❝

Part of a series: Google and AI Content > The September 2026 Spam Update. In Issue #240 we covered who actually sees ChatGPT ads. This week, Google's new written definition of "low effort" and the sites that lost traffic.

Mert Erkal

Hi everyone,

The September spam update is still rolling out, and the forums are full of the same question: "Our traffic halved overnight. What happened?" Marie Haynes, who tracks every update closely, thinks this one targets AI content produced at scale. Many of the site owners hit by it have no idea where the problem is. Over the past two weeks, we've had site owners ask us the same thing.

Here's the twist. In the middle of the spam update, Google quietly changed two of its own documents and, in effect, wrote down the answer.

Grab a coffee.

🎯 THE THESIS

Google isn't punishing AI content. It's punishing unchecked mass production. Google now defines "low effort" in writing: pages generated automatically from feeds, and large amounts of text produced with AI without manual review. That the spam update shook exactly these kinds of sites in the same week may not be a coincidence. The problem isn't "we used AI." It's "we published what AI wrote and nobody read it."

📅 1. WHAT HAPPENED? THREE DATES, ONE STORY

24 September: Google started the September 2026 spam update, the fourth spam update of the year. It applies globally and in all languages, and the rollout takes about two weeks. Google didn't say what it targets and didn't add a new policy.

30 September: A second wave hit. Several site owners told Glenn Gabe they saw big drops on exactly this date. Search Engine Roundtable's report on the second wave shows almost every volatility tracker, from Semrush to Sistrix, flagging the same day.

1 October: Google updated two documents together:

  • Its guidance on AI-generated content now says it's critical to manually fact-check and review all AI-generated content for accuracy and trustworthiness before publishing. Right next to it sit two reminders: generative models don't retrieve information, they predict likely sequences of words, and their output can contain hallucinations.

  • The helpful, reliable, people-first content document gained two new sections: a definition of main content, and the four qualities raters look for in it.

Google's official explanation is modest: it aligned the documents with the slides it uses at developer events. But as Barry Schwartz points out in his Search Engine Land piece, the word "critical" reads less like advice and more like an instruction. I agree. When Google puts "critical" in a sentence, it wants us to read that sentence.

🧱 2. MAIN CONTENT AND FOUR QUALITIES: EFFORT, ORIGINALITY, SKILL, ACCURACY

Google defines main content as everything that directly serves the purpose of the page. It isn't only text. Tools like calculators, user reviews, information inside tabs and page titles all count.

Raters look at four things in it:

  • Effort: how much human work went into the content, or into the system that produced it. Google's high-effort examples include original analysis, translating a poem by hand and building a custom interactive map for a game. Its low-effort examples: auto-generating pages from a feed, and producing large amounts of text with AI without manual review.

  • Originality: whether the content offers information or a perspective that isn't already on other sites.

  • Skill: whether the content shows enough skill to satisfy the reader. Google notes that not everything needs expertise: someone describing their experience shovelling snow doesn't have to be an expert.

  • Accuracy: on topics that affect people's lives, like health and money, content has to be highly accurate and in line with expert consensus.

There's also a new warning about fake author profiles. Using an AI-generated portrait, an invented name or a fake title to make content look expert-written is, in Google's words, a form of deception. And deception makes a page untrustworthy both to readers and to Google's automated quality systems.

In her line-by-line video on the new document, Marie Haynes underlines an important point: these questions aren't a scorecard. Google's machine learning systems try to imitate the "helpful content" this document describes. Ticking the questions off one by one isn't enough. You have to look at the page as a whole.

🇹🇷 FROM THE FIELD: WE APPLIED THE SAME RULE TO OURSELVES

When I read Google's new "deceptive author" line, the first thing we did was check our own site. The About page on stradiji.com shows two people: me, and Emre Ercan, who has been on our team since 2013. Both with real names, real titles and real histories. Our blog bylines link to a real person too.

Then we looked at our process. At Stradiji, every piece of content we produce with AI passes three gates before it goes live:

  1. Brand voice: Is this my voice? Are the terms right?

  2. E-E-A-T checklist: A named author, sourced numbers, correct dates.

  3. Originality score: What does this piece say that isn't already on the internet?

If there's no firsthand data, we don't invent it. We leave a gap in the draft and ask. A real example: in the first draft of this very issue, the AI wrote an opening line saying the same question had "landed in my inbox this week." It sounded completely natural. There was no such data. The sentence bounced at the second gate and got deleted. That's exactly what Google means when it says manual review is "critical."

In Turkey, the number of sites publishing dozens of AI-written posts a week is growing fast, so I expect to see this update's effect in Turkish search more clearly over the coming weeks. If you work in a non-English market, I'd watch for the same.

