Google Scaled Content Abuse Policy: Which AI Store Pages Are at Risk?

Under Google's spam policy, AI-written store pages are at risk when many of them are generated mainly to catch searches and add little a shopper needs: near-identical product descriptions, keyword-swapped collection pages, and copied supplier text. Check your Shopify store by sampling each page type, then edit, merge or noindex what fails.

The policy does not ban AI. Google's wording says scaled content abuse is many pages generated for the primary purpose of manipulating search rankings and not helping users, "no matter how it's created". That cuts both ways: a human who pastes the same paragraph onto 500 pages is just as exposed, and a model that writes 500 accurate, differing descriptions from real product data is not the target.

Follow me on Instagram@sagnikteaches

What the policy actually lists

Google's spam policies page names five examples under scaled content abuse. Read them with a store in mind, because each one has a store version.

Connect on LinkedInSagnik Bhattacharya

• Using generative AI to generate many pages without adding value. Store version: a script that writes a 150-word description for every SKU from the product title alone, so "Blue cotton shirt" and "Green cotton shirt" differ only by the colour word.
• Scraping feeds or search results to generate many pages, including through automated transformations such as synonymising or translating. Store version: taking a supplier's feed, running it through a rewriter and publishing the output.
• Stitching or combining content from different pages without adding value. Store version: collection pages assembled from fragments of other sites' buying guides.
• Creating multiple sites to hide the scale. Store version: ten near-identical shops selling the same catalogue under different names.
• Pages that make little sense to a reader but contain keywords. Store version: "best cheap waterproof jackets for hiking this week 2026" collections with a heading and a grid.

Google also says that sites violating its spam policies may rank lower or not appear at all, and that detection uses automated systems and, as needed, human review that can result in a manual action. It tells site owners with such content to exclude it from Search. When this tutorial was checked, the page footer showed a last update of 28 August 2026, so read the live text if you are reading this much later.

Subscribe on YouTube@sagnikteaches

Which store pages are usually at risk

Not every AI page is equal. Rank your page types by exposure before you touch anything, because you cannot review every URL by hand.

The question underneath every row is the same one Google's guidance asks: what does this page give a person that they did not already have? If the answer is "the same facts as the other 400 pages with one word changed", the page is the risk.

A worked audit of a 200-product Shopify store

Take an illustrative home-and-kitchen store with 200 products, 40 collections and 60 guide articles. The owner used AI heavily last year. The aim is to decide what to keep, edit, merge or hide, within a working week.

Step 1: count the page types

Export products, collections and guide articles to a spreadsheet. Add columns for word count of the description, whether the text matches the supplier's text, and the last month's search clicks if you have Search Console. Illustrative counts: 200 product pages, 40 collections, 60 guide articles, 12 other pages.

Step 2: find duplicates, not by eye

Ask an AI tool to compare descriptions in batches of 20 and cluster the ones that say substantially the same thing.

Below are 20 product descriptions, numbered. Group together any that
say nearly the same thing with different nouns or colours.
For each group, list the numbers and quote the repeated sentences.
Then list any description that contains a fact not present in the
product data I gave (specs, size, material) so I can check it.

[paste product data and descriptions]

Illustrative result for the first batch: 6 of 20 descriptions fall in a cluster of near-identical text about "durable, stylish and easy to clean", and 3 contain a claim not in the spec table ("oven safe to 250 degrees"). The clusters are the scaled-content signal; the unsupported claims are a separate accuracy problem Google's generative AI guidance flags, because models predict likely words rather than retrieve facts.

Step 3: sample, then extrapolate

You do not have to open every page. Take 10% of each page type, rate them with the rubric below, and treat the failure rate as your guide to the whole set. If 7 of 20 sampled products fail, plan for roughly a third of the catalogue.

Step 4: the scoring rubric

Score each sampled page 0 to 2 on each question. A page scoring under 6 out of 10 goes in the edit, merge or hide pile.

