Short answer: no, AI-assisted content is not penalised for being AI-assisted. What gets penalised is unhelpful content, and AI makes it very easy to produce that at scale.
The distinction matters, and most advice on this gets it wrong in one direction or the other.
What Google actually says
Google's stated position is that it rewards helpful, reliable content however it is produced. There is no penalty for using AI, and no requirement to disclose it.
What Google targets is scaled content abuse: producing large amounts of content primarily to manipulate rankings rather than to help anyone. That policy predates the current AI wave and applies equally to content written by humans, spun by software, or generated by a model.
So the question is not "was this AI-written." It is "would a person be glad they landed here."
Where sites actually get hurt
Not by AI itself, but by patterns AI makes cheap:
Volume with no substance. Two hundred pages published in a month, each covering a keyword and saying nothing a reader could not get from the first result. This is the pattern that triggers the policy, and it is the most common way sites get hit.
Content with no first-hand anything. No original data, no experience, no opinion, no example the model could not have produced without you. It reads as interchangeable because it is.
Unverified claims. Statistics and citations that a model produced and nobody checked. This is a credibility problem before it is a ranking problem.
Thin pages targeting near-identical keywords. Five pages for five variants of one query, splitting their own rankings — keyword cannibalisation rather than coverage.
Publishing velocity that does not match the site. Going from a handful of pages to a hundred in a week is a pattern, and patterns get looked at.
What actually works
Use AI for the draft, not the substance. The model can structure and phrase. The information worth reading has to come from you — your data, your experience, your customers, your test.
Add something only you have. One original example, one screenshot of your own result, one number from your own data. This is the single biggest differentiator between content that ranks and content that does not, and it is exactly what a model cannot supply.
Verify every fact and every citation. Models invent sources that look entirely plausible.
Publish at a sustainable rate. Ten good pages a month beats a hundred thin ones, and does not look like abuse.
Write for one query per page. If two pages could answer the same search, merge them.
Edit properly. Write the opening and closing yourself — they carry the voice. Cut the tricolons, the "in today's landscape", the closing paragraph that restates the opening.
The measurable difference
You do not need to guess whether your content is helpful. The behaviour tells you.
Watch time on page, scroll depth, and — most usefully — whether people return to the search results after landing. A page that sends people straight back is failing regardless of what wrote it.
Google Search Console will show you which pages get impressions but no clicks: that is a title-and-description problem. And which get clicks but rank position 8–20: that is content that is nearly good enough, and the cheapest improvement available to you.
Do you need to disclose AI use?
Not for Google's purposes. There is no ranking requirement.
You may want to for your readers' sake, depending on your field and your audience — and in some regulated contexts you may be required to. That is an editorial and legal question, not an SEO one.
Common questions
Does Google penalise AI content? No, not for being AI-generated. It penalises unhelpful content and scaled abuse, whoever or whatever produced it.
Will AI content rank? Yes, if it is genuinely useful. It will not rank on volume alone, and neither would human content.
Should I disclose that I used AI? Not required for SEO. It may matter for your audience or your industry's rules.
How much AI content is too much? There is no number. The risk is thin content published fast, not AI as such. Ten substantial pages a month is very different from a hundred thin ones.
Can Google detect AI writing? It does not need to. It measures whether content is helpful, which is both easier and more relevant than detecting authorship. Third-party detectors are unreliable anyway.
What is scaled content abuse? Producing large volumes of content primarily to manipulate rankings rather than to help readers.
What is the biggest mistake with AI content? Publishing something that contains nothing the reader could not get elsewhere. Add your own data, example or experience, or do not publish it.