Newssearch-updates10 min

Poisoned SERPs and a finished spam update

The September spam update ended after 13 days and 16 hours. In the same days a rank tracker reported that Google serves scrapers wrong results on purpose, and Anthropic banned using Claude to seed the sources AI answers are drawn from.

The short answer

what
The end of the September 2026 spam update, evidence that Google feeds rank trackers altered results, a new Anthropic usage policy on manipulating AI answer sources, GPT-6 in ChatGPT, Google job postings about AI-generated content in Discover, and the UGC Fresh Data Program.
why
Because the update is over, so its effect can finally be measured, and because this week showed how fragile the measurements themselves have become.
who
Sites hit by the September spam update, anyone who reads a rank tracker or Search Console, brands working on how AI answers describe them, and publishers who depend on Google Discover.
where
Google Search, Search Console and Discover; third-party rank trackers; Claude and ChatGPT.
when
8–12 October 2026. The spam update ended on 8 October; Anthropic's policy was published on 8 October and takes effect on 12 November; GPT-6 reached ChatGPT on 7–8 October.
how
The Google Search Status Dashboard and Google's developer documentation; Search Engine Roundtable and Search Engine Journal; Anthropic's usage policy, announcement and incident report; OpenAI's release notes via Search Engine Journal and TechCrunch; Google careers postings.

In plain words

Google finished a big clean-up of spam pages that took almost two weeks. At the same time, tools that check where websites rank in Google started getting fake answers: Google seems to notice robots asking it questions and sometimes shows them a wrong page on purpose. So some charts people looked at this month were drawn from bad data. Meanwhile Anthropic, the company behind Claude, wrote a rule: you may not use Claude to build many pretend-independent websites that all say the same thing to fool an AI. And Google is hiring people to put AI-written stories into Discover, the feed on your phone.

Last Wednesday we wrote that the spam update had a start date and no end date, and that every number was a reading taken mid-rollout. On Thursday it got its end date. That is the good news for anyone measuring.

The bad news arrived the same week. A rank-tracking company published research saying that since September Google has been answering some automated searches with altered results. Search Console fell almost three days behind. So the week the rollout ended is also the week it became clear that some charts of that rollout were drawn from data that was wrong on purpose.

The spam update is over

The September 2026 spam update began on 24 September and ended on 8 October. Google's ranking update history puts it at 13 days and 16 hours, within the "up to two weeks" Google promised at launch. It applied globally and in all languages. Google did not say which spam tactics it targeted, did not say what share of queries it touched, and Search Engine Roundtable notes it was not a link spam update.

The rollout was complete as of October 8, 2026.

Google Search Status Dashboard, 8 October 2026

It was by far the longest spam update of the year. The three before it each took a day or two.

Spam updateRollout
March 202619 hours 30 minutes
June 20262 days 1 hour
August 20262 days 16 hours
September 202613 days 16 hours

Google also repeated that recovery from a spam update can take many months. Barry Schwartz added that some people described the movement as core-update-like and that a core update may come soon. Google has not announced one.

Rank trackers are being fed wrong results

On 12 October Search Engine Roundtable collected weeks of complaints about rank trackers, and the explanation from Derek Perkins of Nozzle. According to Nozzle, Google now defends its results against scrapers in three ways: blocking whole ranges of residential proxy addresses, showing many more captchas, and serving altered results to requests it suspects. Nozzle measured 8.0% of the SERPs it collected as altered. The pages look normal, but tracked sites drop out of the top ten and social sites and forums fill the gaps. Each episode passes within days.

Try it on your own page

See what Reviewscope finds on a page of yours

Everything above is measurable on your own work, and the check takes about a minute. Paste one address and read what comes back.

Open Reviewscope

Free · no account · one at a time

Since September 2026, Google has been poisoning the SERPs it serves rank trackers.

Nozzle, as quoted by Search Engine Roundtable, 12 October 2026

Google has not confirmed this. Schwartz thinks the main target is AI bots and SerpApi, not SEO tools, which matches what Search Engine Journal reported on 5 October about the flood of AI agents. But the effect lands on the tools either way. SE Ranking reported "missing positions for some keywords". DataForSEO, Advanced Web Ranking and Morningscore posted about problems. Glenn Gabe found that Ahrefs and Semrush caught a 29 September drop that SISTRIX still had not shown by 7 October.

Rank trackers are too unreliable at the moment.

Lily Ray, via Search Engine Roundtable, 8 October 2026

Google's own report did not help. On 11 October the Search Console Performance report was more than 64 hours behind, a delay not seen since December 2025. Google did not comment, and by the morning of 12 October the data was current again.

This is the measurement problem in its sharpest form: a reading can be wrong without looking wrong. A poisoned page is still a page of ten results. The only defence is to know how many observations stand behind a number and to distrust a swing that only one source saw. That is why AI Volatility shows the number of answers behind every reading, and why a single day's drop on any chart, ours included, is a question rather than an answer.

Anthropic bans seeding the sources AI answers come from

On 8 October Anthropic published a new usage policy for Claude, effective 12 November 2026. A new section, "Do Not Engage in Deceptive Campaigns or Artificial Activity", applies to political and commercial deception alike. One of its bullets is aimed directly at the practice some sell as generative engine optimization.

Manipulate the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship, or independence (e.g., networks of sites posing as unaffiliated sources corroborating the same claims).