1. Specific facts. Does it state sizes, materials, capacity, care, compatibility or what is in the box? (0 none, 1 some, 2 complete)
2. Different from its siblings. Would a shopper notice a difference if the page were swapped for a similar product's?
3. Accuracy. Is every claim in the page backed by the spec sheet or by you testing the item?
4. Useful beyond the description. Does it answer a real question: sizing, use, limits, who it is not for?
5. Original assets. Your own photos, measurements or a customer review, not just supplier material.

Fixes, in order of preference

Once pages are sorted, apply the lightest fix that works. The order matters: editing saves a page, merging consolidates, noindex hides, deletion is last.

Edit: add what only you know

For product pages worth keeping, the best repair is to add facts the supplier feed lacks. Give the AI your real inputs and ask it to write around them.

Rewrite this product description for a kitchen store. Use only the
facts below; do not add features. Keep it under 120 words. Include a
"Not ideal if" sentence.

Facts: 28cm cast-iron skillet, 3.9 kg, pre-seasoned, works on gas,
induction and in the oven, hand wash only, handle gets hot, our
own testing: heats evenly but takes 6 minutes to reach searing temperature.

Illustrative output: "A 28 cm cast-iron skillet that weighs 3.9 kg, pre-seasoned and ready to use on gas, induction or in the oven. It heats evenly, though in our testing it takes about 6 minutes to reach searing temperature. The handle gets hot, so keep a cloth to hand, and wash it by hand only. Not ideal if you want something light enough to lift with one hand." The facts are yours; the sentences are the machine's. The "not ideal if" line is useful to shoppers and nobody else's page has your test result in it. Shopify's own help page says to avoid copying a manufacturer's text because other sites already use it, and the same goes for lightly reworded versions of it. If the description is AI-generated and goes to Merchant Center, it also needs to be sent in the structured description attribute labelled as generated by AI; the tutorial on rewriting Google Shopping titles explains the equivalent for titles.

Merge: one stronger page instead of five weak ones

If you have five collection pages that all describe "stainless steel water bottles" with slightly different words, choose the best URL, write one page that covers the topic properly, and redirect the others to it. In Shopify, the URL redirects tool (under Content, then Menus) handles this. Shopify only lets you redirect from an address that no longer works, so retire or rename the old collection first, then point its old URL at the surviving page. After merging, check that no menu or internal link still points to a retired page.

Noindex: pages that shoppers need but search does not

Some pages exist for people already in your store: a filtered view, a printable comparison, an internal campaign landing page. They can stay live and still be kept out of Google Search. Shopify's help page offers two routes. The first is a robots meta tag in the head of the theme layout file:

{% if handle contains 'page-handle-you-want-to-exclude' %}
<meta name="robots" content="noindex">
{% endif %}

The second is a metafield called seo.hidden set to 1 on a page, article or product, which hides it from sitemaps, search engines and your store's own search. Shopify notes that this metafield can be used by third-party apps, so you may find the name already taken. For a product you want reachable only by direct link, Shopify suggests the Unlisted status instead.

One rule from Google's noindex documentation matters here: the page must not be blocked by robots.txt, or Google will never see the tag. Leave the page crawlable and let the noindex do its work. Also remember that noindex removes a page from search results, not from your store: shoppers with the link can still open it. Google says some other search engines may interpret the rule differently.

Delete: only when nothing needs the page

If a page has no clicks, no links, no use in your navigation and no reason to exist, remove it and let it return a not-found response, or redirect to the closest useful page if it had any inbound links. Do this last and in small batches, so you can see the effect.

Outcome of the sample audit

Back to the 200-product store. After the sampling, the owner's piles look like this (illustrative):

• Products: 120 fine as they are (own specs, photos, differing text); 55 edited with real facts; 25 are near-duplicate variants of one product, merged into 6 pages with a variant selector and redirects.
• Collections: 18 kept; 12 low-value keyword variants merged into 5 stronger collections; 10 filtered views noindexed.
• Guide articles: 22 kept; 25 rewritten with the owner's own testing notes; 13 deleted, 6 of them redirected to a related guide.

