Google Tightens Its Approach to AI Content: What Changed in Late September and Early October 2026?

• Google does not ban content created with the help of AI, but it expects that content to be manually reviewed, fact-checked and assessed for quality before publication.
• Two events coincided at the turn of September and October: the global September 2026 Spam Update and an update to Google’s official guidance on generative AI and people-first content.
• Today Google puts more emphasis on human input, originality, accurate information, the author’s real expertise and value that cannot be reproduced with a single prompt.
• The biggest risk lies in mass-producing content without oversight, original value or credible authorship. A safer process looks like this: AI → human review → fact-check → quality and value assessment → publication.

On September 24, Google started rolling out the September 2026 Spam Update. The update is global, covers all languages and affects search rankings. Google announced that the full rollout could take up to two weeks.

Google has not named a specific new type of spam targeted by this update. Officially, then, the existing Spam Policies still apply, covering scaled content abuse, site reputation abuse and other practices meant to manipulate search results, among others.

On October 1, the official Google Search Central changelog announced:

“Updated guidance on using generative AI content”.

Google explains that the documentation was expanded with information from the Search Quality Rater Guidelines, among other sources, to align it with what the company has presented at its developer events.

Google’s stance on AI itself remains consistent: generative AI can be used to create content.

What is now described in much more detail is the process that leads from generating the material to publishing it.

The most concrete change in the guidelines concerns content verification.

Google reminds readers that generative models do not work like a verified database of facts. They predict a likely sequence of words based on their training data, which is why they can produce false information and hallucinations.

The new guidance contains a very strong statement:

“It is critical to manually factcheck and review all AI-generated content”.

According to Google, then, manually reviewing and fact-checking AI-generated content before publication is critical.

The simplest workflow usually looks like this:

prompt → AI → publication

A safer process, in line with the guidelines, should look like this:

AI → human review → fact-check → quality and value assessment → publication.

And this is not only about articles.

The rule also applies to automatically generated elements such as titles, meta descriptions, structured data and image alt text.

Automating these elements can still be useful. At scale, however, checking whether the information is correct and actually describes the content of the page becomes even more important.

Source: An excerpt from Google’s official guidelines 

At the same time, Google expanded its documentation on helpful, reliable, people-first content.

It now includes a very interesting section describing four elements that Search Quality Raters take into account when assessing the main content of a page:

Effort, Originality, Talent or Skill and Accuracy.

Source: An excerpt from Google’s official guidelines

The first criterion is particularly interesting.

Google defines effort as the amount of human work put into creating the content, or the system that creates it.

As an example of low effort, it points to automatically generating pages from feeds, or using generative AI to create large numbers of texts without manual oversight and curation.

This is an important shift in perspective. The mere fact that AI was used to create a text says little about its quality. What matters far more is the whole process: the scale of automation, human review, fact-checking, expert input and the value added before publication.

AI can produce hundreds of texts in a short time. Google is paying more and more attention to how much real work and value sits between generating the material and clicking “publish.”

The second element is originality.

Google has long encouraged publishing original information, analysis and research. Now it ties this principle even more closely to the assessment of a page’s main content. This matters especially in the world of generative AI. A model can produce a plausible-sounding article in seconds, based on information that already exists online. Material like that does not necessarily add any new value, though.

That is why, alongside the question:

“Is this text good enough to publish?”

it is worth asking a second one:

“What will users find here that they won’t find in the next five search results?”

Originality does not have to mean conducting a major study for every article. If the content is being created for a client, it is worth asking someone who really knows the subject for even two or three sentences of their own commentary.

What does the problem look like in practice? What do customers ask about most often? What does the company observe in its own data? Is there an exception to the rule that existing materials do not cover? A few sentences like these can add more to an article than yet another paragraph generated from information already available in the search results.

It is precisely the knowledge that exists inside an organization that can be one of the most valuable sources of original content in the age of AI.

In practice, the role of the content process is changing too. What is becoming increasingly important is drawing out knowledge from the client and their experts, and then turning it into content that is useful to the audience.

How Do We Ensure Originality When Creating Content for Our Clients?

At Delante, our content creation process builds on the knowledge and materials our clients provide: from their Tone of Voice and brand communication guidelines, through expert materials and product knowledge, to information from the people involved in a given process. This lets us reflect the brand’s perspective and its experts’ experience in the content, instead of limiting ourselves to information available in the search results.

For large brands, another important part of the process is detailed subject-matter, product and legal review, as well as following the brand’s communication guidelines, such as lists of approved and banned words and phrases, approved claims and specific communication rules. These are resources an AI model simply does not have on its own.

