Content marketing

AI content: which pages endanger your visibility and how we use AI

Key takeaways
  • In May 2026, Lily Ray analyzed over 220 websites that use AI content tools. 54% lost at least 30% of their traffic compared to their peak.
  • According to its spam policies, Google evaluates the purpose of a page. Mass-produced pages without value count as spam, regardless of whether a human or AI wrote them.
  • Templates at large scale are risky: comparisons and lists, short explainer pages and pages that differ only in the city name.
  • For our copy, the rule is: AI helps with layout and structure. Every sentence is an activity, a piece of evidence or a fact with a source.

With AI, hundreds of pages can now be created in hours. In May 2026, SEO analyst Lily Ray investigated what happens to websites that work this way.

In this article, we summarize what Ray found and what Google says about it. Then we show which patterns we changed on our own site and the rules by which we use AI in writing.

Lily Ray analyzed over 220 websites with AI content

Ray observed websites that AI content providers had publicly named as customers. The basis is traffic estimates from Ahrefs, cross-checked against the Sistrix visibility index, as of May 2026.

Loss compared to peak organic traffic Share of websites
at least 30% 54 %
at least 50% 39 %
at least 75% 22 %

According to Ray, the pattern is often the same: first the number of pages rises, then the traffic, and after a few months it drops significantly. Ray calls this pattern “Mount AI.”

Ray herself adds a caveat: these are third-party estimates. She shows a correlation, not proof that the tools caused the losses.

According to Ray, from around January 20, 2026, at least 40 websites lost between 40% and 95% of their organic traffic by April. Self-promotional best-of lists were frequently affected. The loss often hit the entire blog or the entire domain. Google has not confirmed an update for this period.

Google evaluates the purpose of a page, regardless of the tool

Google's spam policies call this pattern “scaled content abuse.” It refers to many pages created primarily to manipulate rankings that offer readers little. According to Google, this applies regardless of whether a human or an AI created them.

The same applies to AI search. In its guide from July 2026, Google writes that the AI features of Search are based on the ranking and quality systems of regular search. Google explicitly classifies building separate pages for every variant of a question in order to influence rankings or AI answers as spam.

In the same guide, Google recommends creating content based on your own knowledge. It advises against copy that merely repeats what others have already said or what an AI model can easily generate.

Templates produced at large scale are risky

Ray describes eight page types that keep appearing on the affected websites. None of them is bad in itself. We sort them into three groups and state what such a page must carry with us before it goes online.

Group Page types according to Ray What the page must carry with us
Comparisons and lists “A vs. B,” “The best X for Y,” best-of lists featuring your own offering, alternatives to competitors disclosed own role, traceable criteria, own experience with the offerings
Short explainer pages Glossaries following the pattern “What is X?”, one page per question bundled questions on one page, own examples or data
Mass in breadth Template-based location and language pages, topics unrelated to your own offering own information per location or language, clear connection to what the company offers

How we use AI in writing

AI helps us work faster. We use the time saved to check claims and find evidence. Since October 2026, fixed rules apply to all copy on unitedads.de:

  • AI builds; it does not supply content. We use it for layout, structure and revision. Claims, figures and examples come from our work or from sources.
  • Every sentence has an origin. It describes an activity from our service pages, states a result from a published case study or a fact with a source. Otherwise it is deleted.
  • If evidence is missing, a gap remains visible. The draft then contains a placeholder in square brackets. We do not invent figures, quotes or customer cases.
  • Figures appear in the text with source and date. We label our own queries with date and data source.
  • A checking script runs before every publication. It finds empty phrases, banned words and promises we cannot keep.
  • The author signs off. A text that AI has revised remains a draft until the responsible person has reviewed it.

We do the work that makes the difference. For copy, that means: we check every claim before it goes online.

We review your content by the same rules

In search engine optimization for our clients, we apply the same standards:

  • Measure results: We measure which pages bring inquiries and revenue. That is where we shift the effort.
  • Answer questions: We build pages that answer your customers' questions with your knowledge, your data and your examples.
  • Fix the technology: We fix whatever keeps Google from reading and evaluating your pages, from load times to indexing.

How we approach AI search is described on our page on GEO. You can find the fundamentals of search engine optimization under SEO.

Sources
  1. It Works Until It Doesn't: AI Content Strategies That Backfire · Lily Ray, May 13, 2026
  2. Spam policies for Google web search · Google Search Central, updated August 28, 2026
  3. Optimizing your website for generative AI features on Google Search · Google Search Central, updated July 10, 2026
Initial consultation

Let's find out where your budget can achieve more.

A personal conversation in which we assess your starting point together.

Julia Bühler

Julia Bühler · Solution Manager

Your inquiry comes directly to me. Then you can pick a time right away.

Schedule a consultation

In the first conversation, we identify where the biggest profit potential in your online marketing lies.

In the next step, you choose your appointment.