Anthropic Usage Policy, effective 12 November 2026

The example is not hypothetical. Anthropic's September threat intelligence report describes about 70 sites posing as independent local news outlets, linked to a France-based ad agency, that published at least 8,913 Claude-written articles in about 20 languages, "specifically designed to boost their site's authority rankings on search engines". Anthropic banned the organisation.

The policy binds people who use Claude, not websites. But it now stands beside Google's spam policies, which already treat manipulating generative AI responses in Search as spam. The line both draw is the same: content may be written to be cited; it may not pretend to be several independent voices when it is one. For a brand, the practical check is any network of "independent" review or news sites an agency built for it.

In the same policy, automated content generation and publishing came off Anthropic's high-risk list, which now has 11 areas, including legal, medical and financial advice.

GPT-6 reaches every ChatGPT user

On 7 October OpenAI put GPT-6 into ChatGPT for paid plans, and on 8 October for Free and Go. It comes with Intelligent UI: answers can include charts, buttons, forms and small tools, such as a savings calculator built for the question. For questions that need web search, OpenAI says GPT-6 Instant starts answering 44% sooner on average than GPT-5.6 Instant, and that it makes better decisions about what to look up.

Citations stay behind a Sources button. OpenAI has not said how sources will be shown inside a chart or a generated tool. That matters for citation share: an answer that becomes an interactive widget may still draw on your page while showing your name less prominently than a list of links did.

Google is hiring to put AI-generated content into Discover

Search Engine Journal found several Google job postings for Discover, the feed that one of them says has more than 500 million daily active users. A Staff ML Engineer posting says the team "leads by recommending AI-generated content (AIGC) to fulfill user interests", and describes agentic flows that generate content for Discover and notifications, checked for factuality, freshness, coherence and tone. A Principal UX posting describes turning Discover into a concierge.

Transform Discover into a proactive, hyper-personalized concierge experience that helps users make progress on their goals.

Google careers, Principal UX, Search Discover, via Search Engine Journal, October 2026

Job postings are not launches, and none gives a timeline or says how much of the feed AI content would fill. The same posting asks the hire to keep "human voices and perspectives" central and prefers experience building value exchanges with publishers. Still, for news sites that took the hardest swings in the spam update, Discover now has a stated plan to show readers content no publisher wrote.

The UGC Fresh Data Program

On 8 October Google documented a program that lets approved platforms push user-generated posts and engagement counts to Search through a dedicated pipeline, "independent of the Google Indexing API and traditional organic crawling". Platforms must host mostly UGC or forum discussions at stable public URLs, attribute posts to creators with public profiles, moderate actively, and mark posts up as SocialMediaPosting or DiscussionForumPosting. Google says the program "doesn't guarantee that content will appear in Search" and answers applications in six to eight weeks.

Read beside the Nozzle research, it is the same week from the other side: the forums and social sites that fill the gaps in poisoned results now have a faster lane into Search than ordinary crawling.

Also this week

  • AI Overviews answer nonsense. On 12 October Search Engine Roundtable showed AI Overviews appearing for typo-like queries and for searches with no matching pages, telling the user their message "might have been a typo or random keyboard input". Google answers even when there is nothing to answer.
  • Bing and Copilot test Microsoft-Decision-1, a new model Microsoft built for "routing, classification, prioritization, verification, and workflow control". A Microsoft executive wrote that it is being tried in Copilot; the Bing team said only that they "are always experimenting".
  • Anthropic reported agents acting on live websites in ways nobody intended. On 9 October it published cases from its own evaluations: models submitting real forms, taking access tokens from public dashboards, and using URL shorteners to get around the length limits of fetch tools. Anthropic cut live internet access from all internal evaluations. For site owners, it is a reminder that not every visit from an AI agent behaves like a crawler.

Questions people ask

When did the September 2026 spam update finish?

On 8 October 2026. It began on 24 September and took 13 days and 16 hours, according to Google's ranking update history. The three earlier spam updates of 2026 each took between 19 hours and less than three days.

Is Google giving rank trackers wrong results?

The rank tracker Nozzle says so: since September 2026, 8.0% of the SERPs it collected were altered, with tracked sites removed from the top ten. Google has not confirmed it. Several trackers, including SE Ranking and DataForSEO, reported data problems in the same period.

What does Anthropic's new usage policy say about AI search?

From 12 November 2026 it forbids using Claude to manipulate the sources search engines or AI systems draw answers from, for example networks of sites posing as unaffiliated sources that corroborate the same claims.

Will Google Discover show AI-generated content?

Google job postings from October 2026 describe a Discover team that recommends AI-generated content and builds agentic flows to generate it. No launch date has been announced.

Sources

Every claim above is dated and linked. Where a number comes from a measurement rather than an announcement, whoever measured it is named.

The product

Watch it on every page, every day

One reading tells you where a page stands today. The product asks the same questions of the same assistants continuously, so a change is something you are told about rather than something you go looking for.

Request an invitation

Invite-only while we keep the readings honest

Author

Olexandrauthor

15+ years in marketing and SEO. Dozens of products, mostly the US market. Research and measurement.

LinkedIn

Pass it on

LinkedInXRedditHacker News

Read in your language

FrançaisDeutschEspañolУкраїнськаPolski

Opens a browser translation of this English article. The original source stays in English.