Table of Contents
- 1. AI Spam Is Still Spam, Even When It Reads Well
- 2. The Bottleneck in Content Scaling Moved From Production to Review
- 3. Decay Is Silent, So Measurement Is Your Early Warning
- 4. Non-Commodity Content Pays the Bills
- 5. Prune Before You Publish
- Scale What You Can Defend, Cut What You Can’t
- Frequently Asked Questions
- Content scaling with AI generated content is being corrected again, and the teams that survive pair automation with real human review.
- Google’s October 1 guidance calls manual fact-checking of all AI generated content critical, which makes review your true bottleneck.
- Audit your archive first, prune what is decaying, and move budget from volume to original research.
Disclosure: This article contains affiliate links to Semrush. If you sign up through them, I may earn a commission at no extra cost to you.
Content scaling with AI generated content was supposed to be the cheapest growth lever marketing had ever seen, and for a while it looked that way. Now the reports are consistent. Google is quietly halting indexing of entire site sections, one 17,000-page programmatic site reportedly remains deindexed months after cleanup, and Andy Crestodina’s data shows the share of marketers reporting strong content results has hit a 10-year low.
That is not a reason to abandon AI. It is a reason to rebuild around editorial safeguards, content pruning, and proprietary insight. If you scaled AI generated content in the last two years, assume some part of your archive is decaying right now. Here are the five things worth knowing before your next publishing sprint, starting with the one that explains all the others.
1. AI Spam Is Still Spam, Even When It Reads Well
Google has punished scaled, low-value content before, and the signals that decide who survives (originality, authority, demonstrated effort) have not changed.
Think back to 2011 and 2012, when article spinning flooded search results. Google answered with Panda and Penguin, two major core algorithm updates from the early 2010s designed to clean up results and penalize low-quality SEO tactics. Today’s content scaling swaps spinning software for language models, but the pattern is familiar.
Jan-Willem Bobbink, writing about his AI search experiments at Advanced Web Ranking, estimates that roughly 95% of programmatic SEO is not working, and he says the industry may “head back to a kind of 2012.” Treat the 95% as a practitioner’s estimate, not a measured industry statistic. The direction, though, matches what other experienced SEOs are saying, another Panda-style correction is coming as query fan-out tactics get copied at scale.
Query fan-out, in plain terms, means splitting one topic into dozens of narrow pages that each match a sub-question. It works until everyone does it, and the pages stop being different from each other or from what already ranks.
Here is the encouraging part, a known pattern has a known ending, which means the survival playbook is known too. Generation is now free, so the scarce resources are editorial judgment, proprietary insight, and the ability to spot decay early. Lead with continuity, not panic. “AI content is dead” and “ship more, faster” are both wrong.
2. The Bottleneck in Content Scaling Moved From Production to Review
Writing used to be the constraint. Now checking is, and Google’s latest guidance makes that checking close to non-negotiable.
On October 1, 2026, Google updated its guidance on using generative AI content. It now says it is critical to manually fact-check and review all AI generated content before publishing. Industry coverage reports that the review extends to title tags, meta descriptions, structured data, and image alt text.
One honest caveat: this sits in Google’s help documentation as guidance. Google has not announced a new penalty, and it has not defined what an adequate fact-check looks like. I would still treat it as the new floor, because it lands on a problem teams already have. They are producing more content than they can fact-check, slop-check, and approve, and small teams say they cannot absorb a mandate to review everything down to the alt text.
Peep Laja has made a related point:
Everyone’s output is up, but impact is not. Attention and trust are the new scarce resources, and the idea matters more than the volume.
A review gate does not need to be complicated:
- Name one human editor for every AI-assisted page before drafting starts.
- Check claims, numbers, and sources first, then titles, meta descriptions, schema, and alt text.
- Cap weekly output at what your team can genuinely review, not at what your tools can generate.
3. Decay Is Silent, So Measurement Is Your Early Warning
Google rarely tells you when it stops indexing a section, so you have to detect decay yourself.
Content scaling raises the stakes here, because a single decaying section can hold thousands of pages. Programmatic site owners describe the same experience: Google stops indexing a whole section without explaining why, and recovery after cleanup takes months or never arrives. Indexing reports lag, and Google only says “not indexed.” That is why some practitioners check Bing and Yandex first, since a page missing everywhere points to a quality problem rather than a quirk.
There is a second, quieter failure. Content can rank and still earn nothing. Executives are discovering that end-to-end AI content is not producing ROI, even when it is producing plenty of content, and traffic is increasingly treated as a red herring compared with leads and signed business. Measure what the content earns, not only what it attracts.
