that happens to land on the wrong folder. Scale determined how dramatic this particular collapse looked. It didn't determine whether the underlying tactic was safe.
Do this next
Audit your own content production the way a Google reviewer would: pull a random sample of pages published in the last quarter and ask, for each one, whether it says something a competitor's AI-generated equivalent couldn't say just as easily. If a meaningful share of your recent output fails that test, that's the section of your content marketing plan most exposed to a scaled content abuse review, regardless of how much traffic it's currently pulling. Growth built on volume alone is borrowed, not earned, and the bill on that loan comes due exactly when a reviewer, a journalist, or a competitor decides to look closely. Fix the originality gap before someone else finds it for you, and treat technical SEO hygiene, canonical, robots, and index coverage, as your early warning system rather than your last line of defense.
One more check worth adding to that audit: if a page's byline claims a specific human wrote it, be able to defend that claim if asked. National Today's collapse wasn't purely a content-quality problem; it was a trust problem, invented reporters attached to lifted reporting, and that combination is what turned an SEO story into a national news story. A scaled-content review looks at volume and originality. A trust review looks at whether the site is telling the truth about who's behind it. Both are worth passing before either one gets asked.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.