The pitch is hard to resist. A vendor offers to generate hundreds or thousands of landing pages built around every keyword your category throws off - "stainless steel water bottles for hot yoga in summer," "small woven storage baskets for narrow bathroom shelves" - a page for every long-tail phrase sitting in the search data.
The pages go live in minutes. Traffic begins to climb.
On a dashboard, it can look like the cheapest growth you've ever bought.
Why isn't every brand doing this?
Surely this must be the secret that other brands just haven't figured out yet, right?
Wrong.
Somewhere between 3-12 months later, traffic doesn't just plateau - it drops like a brick, taking the rest of the search traffic down with it - a phenomenon Glenn Gabe at GSQi coined Mount AI, due to the pattern's distinct rise and fall.
The public evidence is now clear enough that it's worth saying plainly: scaling AI-generated pages to chase keywords is one of the most expensive mistakes a Shopify brand can make heading into agentic commerce.
Not because AI writing is inherently bad, but because of what the pattern triggers - and the extent of the damage once it does.
Scaled and helpful content updates are notoriously difficult to recover from.
By participating in Scaled Content schemes, you're jeopardizing your trust with Google and putting all of your organic traffic at risk.
Image Credit: Semrush - There's a New SEO Traffic Pattern Emerging
Google's Scaled Abuse Spam Policy
Google defines scaled abuse as follows:
"Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users... large amounts of unoriginal content that provides little to no value to users, no matter how it's created."
Two of Google's own listed examples of the practice land almost exactly on keyword-first page programs: "using generative AI tools... to generate many pages without adding value for users," and "creating many pages where the content makes little or no sense to a reader but contains search keywords."
If a program's honest description is "a page for every keyword," it's describing the policy violation.
What Scaled Content Looks Like
The tell isn't always AI. Plenty of good pages are drafted with the help of AI, and Google says as much in it's AI content guidelines.
The tells for non-helpful AI content are:
the same page skeleton stamped out over and over, with each copy built around a different keyword
the focus of the page is the keyword and not the positioning of the brand
i.e. "Shade is important at the beach so you don't get too hot..." vs "Our umbrellas range from 36-65 inch canopies to provide shade for yourself our your whole family."
the pages target search intent with impressive breadth and answer it with startling genericness
the copy is generic enough that you could swap in a competitor's name and the content would read the same - which is the opposite of what a distinctive brand should ever publish
pages filled with prose that reads fine sentence by sentence but says nothing unique to the brand or helpful to the consumer
keywords used forced and awkwardly throughout
real internal linking replaced by an identical block of "related pages" pasted onto the bottom of every generated page accompanied by an early 2000's style html sitemap
the same AI common phrases recuring across a huge share of the pages - "a touch of," "timeless elegance," "elevate this", "elevated that"
of course, mdashes everywhere.
The Penalties are Proven
The potential Google fall-out isn't a hypothetical future update.
In early 2026, SEO analyst Lily Ray documented an algorithmic shift around January 20 that hit companies scaling this kind of content, including one with 2,000 such pages.
Across more than 220 sites she tracked using content-scaling platforms, 54% lost at least 30% of their peak organic traffic, 39% lost half or more, and 22% lost three-quarters or more - most on the same arc of a fast climb followed by a steep fall.
The enforcement can be blunt.
In one documented case, an 850,000-page AI-generated directory was hit with a manual action for scaled content abuse and removed from Google's index wholesale.
That's the part most brands underestimate: a penalty at this level isn't scoped to the junk pages. It can drag on the standing of the whole domain - including the real pages that were never the problem.
The "we'll just win in AI instead" hope doesn't hold either.
AI Overviews, and Google's AI Mode largely ground their answers in the search index, so when a site is removed from that index it tends to disappear from those AI answers as well.
The same case above also saw the content vanish from ChatGPT once Google dropped it. With AI Mode alone past a billion monthly users and AI Overviews reaching roughly 2.5 billion, that's exactly the surface a brand can't afford to be cut out of; there's no side door around the index.
The Bet Most Brands Are Underpricing
We've heard the reasoning out loud: if we get caught, we'll just stop doing it. This mental model assumes the downside is small and contained.
Here it's neither.
The upside is a temporary lift in low-intent keyword traffic that, on the evidence above, tends to fade and often reverses within a year.
The downside is a sitewide demotion or a manual action affecting the whole domain - your genuinely good pages included - that propagates into AI answers too, with recovery measured in years and never guaranteed.
