Glenn Gabe opened a directory and started counting.
Gabe is the president of G-Squared Interactive (GSQi). The directory was the USA section of the popular lifestyle site, National Today. Inside this section, Gabe counted a nearly absurd 850,000 indexed URLs.
The reporters who wrote those pages did not exist. Gabe described the pages as 100% AI-generated. National Today’s team had leveraged AI to scale a massive amount of online content.
By the time Gabe published his findings, the entire section had been deindexed by Google.
The graph he attached to the case rose like a ski jump and fell like a cliff.
Its called “Mount AI”
In a LinkedIn post last January, Gabe gave the shape of this graph a name: “Mount AI“, due to the mountain-like graph, showing a massive rise in traffic, followed by a disastrous fall.
The National Today case became the textbook illustration for Mount AI. The site had been spamming AI-written local news into its USA section at industrial speed.
Maggie Harrison Dupré, a reporter for Futurism, tried to count a single day’s output on National Today and lost track after counting an eye-watering 300 articles. Inside these franken-articles she found real people’s names swapped for “Jane Doe,” a fabricated quote attributed to the Pope, and a quote about San Francisco crime pasted into stories about the Dallas Cowboys and a Boston biotech firm.
In short, it was slop.
Here’s the thing though: The distribution, while it lasted, was enormous. The ascent up the mountain is real. The heights are also real, albeit brief. Unfortunately for publishers brave enough to scale AI content, the other side of Mount AI is steep and merciless.
Gabe documented National Today’s USA ranking in Top Stories, Google News, the News tab, and likely Discover. Then the curve broke. Some time after Futurism contacted Google, most of National Today’s results disappeared from Google Search and Google News.
Gabe’s follow-up showed visibility collapsing across organic Google search results, AI Overviews, and AI Mode, and most ChatGPT citations of the directory vanishing too. The core holiday-calendar content outside their USA section kept ranking.
So what happened? Gabe infers a scaled-content-abuse “manual action” taken by Google. While Google didn’t name a penalty against National Today, his guess is likely right on the money.
The only thing we have from Google on this matter is the following general statement:
“Our policies prohibit producing content at scale for the primary purpose of manipulating search rankings. While we don’t comment on spam penalties against individual sites, we take appropriate action when we identify violations of our policies. We go to great lengths to fight webspam in our search results, and 99 percent of Search visits are spam free.”
Two hundred and twenty sites, one similar pattern
Lily Ray, vice president of SEO and AI Search at Amsive, spent the spring of 2026 observing similar “mountains” elsewhere across the landscape.
On May 13, 2026 she published an analysis of more than 220 customer sites.
She named no vendors and no domains. Her method: Correlate Ahrefs organic traffic with indexed page counts — corroborate with Sistrix.
The numbers are hard to look away from:
Of the 220-plus sites generating AI content at scale, a majority had lost 30% or more of their peak organic traffic. 39% had lost 50% or more. 22% had lost 75% or more. The repeated shape matched Gabe’s mountain: a page-count surge over six to twelve months, a traffic peak arriving three to six months after the content peak, then a drop that typically went below the prior baseline within a year.
And that’s the important bit: The result of mass scaling AI content was net negative over time.
It’s a near term win, with a long-term loss attached.
What scaled content looks like in the field
Ray catalogued eight templates that kept appearing on the declining sites, with most of them using three or four at once:
- Comparison pages at scale
- “What is X” glossaries
- “Best X for Y” listicles
- Self-promotional listicles ranking the publisher first
- Competitor-alternatives pages
- Programmatic location and language clones
- FAQ farms
- Traditional scaled content like jokes baby names and horoscopes bolted onto B2B sites.
At least 40 sites running self-promotional listicles and GEO-style templates saw declines of 40% to a devastating 95% from about January 20, 2026 through April 2026.
While Google didn’t confirm a named January update, search watchers generally agree something massive happened to scaled content farms during that period. By May 2026, many of the case-study URLs had been 410’d or redirected, while the vendors’ case studies praising the same campaigns stayed live.
A second named graph drew the same outline: Grokipedia, xAI’s AI-written encyclopedia, charted a sharp visibility rise into late January 2026. Then it collapsed. Grokipedia’s rapid decline showed-up in Google organic, AI Overviews, AI Mode, and ChatGPT citations.
Just how bad was Grokipedia’s collapse? Put it this way: Barry Schwartz, CEO of RustyBrick and news editor of Search Engine Land, reported that Wikipedia actually outranked Grokipedia when the search term was “Grokipedia”.
So was there a manual action taken against Grokipedia? The official answer is no. But the mountain appeared anyway.
No matter how the pages are made
Google’s live spam policy on Search Central defines scaled content abuse as:
“when many pages are generated for the primary purpose of manipulating search rankings and not helping users,” a practice “typically focused on creating large amounts of unoriginal content that provides little to no value to users, no matter how it’s created.”
The rule is method-neutral on its face: AI, human, or mixed. Its closing instruction: “If you’re hosting such content on your site, exclude it from Search.”
For quality publications, and high-end brands human-made content has become the law of the land. For others, the temptations and the risks of AI content generation are worth it.
To be fair, the appearance of Mount AI tends to correlate with scale. Using AI to write an article here and there is not likely to land your site in hot water. Using AI to write 10,000 articles on a random Tuesday, is basically an invitation to the Mountain.
To scale or not to scale?
Scaling AI content behaves like a leveraged trade.
It amplifies whatever the content already is. Scale original reporting, real expertise, and pages that exist to help a reader, and traffic gains might last. (With an emphasis on might).
Scale commodity pages that a generative model could produce for anyone, in volumes no editorial team could possibly achieve, and the graph will invariably look like Gabe’s mountain: You’ll likely see a peak three to six months after publication date, then a fall that frequently plunges to a level below where the site started.
In other words, a net loss. On the other side of that mountain isn’t a sustained plateau. It’s more like a deep valley.
Meanwhile writers and agencies selling human-written work, keep racking up slow, steady and consistent wins.
This isn’t to say that the use of AI is inherently bad. The workflow of serious writers can still leverage AI without scaling abuse.
Wordmetrics is a case in point: give it a search-term it conducts a real-time Google search. Then Wordmetrics’ real-time crawlers examine the top 50 ranking pages, extract the article text, and use machine learning and matrix reduction to surface the terms and entities that correlate with what ranks. Wordmetrics isn’t doing the writers’ job. It’s providing knowledge and insight to the writer, while human writers still do 100% of the writing.
Tools like Wordmetrics cannot say why Google prefers a specific word or phrase only that it does — empirically.
In Wordmetrics’ writing interface the software highlights writers’ own words in realtime, giving writers a birds-eye view of topical completeness for their targeted search-term — and assigns their article a letter grade for matching content with search-intent. The end result is a purely human-written article, with an AI powered level of strategy behind the words.
The writing stays human. The intelligence behind the writing is AI assisted.
As for the sites sites plunging down the slopes of Mount AI the verdict seems to be in: Google, like readers around the world, prefers human written content which tells a unique, personal story — and has a distinctly human voice.