Google does not penalise a post for being written with AI. It penalises thin pages made mainly to rank, whichever tool wrote them. If you have already published a lot of AI-assisted posts, the question that decides what you do next is not whether Google can tell which ones were AI. It is whether those pages were worth reading.
Here is Google’s own wording first, then a way to work through your posts one at a time, and one check in Search Console that tells you whether you have been handed a penalty or have simply lost ground.
Does Google penalise AI content?
Not for the AI. Google’s own guidance on the subject says that using generative AI tools or similar tools “to generate many pages without adding value for users may violate Google’s spam policy on scaled content abuse”. The trigger is the volume without the value. The tool is only the detail.
Google’s spam policies define the practice in one sentence, and it is worth reading twice.
Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users.
The tail of that policy sentence does more work than most readers notice. The practice “is typically focused on creating large amounts of unoriginal content that provides little to no value to users, no matter how it’s created.”
Those last four words settle most of the argument online. Google’s line is not machine against human. Three hundred thin posts written entirely by hired people break the same rule a generated set does. One AI-assisted post that a person checked, cut down and added to is not what the policy is aimed at.
Using the tools is ordinary here now. IMDA’s Singapore Digital Economy Report 2025, published on 6 October 2025, takes a comprehensive look at Singapore’s digital economy in 2024. It found that the share of Singapore small and medium enterprises using AI tripled, from 4.2% to 14.5%, and that AI-using firms mostly reach for off-the-shelf generative AI tools, 84% of them. Google is not telling small businesses to stop using what they have bought. It is asking for pages that help somebody.
One piece of older advice is now wrong, and it still circulates. In March 2024 Google said it had “enhanced our core ranking systems to show more helpful results using a variety of innovative signals and approaches”, and that there was “no longer one signal or system used to do this”. Helpfulness is not a separate switch you wait out. It sits in the main ranking.
What is scaled content abuse, and how many AI posts is too many?
Many pages, made mainly to rank, adding little for the reader. Google’s own examples are plain: using generative AI tools to generate many pages without adding value, stitching content from other pages without adding value, and “creating many pages where the content makes little or no sense to a reader but contains search keywords”.
There is no number in Google’s guidance. Neither the spam policies page nor Google’s page on generative AI content sets a limit: no percentage of AI writing that is too much, no post count, no recovery timetable. Any threshold on offer is a guess, because those two pages do not contain one.
So run Google’s test instead of a threshold. The four terms the rest of this piece uses are worth pinning down first.
- AI-generated content
- Text made wholly or mostly by a generative AI tool and published with little or no rewriting by a person.
- AI-assisted content
- A draft from an AI tool that a person then checked, cut, reordered and added to before publishing. Google applies the same rulebook to both.
- Scaled content abuse
- Google’s name for producing many pages mainly to manipulate search rankings, with little or no added value for readers. Its policy says the rule applies no matter how the content was created.
- Manual action
- A decision by a human reviewer at Google that pages on your site break its spam policies. It is shown in Search Console, and it is separate from how well your pages rank.
Is your traffic drop a penalty or a ranking problem?
Search Console answers that in one report. Google issues a manual action “when a human reviewer at Google has determined that pages on the site are not compliant with Google’s spam policies”. If your site has none, “you’ll see a green check mark and an appropriate message”.
Most of the time, an action means affected pages “will be ranked lower or omitted from search results without any visual indication to the user”. The fix carries a rule that catches people out: clear the problem on every affected page, because “fixing the issue on just some pages will not earn you a partial return to search results”. Then you select Request Review in the report.
If the report shows a manual action
Work through every affected page, not the easy ones. Check that Google can reach each one: no login, no paywall, nothing blocked by robots.txt or a noindex directive. When the fixes are done on all of them, use Request Review in the report and describe the exact problem, the steps you took and the result you saw.
If the report shows a green check mark
There is no decision to appeal, so there is nothing to request. What you have is a ranking problem, and it is fixed the slow way, by making the pages worth reading. If you have never opened that report, our free walkthrough of checking your own SEO in ten minutes takes you to it.
How to recover, one post at a time
Start with a list, not a rewrite. This is the order we would work in.
A note on the pictures: both of ours were made with an AI image model.
- Open the Search Console performance report and put your post list beside it, with the clicks and impressions for each page. An impression counts a link appearing; a click counts someone choosing it.
- For each post, answer two questions: was it published mainly to rank, and would it still be worth a reader’s time if nobody ever searched for it?
- Mark every post improve, merge or remove.
- Improve the posts whose topic is worth keeping but whose body is thin. Add what only you have.
- Merge the thin posts that answer the same question into one page that does the job properly, then point each old address at it with a redirect. Google treats a redirect as a strong signal that the destination should become the canonical page.
- Remove what is worth nothing. Google’s own advice for a site hosting such content is to exclude it from Search.
Two things in that list matter more than the order. A redirect is not a deletion. The old address keeps working and sends the reader to the page you kept.
And the improve step is the only one that adds something new, which is why the effort belongs there. That material is not on the internet. It is the example from your own work, the photograph from your own premises, the number from your own month, the name of the person who did it.
Google’s own self-assessment asks whether it is clear who created a piece, how it was produced and why it was created, and it puts the weight on the last of those three: content created “primarily to help people, content that is useful to visitors if they come to your site directly”. Its quality raters are trained on what Google calls E-E-A-T, though their ratings do not change how a page ranks. The practical test is simpler than the acronym. Could a reader tell that somebody who has actually done this wrote it?
If you would rather watch someone else work through the same question, this is the closest thing we found.
Watch on YouTube: Does Google Penalize AI Content? New SEO Case Study (2026), by Nathan Gotch.
We kept it because it is the one candidate that stays on this exact question and shows a site’s numbers either side. The others the search returned were about ranking in AI Overviews, or about how to write past a detector, which is the advice this piece argues against.
Does Google detect AI content, and do you have to say you used it?
On disclosure, Google’s guidance is measured. Its own words: “AI or automation disclosures are useful for content where someone might think ‘How was this created?’”, and they are worth adding “when it would be reasonably expected”. So a short note on a page where a reader would plausibly wonder is worth having.
On detection, neither of those two Google pages describes how it identifies AI text, and neither asks you to declare AI use on every page. What they do ask for is accuracy, quality and relevance, “especially when automatically generating the content”. Google judges the page, not the tool that drafted it, so running posts through a detector to see whether they pass is measuring the wrong thing.
What to do today
Pick your three thinnest AI-assisted posts. Put each one through Google’s own self-assessment questions in “Creating helpful, reliable, people-first content”, the page where Google asks creators to grade their own work. Then mark each of the three improve, merge or remove, and start with the one you can finish this week.
When you want the rule rather than the summary, read Google’s spam policies for Google web search. It is not a long document, and the scaled content abuse section is the part that applies to you.