Writing for the web has changed significantly in recent years. Until not too long ago, when people talked about SEO writing, they obsessed over keywords, related terms, and synonyms—with the result that they often catered to His Majesty Google at the expense of the poor user, who was forced to read concepts that were repeated and rehashed over and over.
Today, the trend seems to be shifting. When it comes to web content, in fact, the content that is rewarded with the most visibility is primarily that which offers something original, useful, and specific. The point is: SEO is fine, but content must be designed first and foremost for people. Google itself, in its most recent guidelines, urges creators not to simply recycle what others have already written online or what could easily be produced by a generative AI model.
Of course, talking about “human” writing at a time in history when practically everyone uses AI raises a significant issue.
And that’s exactly the point: even though Google doesn’t prohibit the use of artificial intelligence in content creation, it all comes down to the quality of the content itself. So no, using ChatGPT or other AI tools to write an article doesn’t automatically mean you’ll drop in search rankings. There is no automatic penalty simply because content was generated, in whole or in part, by AI. The issue is rather one of originality and, above all, the methods and purposes by which that content is produced. But let’s take a closer look at the matter.
Google does not penalize an article simply because it was written using AI
First of all, Google provides a set ofguidelines and policies regarding spam that explain which practices can negatively affect a website’s visibility in search results. This is not, therefore, a specific rule targeting artificial intelligence: the same policies apply regardless of how content is produced.
In fact, when it comes to the use of AI, the search engine primarily evaluates the result and the value it offers the user—not simply the tool used to produce it. And there isn’t even a so-called “maximum AI percentage” allowed on a website.
In other words, content created with the help of artificial intelligence can easily rank well, whereas content written very poorly by a human may perform very poorly or, if it violates anti-spam rules, face far more serious consequences.
In short, the key issue remains the content. And, above all, the way it is produced. Because what Google is trying to combat is not automation itself, but its use to create large quantities of low-value pages, primarily for the purpose of manipulating search results.
From repeated keywords to content created solely to target search queries
And this is where things get more interesting.
For years, part of SEO writing followed rather mechanical methods: primary keywords, secondary keywords, related keywords, and synonyms to be distributed throughout the text according to more or less rigid patterns. Taken to the extreme, this approach produced extremely clunky texts, with keywords repeated ad nauseam and unnatural phrasing. The page ended up being built around the keyword rather than around what actually needed to be explained.
Please note: this isn’t simply a matter of meeting a deadline or using a keyword correctly. We’re talking about taking it to such extremes that the text becomes redundant, unnatural, and—in the worst cases—almost devoid of information that’s actually useful to the user.
Google has long viewed techniques of this kind as problematic. This does not mean, however, that they haven’t worked for years or that they haven’t helped shape a certain approach to online content: identifying what people are searching for and creating a page specifically designed to capture that search.
Today, this dynamic may be even less obvious. An article may be accurate, readable, and well-structured, but exist primarily because a particular search query has a high search volume.
The result can be a potentially endless sequence of articles:“How to store zucchini,”“How to store eggplant,”“How to store bell peppers,”“How to store carrots.” All are grammatically correct, all follow the same structure, and all are capable of answering a query. But they don’t necessarily add anything original compared to the dozens of pages that already answer the same question.
The real breakthrough in AI is speed
At this point, it was above all the speed of production that truly changed the publishing market.
Before artificial intelligence, generating this type of content still required a certain amount of work and time on the part of the human copywriter who produced it. And, as a result, it entailed a certain cost for the copywriter’s employer. Creating hundreds of articles meant employing editors, dedicating hours of work, and allocating resources.
With the advent of artificial intelligence, generating content of the same type has become significantly easier, faster, and less expensive. And in some ways, it’s even more natural: a generative model doesn’t need to mechanically repeat the same keyword a hundred times to produce an optimized, readable, and seemingly complete text.
This is where the scale of the phenomenon changes. Whereas producing 100 pages used to mean commissioning 100 articles, today part of the process can be automated. The same structure can be applied to dozens or hundreds of topics, multiplying the number of pages on a website and making it possible to systematically cover a huge number of queries.