The work takes about two weeks of evenings, not a weekend, and the owner records every change in a sheet so that the next batch of AI drafts goes through the same rubric from the start. For the guide-article side of this, the tutorial on writing website articles with AI that still rank covers how to add original value to each post.

Collection pages: a before and after

Collection pages are where most stores overreach, because they are cheap to multiply. Here is an illustrative before and after for a collection of cast-iron cookware.

Before (AI, no inputs). "Discover our wide range of cast iron cookware. Our cast iron cookware is perfect for every kitchen. Browse cast iron cookware today." Sixty words of that, repeated on "cast iron pans", "cast iron skillets" and "cast iron frying pans" collections with the nouns swapped.

After (one page, real inputs). A single collection with a short guide above the grid: which sizes suit one, two or four people; the weight range (from 2.1 kg to 6.8 kg across the range, which matters for people who struggle to lift); which pieces work on induction; how to season; and when enamelled is the better buy. The three old collections redirect to it. The page is longer, but it exists once and answers a shopper's actual question: which pan should I buy?

Prompt for the guide above the grid

Write a 150-word buying guide for the top of a cast-iron cookware
collection page. Use only these facts and nothing else:
- Range weights: 2.1 kg to 6.8 kg
- Induction-compatible: all skillets, not the grill pans
- Enamelled pieces cost about $30 more and need no seasoning
- Customers most often ask: "Which size for two people?" Answer: 26 cm.
End with one sentence telling someone who wants a light pan to look
at the carbon steel collection instead.

The final sentence matters: a page that sends the shopper elsewhere when it is the wrong fit is clearly written for the shopper.

Product variants, FAQ blocks and translations

Three smaller patterns catch stores out.

• Variant pages. If each colour or size has its own URL with its own description, you have created near-duplicates by design. Use one product page with a variant selector, as Shopify does by default, and only give a colour its own page if the product genuinely differs.
• FAQ blocks. A block of eight generic questions pasted under every product adds repetition. Write two or three questions that real customers asked about that item, and leave the block out where there are none.
• Translated or reworded pages. Google's examples include automated transformations such as translating or synonymising content with little value added. Translating your own accurate pages for customers who speak another language is a service, and it is different from spinning other people's pages. The test is whether the result helps the person reading it.

When a big catalogue is unavoidable

An illustrative B2B software reseller or wholesale distributor may list 3,000 SKUs, such as cables, fixings or licence tiers, where the data does the talking and there is no room for a long description. Scale on its own is not the problem Google describes. A page for each SKU with a clean specification table, compatibility notes and stock information is a legitimate page; what makes it risky is padding that table with 200 words of AI prose that says nothing.

For such catalogues, keep the template short and factual, put the effort into accurate attributes, and write extra copy only for the top 5% of products by revenue. If thousands of pages are likely to be nearly identical, noindex the low-value ones rather than publishing them all to Search. A tidy catalogue has fewer, better pages in the index, and if you use the seo.hidden metafield, Shopify also drops those pages from its sitemap.

How to tell whether the clean-up worked

You will not get a message from Google saying a policy concern is cleared unless you had a manual action. Instead, watch these:

1. Pages indexed in Search Console, by type, fall to the number you intended.
2. Clicks per indexed page rise over the following weeks.
3. Reports of crawled but not indexed pages stop growing.
4. Your shopping feed shows no new mismatch warnings after you edited descriptions.

Give it weeks, not days. Search systems take time to recrawl, and noindex and redirects only apply when Google next visits. Resist the urge to republish unchanged pages. Write down the date of each batch of changes so you can match it with the data.

A mistake that looks like thrift

Here is how a store gets into trouble without meaning to. A seller finds that adding "in [town]" variants of a collection page brought a few visits, so they have AI produce 400 versions, each with the same body and a different place in the heading. Visits rise briefly, then fall; the pages are all but identical, shoppers bounce, and Google's systems have a clear pattern to detect. The cheaper path was one genuinely useful delivery-and-returns page for the area the store serves. The lesson is to ask what the shopper in that town gets that the shopper elsewhere does not.

