Google has already put SAFE into production, a new AI-based system designed to identify massively generated spam content. This is not an update announced for the future: it's active now, and it's likely part of the September 2026 Google Spam Update.
Those who used AI to produce large volumes of text without real quality control are the most exposed. It's not about penalizing AI as such, but content that doesn't add value: repetitive texts, lacking original perspective, optimized for volume rather than usefulness.
The concrete move to make now is an audit of AI content published in recent months: check which pages lost traffic after the update, evaluate the informational depth of each piece, and decide what to rewrite, consolidate, or remove. Waiting for the data to worsen is not a strategy.
SAFE: what it is and why it's already in play
SAFE is the name of the new spam detection system that Google has developed and—importantly—already deployed in its ranking systems. The news, reported by Search Engine Journal , points to SAFE as almost certainly one of the technical engines running the September 2026 Google Spam Update.
In two sentences: SAFE is an AI classifier that analyzes content published on the web and evaluates it based on spam signals. It doesn't just look at keywords or links, but the overall pattern of the content, looking for the typical characteristics of low-quality automated production.
The distinction that matters is this: SAFE doesn't penalize those who use AI to write. It targets those who use AI to fill pages without bringing anything useful to the reader.
The profile of who is at risk—and who isn't
Not every site using AI content is in the same boat. The real danger is locked into specific types of profiles:
- High volume, zero review. Anyone who has automated the production of hundreds of articles without substantial human editing.
- Interchangeable content. Pages that cover the same topic almost identically, with minimal keyword variations.
- Lack of original perspective. Content that pulls together info already found elsewhere without bringing a new angle, first-hand data, or direct experience.
- Optimization for the crawler, not the reader. Structures built to satisfy an algorithm, not to answer a real question.
Those who instead use AI as an acceleration tool — for the draft, the outline, the research — and then invest time in review, deeper exploration, and differentiation, are in a much stronger position. This is even more true when considering how Google AI Mode is already changing how organic clicks are distributed : less total traffic means every spot you move up has to be backed by content that really holds up under quality checks.
What you need to do a serious audit right now
A post-SAFE audit isn't a quick once-over. It takes a real strategy. Here are the practical steps:
- Extract pages with dropping traffic. Compare Google Search Console data from the weeks before and after September 25, 2026. Pages that have lost impressions or clicks are the first sign.
- Evaluate information density. For each page: does it answer a specific question? Does it add something not already found in the top three search results? If the answer is no, the page is a candidate for review.
- Consolidate instead of deleting. Often it is more effective to combine five weak articles into one solid piece, using a 301 redirect, rather than deleting and losing any acquired link signals.
- Distinguish between content to rewrite and content to remove. If a page hasn't gotten any traffic in over six months and doesn't have any incoming links, deleting it is usually the cleanest move.
If you are working on a site with significant volumes of AI content, this audit ties into what we described in the analysis on Google Spam Update of September 2026 : the two phenomena—the update and SAFE—work together.
The link with visibility in AI answers
There's a second level of impact that goes beyond traditional ranking. Content flagged as spam by SAFE not only loses positions in classic SERPs: it also loses the chance to be cited in AI search systems like Google AI Overview or ChatGPT.
We already looked at how traditional SEO rank metrics no longer capture real visibility in an ecosystem where AI answers intermediate search. Content that Google classifies as spam will never enter the corpus that feeds these answers. And regaining that visibility is slower than recovering a ranking.
For anyone looking to figure out how to systematically map the sources driving AI answers in their niche, the starting point is the analysis of AI citations and competitor sources : it helps figure out where you stand today and where you risk disappearing.
The mistake not to make when reading this news
The most common mistake in these cases is reacting with a mass cleanup: deleting all published AI content, blocking crawlers on entire sections of the site, freezing editorial production. It is an emotional response, not a strategic one.
SAFE is not an anti-AI filter. It's an anti-spam filter. The difference is substantial. Blocking AI content production doesn't solve the problem if the content already published is weak. And it doesn't provide any advantage if the content you were creating was already high quality.
The variable that matters is the quality perceived by the reader, not the tool used to produce the text. This is the criteria by which Google has always stated it evaluates content, and SAFE seems to be the technical tool that makes it more operational.
For anyone who wants a fresh look at how to shift SEO budgets around right now, we dive deep into it over at metrics, budget and SEO priorities for 2027 . The starting point to find your way around the whole cluster remains the section dedicated to SEO and AI visibility on SHM Studio.
Related articles
Discover more articles exploring similar topics, selected to offer you a more complete and stimulating perspective. Each piece of content is carefully chosen to enrich your experience.