That's not a parking ticket against one page.
It's a bet with your whole storefront's standing as the stake, placed with a fading, low-quality return on the other side. Weighed honestly, the reward rarely justifies the exposure.
Even Good Content Might Not Save You
The natural objection to scaled content is that if the pages are well-written and genuinely useful, they should be safe.
That instinct isn't wrong about content in general - Google rewards helpful, people-first pages, and says plainly that using AI to help write them is fine.
The problem remains two things sitting underneath a page-scaling program:
- why the pages exist, and;
- how they're made.
pages "generated for the primary purpose of manipulating search rankings and not helping users."
Purpose is the first test: a page built to rank for a keyword rather than to provide use to a real person is already on the wrong side of the line.
Production is the second: when that intent is executed at scale, fast and programmatically, across hundreds of near-identical pages, it fits the pattern the policy describes.
A single page can read perfectly well and still meet that definition, because decent writing doesn't change why the page exists or how it was produced.
That's the work no matter how it's created is doing in the policy - the polish on any one page isn't what clears you.
Practitioners who study these penalties describe the same thing.
The technical SEO David Quaid frames enforcement as being "not about quality... it's about a mechanism" - publishing velocity, the share of a site that's templated, repeated title and URL patterns.
His heuristics are informed inference rather than confirmed Google signals, but they line up with the policy: what gets read is how and why the pages were produced, not whether the sentences scan.
There's a timing trap inside this, too.
New pages get an early lift - a freshness boost - that can look like the program is working. When it fades, Google re-samples the pages against an evolving quality bar and decides whether they've earned continued indexing and crawl budget.
As Dan Taylor at Search Engine Journal put it:
"AI is simply the latest, and easiest, scapegoat for a fundamental breakdown in the content pipeline," and leaning on volume is "a vanity metric that guarantees long-term resource waste."
The spike you saw in month two was the boost, not durable performance.
And there's a risk specific to how these pages are usually made.
Many page-generation platforms like Optiversal run an an easily identifiable underlying template across clients.
So your exposure isn't only your own publishing pattern - if Google moves against that footprint at the vendor level, every site using it is caught in the same net at once.
You're taking on a risk you don't fully control, on behalf of a tool you don't own.
The Way Through: Content that Actually Clears the Bar
The alternative isn't "publish less and hope." It's publishing the kind of content the search algorithms and LLM systems are built to reward.
Google's own guidance asks for content "created primarily for people, and not to manipulate search engine rankings," and looks for genuine Experience, Expertise, Authoritativeness, and Trustworthiness (E-E-A-T).
For a brand, that's not an abstract standard - it's the one thing a generic tool structurally can't produce: the brand's unique selling proposition, your firsthand perspective, your real product detail, the questions your own customers actually ask.
That specificity is rarer than it sounds, which is exactly why it works.
Analysis of top-ranking pages found they carry only about four genuinely unique data points on average, so a page grounded in real detail doesn't have to be encyclopedic to stand out - it just has to say something true that the AI slop pages can't.
Pair that with interlinking that's actually structural - a hub connecting down to its sub-collections and back up again, building a real silo that offers customer direction, rather than the same footer block on 500+ URLs - and you have pages that earn their rankings instead of borrowing them.
There's also a quieter reason the scaled-page approach is aimed at the wrong target entirely.
A long-tail query like "small woven storage baskets for narrow bathroom shelves" doesn't need its own bespoke landing page - an answer engine can assemble that answer itself from your structured product data, provided the products are set up properly.
The size, material, color, and dimension values that make a basket "small," "woven," and fit for a "narrow shelf" all live in your catalog.
When those attributes are exposed cleanly - through proper variant setup, complete metafields, and ProductGroup schema - LLM-driven surfaces can read them directly and match the product to the intent, no keyword-stuffed page required.
The work that actually moves the needle isn't spinning up hundreds of long-tail keyword pages; it's making your real products legible to the models, so the answer engine can do the matching on structured, trustworthy data instead of guessing from thin prose.
This is the heart of Agentic Commerce Optimization (ACO): fewer, better, brand-true pages that compound, over more pages that decay.
It's slower, but it also survives the updates that clear the shortcuts away - and it builds the brand recognition AI engines increasingly lean on when they decide who to surface.
The shortcut's entire appeal is speed. The bill, when it comes, is charged to the whole site.
We'd rather help a brand win slowly and keep it.