This is where tooling earns its keep. The Semrush SEO Toolkit has several AI-related features that help with early detection:
- Position Tracking monitors daily rankings across Google and Bing, and it can track visibility in AI Overviews, ChatGPT search, and Google AI Mode.
- Site Audit includes an AI search category and crawler settings, so you can see how AI crawlers reach your pages.
- Copilot flags visibility drops and names the pages that took the hit.
- Sensor shows SERP volatility, which helps you tell a site problem from a wider Google shake-up.
Set a monthly check by site section, not just by page. A drop that hits one folder is the signature of a section-level problem.
Find Out Which Pages Are Decaying
Before you decide what to keep, update, consolidate, or remove, see how your content is actually performing. Semrush’s SEO Toolkit can track your rankings across Google, Bing, and AI search, audit your site for technical and AI search issues, and flag visibility drops early.
4. Non-Commodity Content Pays the Bills
Original research and firsthand expertise outperform generic content, and fewer marketers are producing either.
According to Crestodina’s data, original research and firsthand expertise deliver roughly 1.5x to 2.6x average results. Yet fewer marketers are doing either. The market already knows where this is heading, differentiation is migrating to data your competitors do not have.
Non-commodity content is anything a language model cannot assemble from public text. That means your own numbers, your clients’ outcomes, your tested methods, and your documented mistakes. Building the power of thought leadership starts with a point of view you can defend with evidence only you hold.
A few formats that fit small teams:
- A short survey of your newsletter list, published with the raw results.
- Before-and-after numbers from your own pruning or refresh projects.
- Screenshots and notes from tools you have actually used, with what worked and what did not.
Then pair the evidence with data-driven storytelling, so the numbers carry an argument instead of sitting in a chart nobody reads. This is also the cleanest way to build a content scaling strategy around insight instead of volume, without abandoning AI. Use it for drafts, outlines, and formatting, and spend human time on what only you can add.
5. Prune Before You Publish
Deleting and consolidating weak pages is now an offensive move, not an admission of failure.
Decaying archives can drag a whole domain down. Multiple high-engagement posts from practitioners report that fewer pages produced better rankings. Bobbink says he has been reducing site size since Panda in 2012, and that doing so tends to raise average link value and content quality.
Run an honest portfolio audit and put every page into one of four buckets:
| Decision | When to use it | Example |
|---|---|---|
| Keep | The page earns visibility or leads and is accurate | A tested how-to with original screenshots |
| Update | The topic is right but facts or examples are stale | A guide citing last year’s figures |
| Consolidate | Several thin pages answer one question | Five near-duplicate location pages |
| Remove | No traffic, no leads, no unique value | Templated pages that add nothing |
Consolidation gets much easier when your content architecture is clear. One pillar page per topic, supported by pages that each answer a distinct question. The Semrush Keyword Strategy Builder can map pillar pages and subpages from up to 10,000 keywords, and the Topics report in Organic Research (available on Guru or Business plans) shows how competitors group their content. Use both to decide which pages deserve to survive.
Fewer, stronger pages also lower your link building cost, because every link you earn lands on a page worth linking to. Put the review gate from the second point in front of the publish button, and your content scaling stops outrunning your ability to maintain what you publish.
Scale What You Can Defend, Cut What You Can’t
The pattern is old and the ending is known. Scaled, low-value content gets corrected, and the sites that come through are the ones with originality, authority, and visible effort behind every page. AI did not change those rules. It made it easier to ignore them for a while.
For content scaling, the winning move is measurement plus discipline. Audit what you have published, prune what is decaying, put a human review gate in front of everything new, and reinvest in content AI cannot replicate. Waiting for a recovery that practitioners say takes months, or never comes, is the expensive option.
You do not need to choose between AI and quality. You need to decide, page by page, what you can stand behind. Start with one section of your site this week, and let the data tell you what to keep.
Frequently Asked Questions
Is AI generated content bad for SEO?
Not by itself. The risk is publishing unreviewed, low-value pages at scale, which Google’s guidance says to prevent by manually fact-checking everything.
How do I know if Google has stopped indexing part of my site?
Google often reports only “not indexed,” so compare its indexing report with Bing and track rankings by site section. A sudden drop across one folder of pages is your cue to audit it.
How many pages should I prune?
There is no fixed number. Sort every page into keep, update, consolidate, or remove based on visibility, leads, and revenue, and start with pages that earn none of them.
No Comment! Be the first one.