The cost and time required to produce each new article decrease, while the amount of content that can be published increases dramatically. And when the criterion becomes simply“this query exists, so we can create a page for it,”there is a risk of ending up with hundreds of articles that are very similar to one another—formally correct but essentially interchangeable.
Artificial intelligence, therefore, did not invent this type of content. It has made it much easier to produce it on an industrial scale.
"Scaled Content Abuse": When Quantity Becomes an SEO Problem
Google has a fairly specific term for one of the variations of this system: " scaled content abuse," which is translated in the Italian guidelines as“abuso di contenuti su larga scala.”
This definition is important because it also clarifies one of the most common misconceptions about the relationship between AI and SEO. Google refers to large-scale content abuse when a large number of pages are generated with the primary goal of manipulating search rankings rather than helping users. This typically involves large amounts of unoriginal content that offers little or no added value.
And artificial intelligence is just one of the ways this can happen. Google explicitly mentions the use of generative AI tools to produce many pages with no added value, but it also includes scraping and the automatic reworking of other content, the assembly of information from different pages, or the creation of many pages containing keywords but offering little value to readers. Above all, it specifies that the problem exists regardless of how the content is created.
So a website can generate scaled content abuse using AI, human copywriters, or a combination of the two.
The result, however, does not necessarily have to consist of obviously poor writing or false information. And this is precisely one of the most interesting aspects. The articles may appear formally complete, well-organized, and readable, but they offer nothing that makes them truly necessary: the information is assembled without any real curation, there is no original contribution, and each page could be replaced by dozens of others without the reader missing much of anything.
If these practices violate spam policies, the consequences can be significant: Google uses both automated systems and human reviews and states that sites that violate the guidelines may lose rankings or not appear in the results at all.
Editorial decisions also make a difference
Once the rules have been clarified, however, the ball is in the editorial teams' court. Because the problem lies not only in the creation of individual content, but even earlier, in the development of the editorial plan.
With tools capable of quickly producing large amounts of content, it can be tempting to turn every SEO opportunity into a new article. If a search query has a high search volume, a page is created. If there are ten variations of the same question, ten pieces of content are produced. If a topic generates traffic, it’s covered even when it has little to do with the site’s identity or expertise.
The alternative isn't to give up on SEO. It's to ask yourself whether, in addition to the potential for ranking, there's also an editorial reason to publish that content.
This means selecting topics on which the site can truly offer something of value: specific expertise, a worthwhile source, an interview, original data, firsthand experience, or a distinctive point of view. It also means accepting that not every query needs to result in a new page, and that two very similar searches can sometimes be answered by the same content.
It’s no longer just a matter of “covering a keyword,” then, but of deciding why that content should exist on that particular site.
This distinction is particularly relevant today. In its 2026 guide, which also covers search features based on generative AI, Google emphasizes specific, useful content that offers a unique perspective, contrasting it with generic content that merely repackages information already widely available.
So how can you use AI without creating content for Google?
None of this means that a newsroom should ban ChatGPT. If anything, it means understanding where artificial intelligence can be useful and where editorial judgment must continue to play a role.
It can help with preliminary research, organizing information, developing an outline, summarizing materials that have already been gathered, or revising a text. It can speed up parts of the work that used to take much longer. But speeding up the process doesn’t necessarily mean delegating the decision of what to publish to it as well.
That remains an editorial decision.
Content that is difficult to replace with other content still stems from the selection of sources, the verification of information, firsthand experience, independently collected data, interviews, the writer’s expertise, and the context that an editorial team chooses to give to that information.
That’s why the problem isn’t asking AI to help you write an article. It’s publishing an article simply because AI makes it possible: because it’s fast, inexpensive, has all the H2 headings in the right places, and, at first glance, seems perfectly complete.
The structure may be flawless. It remains to be seen whether, behind that structure, there is truly anything worth publishing.
That’s exactly why learning to use artificial intelligence also means understanding its rules, limitations, and responsibilities. In our course on the AI Act, we take an in-depth look at European regulations and what they mean in practice for those who use these tools in their work. Discover here the Viasky’s mandatory training course on Artificial Intelligence here.