Review habits that stop the problem coming back

The cleanup is only half the job. Without a habit, the next AI batch recreates it. A short rule set is enough:

1. No AI page goes live without at least two facts the AI was given, not invented.
2. A person reads each description against the spec sheet before publishing. Google's guidance says to manually fact-check AI content, including titles, meta descriptions and structured data.
3. Never publish more than one page for the same search intent.
4. Run the near-duplicate comparison on every new batch of 20 or more pages.
5. Keep a log of the date, the tool and the prompt, so you can find the batch if something goes wrong.

Titles and meta descriptions need the same discipline, which is what the tutorial on fixing Shopify titles and meta descriptions in bulk sets out. And if you are writing fresh descriptions rather than repairing old ones, writing Shopify product descriptions with AI shows a prompt structure that supplies the facts first.

What to check next time Google updates the policy

Google revises these pages from time to time. When it does, reread the examples under scaled content abuse, compare them with your page-type table, and repeat the sample audit on whichever type looks closest. If you never change your pages' value but only your wording, the risk does not change. If you add real information to each page, it falls, whatever the tool was.

What store owners ask about the spam policy

Is using AI for product descriptions against Google's rules?

No. Google's policy says the problem is generating many pages without adding value for users, and it applies however the content is made. AI-written descriptions that are accurate, specific to the product and reviewed by you are not what the policy describes. A thousand near-identical pages are.

How many AI pages is too many?

Google gives no number. The test is purpose and value: does each page exist to help a shopper, or to catch a search term? A store with 300 products that each have a real description is fine. 3,000 location or keyword variations of the same collection with swapped words are the pattern to remove.

Should I delete thin pages or noindex them?

Edit the ones a shopper might want, merge the near-duplicates into one stronger page with a redirect, and noindex the pages that must exist for shoppers but add nothing to search, such as filtered or comparison views. Delete only when there is no reason for the page to exist at all.

Does noindex hurt my shopping ads?

A noindex tag keeps a page out of Google Search results, and Shopify says its seo.hidden setting does not remove a product from collections or third-party sales channels. Shopping ads still need the landing page to load, so do not confuse hiding from search with removing the page. Test one page before changing a batch, and never block the page in robots.txt.

Further reads

• Scaled Content Abuse: An AI Content Checklist for SEO Agencies — The same policy from an agency's point of view.
• Can a Small Online Shop Use AI to Compete With Bigger Brands? — How a small shop can use AI against bigger brands.
• Shopify Magic vs ChatGPT for Product Descriptions — Shopify Magic against ChatGPT for descriptions.
• Does Writing Etsy Listings With AI Hurt Your Search Ranking? — Does AI-written listing copy hurt marketplace ranking?
• Taking a Local Shop Online With AI: Product Listings in a Weekend — Taking a local shop online with AI product listings.
• Missing GTIN in Google Merchant Center: What AI Can and Can't Fill — A feed problem that sits alongside page quality.
• Ecommerce Store Audit Checklist: Run It With ChatGPT or Claude — A page-by-page store audit you can run in an afternoon: what to screenshot, what to export, how to rank the fixes and what AI cannot see.
• WooCommerce to Shopify Product Import: Variants and Redirects With AI — Map WooCommerce export columns to Shopify's import file, fix variants and images, build a redirect list and audit a sample before go-live.
• Applications of AI in E-Commerce: 9 Uses Mapped to a Small Online Store — Nine uses of AI for a small store, ranked by weekly hours saved and matched to Sidekick, Shopify Inbox, Klaviyo and Gorgias, with plan costs.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Google Search Central spam policies and guidance on generative AI content, noindex documentation; Shopify Help Center pages on SEO and hiding pages from search engines; Google Merchant Center Help on generated content.