One of the most interesting changes concerns authorship. Google has long recommended making it clear who created the content and what expertise they have. Now the documentation also explicitly mentions examples of deceptive authorship.

Among the dishonest practices Google lists is creating fictional author profiles using:

• an AI-generated photo,
• a made-up first and last name,
• false qualifications or experience.

This is especially relevant for sites that, in recent years, have scaled their content by creating dozens of “experts” that exist only as a generated avatar and a short bio.

Authorship and stated expertise should reflect reality. AI can support an expert’s knowledge and help turn it into content, but the expert should be real.

Google has long encouraged site owners to look at their content through three questions:

WHO

The author should be identifiable wherever users would naturally expect information about who wrote the content. The information about their experience and expertise should be true.

What’s more, an expert can be a source of real value for the content. Instead of adding their name to a finished article, it is worth drawing on their experience while the material is being created.

HOW

If automation or AI played a significant role in creating the material, Google suggests considering an explanation of how and why the technology was used.

How much to disclose naturally depends on the context. We expect a different level of transparency from a simple product description than from a financial analysis, a test or health-related content.

WHY

According to Google, this is the most important question.

Content should be created first and foremost because it is useful to its audience.

If automation is used to mass-produce pages whose main purpose is to rank in search engines, it may fall under scaled content abuse.

What counts, then, is both the quality of each individual piece and the purpose behind the entire content production process.

Google has not said that the September 2026 Spam Update targets AI content. Nor do we know which specific systems or types of violations it changed.

What we do know is that:

On September 24, a global Spam Update began, covering all languages and markets.

On October 1, Google significantly expanded its documentation on AI content and content quality.

In the new materials, Google puts more emphasis on human oversight, effort, originality, talent or skill, accuracy and genuine authorship.

At the same time, the existing rules still clearly point to the risk of mass automation created primarily to gain search traffic.

For large websites, the problem is more complex than an “AI or human” decision.

An e-commerce store can have tens of thousands of products. A marketplace can have hundreds of thousands of URLs. A site operating internationally has to keep information consistent across many languages and markets.

Automation is therefore often a necessity.

What becomes key is designing a system that lets you use the scale of AI without losing quality, credibility or control.

How Do We Approach This at Delante?

For a large beauty and drugstore retailer in Poland, we run a process for creating and optimizing product descriptions. Because of the very wide product range and the large number of product pages, the key was to develop a model that would let us efficiently increase content coverage and fill content gaps while keeping communication consistent and the descriptions at the right quality.

We began by working out the communication rules together with the client. We prepared materials on Tone of Voice and communication style, which became the reference point for the entire process. Together, we also agreed on the structure of the product descriptions and the scope of each section. As a result, every description addresses the user’s most important needs and delivers information in a way that is consistent with the brand’s communication.

The descriptions are based on the available product information, including ingredient lists and the properties of individual ingredients. This lets us create content that does more than describe a product: above all, it helps users understand what the product is for and choose one that fits their needs.

Another important element of the process is multi-stage content review. First, the content team checks the descriptions for accuracy, consistency with the agreed structure and the brand’s Tone of Voice. The materials then go to the client for subject-matter and legal review. A process designed this way lets us work on a large number of products and systematically increase the site’s content coverage without sacrificing quality or consistency of communication.

The new guidelines are no reason to mass-delete content just because it was created with the help of AI.

They are, however, a good moment to audit your process.

Above all, it is worth checking whether your content has a real owner and a QA process, whether the facts in it can be verified, whether author profiles correspond to real people and real expertise, whether the materials offer value of their own, and whether automation has created large groups of near-identical pages.

It is also worth checking how much of the knowledge available inside the company actually makes it into its content.

Sometimes improving a piece may mean adding company data, an expert comment, a real customer example, product experience or simply two or three sentences from someone who knows the subject from practice.

See also: Will AI-Generated Content Need to Be Labeled? What the AI Act Means for Marketing from August 2, 2026

The full impact of the September 2026 Spam Update can only be assessed once the rollout is complete. Google has not shared details that would allow us to say the update is directly related to AI-generated content.

The October documentation changes, however, offer a much more long-term signal.

AI does not abolish the rules of good content. It makes enforcing them more important.

In a world where anyone can generate a correct text in seconds, the advantage increasingly comes from elements that are harder to generate with a single prompt: brand experience, proprietary data, real experts, original knowledge, methodology and actual work done for the user.

Sources:

https://developers.google.com/search/docs/fundamentals/using-gen-ai-content

https://developers.google.com/search/docs/fundamentals/creating-helpful-content