Google now takes the use of AI into account in its website rankings. Some practices could result in heavy penalties.
The question of Google’s positioning regarding AI-generated texts has been a topic of discussion in the SEO world since the end of 2022. Until now, the company’s message has always been the same: whether the text is generated using AI or not, it’s the quality of the content that counts, particularly compliance with EEAT (Experience, Expertise, Authority, and Reliability) criteria. But the search engine seems to have changed its tune at the start of this year.
Google: AI-generated texts will now be penalized
In a talk at Search Central Live Madrid, spotted by SEO specialist Aleyda Solís, Google Search relationship manager John Mueller highlighted the new feature: “The lowest rating applies if all or almost all of the main content on the page (including text, images, audio, videos, etc.) is copied, reworded, embedded, generated automatically or by AI, or taken from other sources with little effort, little or no originality, and little or no added value for site visitors.” This directive is a slight departure from Google’s previous position that AI use was not taken into account in rankings.
Google’s Search Quality Evaluator Guidelines, dated January 23, 2025, also mentions this direction. It states that “content from a single page or many websites that is summarized, reworded, or paraphrased by humans or generative AI tools” is now considered “copied or paraphrased content .” But Google’s approach is still intended to be balanced. The document states: “The use of generative AI tools alone does not determine the level of effort or the quality rating of the page. Generative AI tools can be used to create both high-quality and low-quality content.”
How Google Detects AI-Generated Text
How can Google detect that a text is generated by artificial intelligence when no detection tool has shown 100% reliable results to date? In its document, the search engine shares some leads. Sites displaying the following characteristics are mentioned:
- Contradictions between content and legal notices: For example, a page may be presented as a guide for parents, while the site’s terms of use specify that it is a project intended for AI enthusiasts.
- A lack of transparency regarding the origin of the content: Some sites indicate that articles are generated (or partially generated) by AI, without specifying which ones. This vagueness makes all content less reliable in Google’s eyes.
- A stated purpose as experimental: When a site specifies that its content was published solely for research or demonstration purposes, this calls into question its value to a wider audience. The same goes for sites that admit that the information may contain errors or be outdated.
- Fictitious or deceptive authors: Some sites use fabricated or AI-generated profiles (photos, biographies, identities) to make content appear to be written by real people. This practice can be perceived as manipulative.
- Textual clues indicative of machine-generated content: Some pages contain phrases characteristic of AI-generated text, such as “As a language model…” or references to data access deadlines (typically: data limited to September 2021). This indicates automated production with little or no human intervention.
Google’s Consideration of AI: What Are the Consequences for Websites?
Google isn’t tackling the use of AI per se, but it now recognizes that generative AI tools can help produce poor content. The Search Quality Evaluator Guidelines specifically penalize “the use of automated tools (generative AI or otherwise) to produce, with little effort, a large number of pages that provide little or no added value to visitors, compared to other content available online on the same topic.”
The search engine therefore appears to be targeting sites that are entirely, or largely, designed by artificial intelligence and whose content is not intended to provide real added value for the reader. The integrated penalty criteria even seem to indicate that only the crudest uses, which involve almost no form of human intervention, will be disadvantaged. An AI used to produce a useful text should therefore not be subject to any penalty, for the time being.
