There is a persistent idea in SEO that Google has developed a particular dislike of AI-generated content.
According to this version of events, Google can detect that a machine was involved, take offence and quietly remove the page from the search results. The Google boogeyman arrives, eats the content and throws it into a digital hole.
It is a convenient explanation. It is also mostly nonsense.
Google does not need to hate AI content. It already has a perfectly good reason to demote weak content: it is weak.
That distinction matters. Many businesses are concentrating on whether Google can identify AI involvement when they should be asking whether the content was worth publishing in the first place.
This article develops the argument from my original YouTube video. If you prefer to watch rather than read, you can view it below.
The evidence does not support the panic
Ryan Law at Ahrefs recently published a study examining the apparent presence of AI-generated content in Google’s search results.
Ahrefs analysed pages appearing in the top ten results across 100,000 search results pages. Of the pages for which it had enough crawlable text, its detector classified 5.3% of top-three results as entirely AI-generated and 9% as containing at least 80% AI-generated content.
That does not prove those classifications were correct. Ahrefs explicitly acknowledges that AI detection is imperfect. Nor does the research prove that AI content performs particularly well.
In fact, the study found that greater estimated AI use correlated with lower indexation rates, lower rankings and fewer impressions. There could be several explanations for that relationship, including differences in website authority and content quality, so correlation should not be mistaken for an AI penalty.
The narrower finding is still useful: there was no obvious threshold at which Google appeared automatically to exclude content because a detector considered it AI-generated.
You can read the full Ahrefs study and its methodology.
Google’s own guidance is less ambiguous. It says that generative AI can be useful for research and for adding structure to original content. Using it to produce many pages without adding value, however, may violate Google’s policy against scaled content abuse.
In other words, the problem is not that a machine touched the words. The problem is producing content primarily to manipulate visibility rather than help anybody.
That should not come as a great surprise. A human is perfectly capable of producing useless content. The internet has provided ample evidence of that.
We have seen this before
AI content is often presented as an entirely new problem. It is not. It has simply made an old problem faster, cheaper and easier to disguise.
In the mid-2000s, article-spinning software allowed website owners to take an existing piece of content, replace words and phrases, and produce what was supposedly a new article. The results were usually dreadful, but they were cheap and scalable. That was enough to make them attractive to a certain kind of SEO practitioner.
The logic was familiar: more pages meant more opportunities to rank.
For a while, that worked often enough to encourage the behaviour. Google subsequently became better at identifying shallow, duplicated and manipulative content. Panda, introduced in 2011 and later incorporated into Google’s core ranking systems, was intended to help original, higher-quality content appear more prominently.
The lesson should have been that content needs substance.
Instead, parts of the industry concluded that the method of disguise simply needed to improve.
Generative AI is now being used in much the same way. The software is more sophisticated and the output is considerably more readable, but the underlying behaviour is often unchanged. Existing information is gathered, rearranged, repackaged and published, with little evidence that anyone involved had anything useful to add.
AI can still be a sophisticated word spinner
Used badly, AI is little more than a highly advanced word spinner.
It can process more material, produce smoother prose and imitate a wider range of styles than the crude tools of the past. This makes the result appear more original, but appearance is not the same as substance.
When the input contains no original thinking, first-hand experience, useful evidence or proprietary data, the finished article is unlikely to add much. It may be grammatically correct and neatly structured. It may even sound authoritative.
None of that means it deserves to exist.
This is where many organisations are getting AI content wrong. They are using the tool to avoid the expensive part of content creation: actually knowing something.
AI is being asked to replace expertise rather than help communicate it. That can produce a convincing imitation of useful content, but the imitation becomes increasingly fragile when every competitor has access to the same tools, the same public information and broadly similar prompts.
The result is a growing mass of competent-looking sameness.
Google is not the only audience
The discussion is often reduced to whether a page will rank. That is too narrow.
Even when generic content attracts traffic, it still has to do something useful for the business. It should build trust, demonstrate judgement or help a prospective customer understand why the business is credible. Ideally, it should move somebody closer to a sensible decision.
A generic summary of information already available on twenty competing websites rarely achieves much of that.
Traffic without relevance, confidence or intent is not a business outcome. It is an analytics event.
The same underlying problem matters as search incorporates more generative features. Google’s guidance for appearing in its AI search experiences still emphasises valuable, unique and non-commodity content rather than a separate collection of supposed GEO tricks.
Specific experience, evidence and original analysis give both readers and automated systems something identifiable to work with. Repeating the existing consensus in slightly different words does not.
That does not guarantee an AI citation or brand mention. Nothing does. But content that contributes no distinctive information gives a system very little reason to use or attribute it. An AI Search Audit should therefore examine how clearly a business is understood and evidenced, not merely count mentions.
AI is still useful
None of this means businesses should avoid AI. That would be an equally simplistic response.
AI can be useful for organising ideas, improving structure, tightening language and identifying gaps in an argument. It can turn a rough transcript into a readable article, help an expert explain something more clearly or suggest a better structure for a complicated subject.
It can also challenge assumptions, surface objections and remove repetition. These are sensible uses because the substance still comes from somewhere real.
The experience belongs to the person. The evidence belongs to the business. The judgement belongs to the expert.
AI helps with expression and production.
That is very different from asking it to manufacture expertise on demand.
Ask whether the content adds anything
The question is not whether AI wrote the content. The better question is whether the content adds anything.
Does it contain an identifiable point of view? Is the argument supported? Could the same article have appeared on any competitor’s website? Would a prospective customer learn anything about how this business thinks? Would anybody notice if the page disappeared?
Those questions are more uncomfortable than running content through an AI detector, but they are also more useful.
AI detectors are not reliable enough to be treated as definitive judgements about individual articles. Even if they were, they would still be answering the wrong question.
The presence of AI is not the strategic issue. The absence of value is.
Stop worrying about the boogeyman
Google is not waiting behind the curtain to punish every business that uses AI to edit a sentence. It is trying, imperfectly, to determine which results are useful and which exist mainly to capture search traffic.
That process is neither transparent nor perfectly consistent. Search never has been.
The sensible response is not panic. Nor is it to publish as much machine-generated content as possible before some imagined loophole closes.
Start with a useful observation, genuine experience, original evidence or data the business actually owns. Use AI to help shape and communicate it, then apply human judgement before publishing.
That is not an AI content strategy.
It is simply a content strategy with a newer tool involved.