🔍 3. IS YOUR DROP ABOUT THE SPAM UPDATE? FIVE SIGNS

Not every traffic loss comes from a spam update. AI answers reduce clicks for everyone, plugins can break URL structures, and servers can block AI bots. If two or more of these signs apply to you, take the spam update seriously:

  1. The dates match. In Search Console, the drop started around 25 to 27 September or around 30 September.

  2. The loss is concentrated in one section. Mass-produced blog, city or product-variation pages fell, while hand-written pages held.

  3. There's a manual action. Check "Security & Manual Actions" in Search Console.

  4. Whole folders are dropping out of the index. In the Pages report, "Crawled, currently not indexed" jumped suddenly.

  5. The authors aren't real. The site has author profiles with AI-generated photos and no clear identity.

In my view, spam updates have one key difference. With core updates, "wait for the next update" can sometimes work. Here it rarely does. As long as the problem content stays on the site, so does the signal.

🛠️ 4. WHAT TO DO

  • Take inventory. List the pages you mass-produced with AI. How many, when, and who checked them?

  • Prune. Remove or merge pages with no traffic and no links. Noindex the ones in between. As Marie Haynes puts it, if you have hundreds of pages that wouldn't exist if you weren't writing for search, you may need to remove most of them.

  • Add effort to what stays. Firsthand data, real experience, screenshots, original analysis.

  • Make your authors real. Real names, real photos, real histories.

  • Turn pre-publication review into a process. Don't leave it to one person's mood.

💡 THE ONE THING TO DO THIS WEEK

Open Search Console. In the Performance report, set the date range to the last three months and mark 24 September and 30 September. Then, in the Pages tab, sort URLs by click loss. If most of the loss comes from a section you mass-produced with AI, you've probably found your cause.

If the chart is hard to read, or you're not sure whether the problem is spam, AI answers or a technical error, you're not alone. That's why we built a four-step Traffic Loss Recovery Plan for sudden drops. Not just for this update, but for whenever you need help:

  1. Quick Look: We compare your Search Console data with update dates and score your 10 biggest losing pages. A two-page memo in 2 business days. $150, credited toward a Diagnosis.

  2. Diagnosis: We crawl the whole site, analyse the losing pages against Google's four criteria and list what to remove, merge or strengthen, in order. 5 business days. $500, credited toward Recovery.

  3. Recovery: We run the pruning and strengthening plan with you and measure it every month. From $1,500 a month, 3-month minimum.

  4. Content operations: Once you've recovered, we set up an AI-assisted, human-reviewed content process, or produce the content for you in several languages.

Details, pricing and the application form are on the Traffic Loss Recovery Plan page. The first step, Quick Look, gets you an answer in 2 business days.

📰 ALSO WORTH KNOWING THIS WEEK

  • Google is paying around 100 publishers for AI contributions. According to The Information, Google pays publishers whose content feeds AI Overviews, AI Mode and Gemini roughly a tenth of a percent of its ad revenue. One publisher gets more than $1 million a year, while small sites earn under $1,000 over several months. Details in Search Engine Roundtable's report.

  • Penske's AI Overviews lawsuit was dismissed. Judge Amit Mehta ruled there was no "content for traffic" agreement between publishers and Google: an expectation isn't a contract. Read the details of the dismissal.

  • The "AI Overview tracking parameter" was an extension. The st_source=ai_overview parameter appearing on links was first taken for a Google test. It turned out to come from a Chrome extension that may sell the data. If you see it in GA4, don't assume Google is testing.

🎤 CLOSING

Producing content with AI isn't a crime. I write this newsletter with AI too. The difference is whether a human reads what comes out, fixes it and puts their name on it. This week, Google wrote that down more clearly than ever before.

So, my question: on your team, who reads AI-generated content before it goes live?

And one request. If your traffic dropped around 25 to 27 September or 30 September, reply and tell me the date and which section of the site took the hit. No URLs needed. I'm building a picture of what this update actually hit outside the U.S., and right now nobody has it.

See you next week. Take care.

Mert Erkal

☕ WANT TO SUPPORT THIS NEWSLETTER?

I genuinely enjoy sharing what's moving in SEO and AI search with you every week. If you find the work useful, you can keep it going by buying me a coffee. ☕ Click 'Buy Me a Coffee' to contribute to the knowledge-sharing. Let's build more together.

👤 ABOUT MERT ERKAL

Mert Erkal is the founder of Stradiji. Since 2010, he has advised enterprise brands on SEO, generative engine optimization (GEO) and conversion optimization. Based in Istanbul, he works remotely with clients in the U.S., the U.K., Australia and Turkey. He writes the weekly newsletter The SEOs Diners Club.