Google does not penalize AI writing. It penalizes pages that exist to rank and do nothing else. That distinction decides whether your content program compounds or quietly stops working, and most of the posts that lost traffic this year lost it for the same reason: nothing happened between the model's output and the publish button.
What is Google actually penalizing?
The offence has a name, and reading it closely saves a lot of money. Google's spam policies define it as scale without value, not authorship.
Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users.
Two points carry the weight. First, the test is primary purpose: if a page was built to move a ranking, it qualifies. Second, the policy says plainly that the method of creation does not matter, which takes the tool out of the question entirely. A page a person wrote and left thin is the same offence as a page produced in bulk. The policy page was last updated on August 28, 2026, so this is the current wording, not a forecast.
Does edited AI content still rank?
Semrush compared more than 20,000 pages and concluded that content quality — not whether a model was involved — was the strongest predictor of rankings. Ahrefs surveyed 879 marketers and found that sites using AI grew organic traffic about 5% faster than sites that did not, while only 14% of those same marketers thought model output was better than human writing. Read the two together: AI is a throughput gain, not a quality gain, and throughput only pays if the quality comes from somewhere else.
Why does unedited AI content fade after about three months?
SE Ranking published roughly 2,000 AI-generated articles over 16 months and watched most of them slide sharply in rankings after about three months. The pattern matters more than the number. A new page gets a provisional read: it is indexed, tested against real queries, then re-scored once Google has behaviour data and a fuller view of the site. A page with nothing original in it survives the first read and fails the second. Publish in bulk and you get a flattering first month and a bill in the fourth.
The three-month lag is why volume feels like it is working
Bulk publishing looks successful for exactly as long as it takes Google to re-score the pages. Judge a content program on a six-month window, never a thirty-day one.
What does editorial work mean in practice?
Not proofreading. Editing is where the things a model cannot have get added: your numbers, your constraints, your customers' objections, and the thing you say out loud but never write down.
- Add one fact only your business has — a number, a job you quoted, a mistake you made.
- Cut every sentence that would survive unchanged on a competitor's site.
- Check every statistic against the original source, not the summary.
- Name the trade-off the draft avoided.
- Give it a real byline, with someone accountable for it.
- Rewrite the opening until it answers the question in one sentence.
How much should you publish?
Fewer posts than your tool can produce, and more than you publish by hand today. The honest answer is: as many as you can put real editorial hours behind. For most of the small teams we work with, that is two to four posts a month, each carrying something the model could not have known. Your ceiling is your editing capacity, not your generation capacity — and that was true before AI. AI only moved the generation cost to near zero and left the editing cost where it was.
We stopped counting posts per month and started counting posts we would defend in a meeting. The second number is smaller, and it is the one that moves anything.
What about the archive you already published?
If you shipped a batch of thin pages, the fix is subtraction before addition. Pull the list of pages that earned impressions but no clicks over the last six months. Merge the near-duplicates into one page worth reading, add the specifics, and remove what cannot be saved. Consolidating is a faster way back than publishing over the top of it, because the pattern Google is scoring is site-wide, not page by page.
Want the archive audited before you publish again?
Hamed Azimi
Founder, NexusWave
Hamed Azimi is the founder of NexusWave Technologies, a Vancouver-based AI-native digital agency and product studio.
We write these guides from real client work — GEO, SEO, and builds we've shipped. If a piece raised a question about your own site, we're happy to look at it.
- Straight answers, no retainer required
- We reply in your language — EN · فارسی
- Based in Canada, working worldwide
