Chances are you have already read AI-generated text today without noticing. That is not inherently bad. But it raises a central question for businesses: how do you use AI for your content without it becoming generic, interchangeable and ultimately worthless? This article explains what Google actually asks for, what a reliable workflow looks like and which transparency rules apply in Switzerland and the EU.
The problem with "AI slop"
"AI slop" has become the common term for mass-produced, superficial content that reads the same across many websites. You can spot it by the same formulaic phrases, a lack of depth, missing evidence and no original perspective.
Google targets the pattern, not the technology. On 5 March 2024, alongside a core update, Google introduced new spam policies, including "scaled content abuse": many pages generated for the primary purpose of manipulating search rankings rather than helping users. The policy explicitly applies no matter whether content is produced through automation, human effort or a combination of both. Policies against expired domain abuse and site reputation abuse were added at the same time.
Google's guidance on generative AI content says accordingly that using generative AI tools to generate many pages without adding value for users may violate this policy. Anyone producing content at volume therefore risks losing visibility not because of AI, but because of missing value.
AI as a tool, not an author
The key lies in defining roles. Google itself notes that generative AI can be particularly useful when researching a topic and to add structure to original content. In practice, AI works well for:
research preparation and briefings whose facts are then checked
outlines and structuring complex topics
first drafts that a subject expert then reworks
suggestions for titles, meta descriptions and alt text
translations as a basis for revision by a fluent speaker
AI is a poor substitute for original expertise and opinions, personal experience and case studies, current industry insight and nuanced argument. The formula: AI for efficiency, humans for value.
How AI-assisted content works in practice
In our agency, we use a three-step process. Step 1: AI supports research and creates a briefing with an outline and pointers to sources. Step 2: a subject expert writes the content based on the briefing, with their own expertise, their own examples and a clear point of view. Step 3: AI helps with the finishing touches, such as metadata, readability and structured data.
What matters most is the review between the steps. Google points out that generative models do not retrieve facts but predict a likely sequence of words, and that it is critical to manually fact-check and review all AI-generated content before publishing. According to Google, this also applies to metadata such as title elements, meta descriptions, structured data and image alt text, which can appear in search results. A simple editorial checklist:
Every figure, date and quotation is backed by a source that someone has actually opened.
The text contains at least one thing only you can offer: your own experience, your own data or a reasoned assessment.
A named person with subject expertise has read the text and takes responsibility for it.
Metadata and structured data match the visible content.
Online shops face stricter rules: according to the same guidance, Google Merchant Center requires AI-generated images to contain IPTC metadata with the DigitalSourceType "TrainedAlgorithmicMedia", and AI-generated product data such as titles and descriptions must be specified separately and labelled as AI-generated.
What Google actually evaluates
Google's position is clear: what counts is not whether AI was involved, but whether content is helpful, reliable and created for people. Google suggests asking three questions:
Who created the content? It should be self-evident to visitors who authored it. Google explicitly warns against deceptive authorship information, and treats fabricated creator profiles, such as AI-generated headshots or made-up names, as deception.
How was it created? According to Google, AI or automation disclosures are useful where readers might wonder how a piece was made. If you generate content automatically, explain how it was created.
Why was it created? Using automation, including AI, primarily to manipulate search rankings violates Google's spam policies.
E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is often misunderstood. Google states that E-E-A-T itself is not a specific ranking factor. Search quality raters assess the quality of results, but they have no control over how pages rank. Of the four aspects, Google considers trust the most important. More useful than the acronym are Google's self-assessment questions, for example whether content provides original information, reporting, research or analysis, or whether it mainly summarises what others have said without adding much value. In practice: show experience through your own examples, demonstrate expertise through depth and precision, and build trust through sources, clear authorship and accurate information. How such content also becomes visible in AI answers is covered in our article on Generative Engine Optimization.
Transparency and the law: Switzerland and the EU
The Swiss Federal Data Protection and Information Commissioner (FDPIC) states that, regardless of future regulation, the Federal Act on Data Protection, in force since 1 September 2023, is directly applicable to AI-supported data processing. This includes transparency about the purpose, functionality and data sources of AI-based processing. Users of chatbots also have a right to know whether they are speaking or corresponding with a machine, and deepfakes of identifiable people must always be clearly indicated. For content production, the main consequence is simple: do not enter customer data or confidential information into AI tools before checking how the provider processes it.
In the EU, the AI Act applies in general from 2 August 2026 under Article 113. Article 50(4) requires deployers of an AI system that generates or manipulates text published with the purpose of informing the public on matters of public interest to disclose that the text has been artificially generated or manipulated. The obligation does not apply where the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for its publication. Whether and how this affects a Swiss company publishing content for an audience in the EU should be assessed case by case with legal advice. Either way, a documented editorial process is the best foundation.
Common mistakes
publishing AI drafts without fact-checking because they sound fluent and plausible
generating many near-identical pages for keyword variants or towns that differ by only a few words
publishing articles without an identifiable author
copying confidential documents or personal data into public AI tools
letting metadata be generated automatically and never checking it
Sidora AG develops content workflows that combine AI efficiency with editorial quality and aligns them with SEO and GEO optimisation and content marketing planning. If you want to integrate AI tools securely into your own processes, our AI solutions cover the technical groundwork.
