Google penalises bad content, whoever wrote it. AI just happens to produce a lot of bad content.
Google’s position hasn’t changed since 2023: it judges quality and doesn’t care how the content was produced. What has changed is that the algorithm has become much better at spotting scaled content abuse, the mass production of pages with no value.
Below: exactly what gets penalised, what gets rewarded, and how to use AI without wrecking your organic traffic.
What actually gets penalised
The tool itself is irrelevant. What counts is the pattern:
| Pattern | What it means in practice |
|---|---|
| Volume without depth | Hundreds of pages covering variations of the same keyword and adding nothing |
| Zero originality | Recycling what the top ten results already say |
| No human editing | Publishing raw AI text with no fact-checking |
| No real author | No name, no bio, no proof of experience |
| Identical structure | The same template in every article, obvious from a mile off |
Sites that published hundreds of pages like these saw losses of 60-80% of their traffic. People called it an “AI penalty”. In reality it was a quality penalty, and AI simply made it easy to do damage at scale.
What gets rewarded
Over the same period, sites that published fewer, carefully edited articles with AI assistance gained traffic. The tool was the same. The difference was what people added on top of it:
- Original information: data, prices, results and measurements you won’t find anywhere else.
- Real experience: “we tried it and this is what happened”, in place of “according to experts”.
- A name and a face: an author with a bio, instead of “Editorial team”.
- Fact-checking: every number and date verified.
- Updates: content is no monument; it needs maintaining.
The real risk lies elsewhere
It lies in mistakes that sound right.
A language model will give you, with complete confidence, a wrong number, a wrong date, a source that doesn’t exist or outdated legislation. The text will read perfectly. And it will be wrong.
If you are a business that sells advice, a wrong number in one of your articles costs you more than rankings. It costs you credibility, and that is far more expensive.
How we use AI at WebMate
| AI does this well | AI does this dangerously |
|---|---|
| Article structure and outline | Numbers, dates, legislation |
| Variations of titles and meta | Claims about results |
| Summarising long texts | Sources and citations |
| Translation and adaptation | The business’s own “point of view” |
| Ideas and angles | Anything that needs experience it doesn’t have |
Our rule is simple: AI writes the draft, a person signs it off, and whatever they sign off, they have checked.
What about AI search?
The same content that survives Google’s updates is also the content that gets cited by ChatGPT, Perplexity and AI Overviews: specific, well structured, with original data and a clear source. More detail: Generative Engine Optimization (GEO).
If your content says nothing the model doesn’t already know, there is no reason for it to cite you. And if it doesn’t cite you, you don’t exist in the new search.
What to do tomorrow morning
| Step | Action |
|---|---|
| 1 | Review the articles you published over the past year: how many say something that isn’t already on ten other sites? |
| 2 | Add a real author with a bio. “Editorial team” doesn’t count as E-E-A-T. |
| 3 | Verify every number and date in anything AI has written. |
| 4 | Add what only you have: prices, results, photos, mistakes. |
| 5 | Fewer articles, better ones. Volume has stopped being a strategy. |
Need help with your content and organic traffic? Take a look at our SEO & website promotion service.
Frequently asked questions
Does Google penalise content written with AI?
Not in itself. Google’s official position is that it judges the quality of content, regardless of how it was produced. What gets penalised is mass-producing low-value content to manipulate rankings, known as scaled content abuse, whether a person or a machine wrote it.
Can Google tell that a text was written by AI?
It doesn’t matter whether it can. Google isn’t looking for an AI signature; it looks for quality signals: original information, experience, consistency, usefulness. A text without these gets demoted, whether a machine wrote it or a person in a hurry.
What is scaled content abuse?
Creating lots of pages with little or no added value, aimed at ranking for lots of keywords. Algorithm updates target it explicitly, and the losses on sites with hundreds of such pages were dramatic.
So should I avoid AI altogether?
Use it, as a tool. Leave the authorship to people. AI is excellent at structuring, summarising, suggesting and speeding things up. It is bad at knowing what happened on your own project, what price you quoted and what the client said.
What is the biggest risk of AI content?
Mistakes that sound right. A model will give you, with complete certainty, a wrong number, a wrong date or a source that doesn’t exist. If you publish it as is, you lose rankings and, worse, credibility.
How do I write content AI can’t reproduce?
With what isn’t on the internet: your own data, prices, project results, mistakes you made, photos, measurements. This is also the content that gets cited by AI search engines.
Sources: Google’s official guidance on AI-generated content (Google Search Central) and its spam policies (scaled content abuse). Last updated: July 2026.
Leave a comment