Was AI-Rewritten Content Reliable? A 2022 SEO Commentary

This is an original article. Please credit the source if you republish it.

The Chinese original contains 2,689 characters and has an estimated reading time of seven minutes.

On the day this article was written in February 2022, I shared an AI-content website on WeChat Moments. That post prompted me to explain the subject in more depth here.

This is the WeChat Moments screenshot.

Part One: full-size screenshots of the sites:

An early-stage AI content site.

The site above was only one small example among the many AI-content sites then appearing. At most, it could be described as being at an early stage.

Now consider sites at other points in the same deliberately playful life-cycle comparison.

The next AI site looked more mature. The screenshot reported roughly 5.7 million organic visits per month at the time.

A mature-stage AI content site.

The next example was the midlife-stage site in the metaphor. Its screenshot reported about 8.3 million monthly organic visits.

A midlife-stage AI content site.

The last example could be viewed as an old-age AI content site:

An old-age AI content site.

The enthusiasm for AI at the time was closely connected with advances in natural-language processing.

Many vendors were promoting GPT-3 technology. Naturally, the tool provider's motive was to earn money. Whether the webmaster would profit was another question; that is why I laughed in the original.

What follows explains the pattern and records my prediction at the time.

2. What Were AI Content and Reassembled Aggregation Sites?

First, let us clarify what I meant by an AI-content site and a reassembled aggregation site.

In earlier years, many people working on Baidu used tools to collect material from large media sites. Sina News, Tencent News and similar publishers were frequent sources.

That method could work on Baidu at the time. It did not necessarily work on Google. My argument was that Google's systems also had weaknesses, which I discuss below.

A site of this kind typically sent crawlers to collect online material, translated the material in bulk and recombined it into pages.

In my 2022 description, AI content was a more advanced method: a system generated new passages and assembled them into an article. Calling those passages completely original was my wording then, not evidence that the result was independent of its training data or source material.

I believed the AI-generated version tended to rank better than the scraped-and-reassembled version because Google was less tolerant of obvious copycat content. That was an observation from the time, not a universal ranking rule.

3. Why Could These Sites Gain Traffic So Quickly?

This part of the argument concerned the underlying logic of a search engine.

I believed these operators had found weaknesses in that logic and used them to gain a large amount of traffic in a short period.

What exactly was the weakness?

Before addressing it, consider the basic purpose of a search engine. This is also part of basic SEO.

Every search engine aims to provide users with the best and most relevant results.

AI-generated content therefore still had to provide information that users wanted before a search engine would treat it favorably.

Content alone was not enough. A site also had to be discovered and indexed quickly and had to satisfy other important ranking considerations.

I did not explore those topics in detail in this article.

In the shorthand I used then, solving those problems could produce a site with a million monthly organic visits. The screenshots were third-party Ahrefs estimates, however, and did not prove a repeatable formula.

The subject matter of these sites also followed a recognizable pattern. Most were question-and-answer or how to sites.

I tried reading some of the pages. They were useful, but they did not fully solve my problem.

As I wrote in the social post, these sites met a user need in one sense, yet did not provide sufficiently detailed information.

4. My Position on These Sites

By then I considered myself a strictly white-hat SEO practitioner. It was not that I enjoyed acquiring traffic slowly; experience had taught me that SEO growth that comes extremely fast is rarely stable.

I therefore hoped readers would not put tactics ahead of fundamentals. The longer you work in SEO, the more clearly you see the importance of the basics.

Historical experience led me to assume that Google had already noticed these sites. Their most obvious characteristics were a rapid flood of content, similar structures and what I called 50-point or mediocre content.

If SEO practitioners could see those characteristics, was it plausible that Google's engineers had not?

I speculated that many such sites were still alive because Google was waiting for its systems to learn their patterns more deeply. This was my theory, not a disclosed Google process.

The 2022 source then said that an October MUM AI Update had been intended to attack AI content; I added that I saw little effect and assumed it was a small test. That premise was inaccurate. Google's official MUM announcement described a model for understanding complex information, and its 2021 Search Central summary said the discussed MUM changes were not yet live and required no website changes. It was not announced as an AI-content spam update.

I predicted that within one or two years the traffic of these AI sites would collapse like the example below. That prediction should remain identified as a dated prediction. Google's later guidance says that appropriate AI assistance is not categorically against its rules; the concern is quality and the use of automation to manipulate rankings. Current spam policies define scaled content abuse by whether pages are produced mainly to manipulate rankings and fail to help users, regardless of whether they are made by AI, scraping or other methods.

The longer I worked in SEO, the more it seemed to move through historical cycles. Backlinks once dominated the conversation; AI dominated it then. What would define the next SEO era? That question made me laugh.

Final Thoughts

My closing argument was that Google would defend the usefulness of its search ecosystem against techniques that damaged it. The original used much more absolute language; it expressed my position rather than a documented promise about enforcement.

I ended with one sentence for readers: in SEO, fast can be slow.

The 2022 post set its first social-sharing threshold at more than 50 reposts and likes.

At that point, I would publish the URL of the early-stage site.

At more than 100, I would publish the mature-stage site's URL.

At more than 150, I would disclose all of them. These thresholds are retained as historical closing context, not renewed as a current offer.