Spam update ends ๐Ÿงน,Claude bans AI answer manipulation ๐Ÿšซ,AI Overviews on brand searches ๐Ÿ“Š

Issue 12, 12 October 2026

What changed in search and AI answers, for marketing teams, in five minutes.

Headlines & Launches

Google's September spam update is over (6 minute read)

Google confirmed on 8 October that its September 2026 spam update finished rolling out, taking just under 14 days, close to the two week estimate it gave at the start. If your traffic moved between 24 September and 8 October, compare Search Console clicks and impressions across that window before blaming anything else, since real spam update effects tend to persist across consecutive days rather than spike and vanish.

Anthropic bans using Claude to fake out AI answer sources (5 minute read)

Anthropic's usage policy, updated 8 October and taking effect 12 November, now explicitly bars using Claude to seed search engines and AI systems with content that hides who actually wrote or published it, naming networks of sites that pose as independent sources backing the same claims. It also bans building or selling tools for that purpose, though it does not say how violations get caught. For marketing teams, it is a signal that the kind of fake authority networks sometimes used to influence AI answers are now a policy violation with a real penalty, losing access to Claude.

Bing tests local pack without borders (2 minute read)

Microsoft is testing a version of Bing's local pack results with the usual outline border removed, spotted by Khushal Bherwani and confirmed by Barry Schwartz. It is a cosmetic test with no confirmed rollout date, but if your business relies on local pack visibility it is worth a quick look at how your listing appears on Bing right now.

Deep Dives & Analysis

AI Overviews now show up on most big brand name searches (7 minute read)

Ahrefs tracked Google's US results for 100 global brand names weekly from July to September and found AI Overviews on just 5 of 98 searches on 1 July, rising to 63 of 100 by 30 September, with a separate broader crawl putting the figure at 83%. Two other trackers, Nectiv and DemandSphere, ran their own samples and got different numbers (93 and up to 90%), a reminder that these counts vary by method rather than a single fixed truth. If you search your own brand name, there is now a good chance an AI Overview sits above your usual top organic result, and it is worth checking whether it cites your own site or someone else's.

Tactics & Tools

RankZero's free AI SEO audit (Sponsor)

See whether ChatGPT, Perplexity and Google AI Overviews can find and cite your brand. Free, results in 30 seconds, no signup.

Google redesigns the Reader Revenue Manager homepage (3 minute read)

Google overhauled this publisher tool's homepage, and the key wording change is that content from subscribers can now appear in a "From your subscriptions" panel across Search, AI Mode, AI Overviews and Gemini, not just Search, Discover and News as before. Publishers running subscription or membership models should note AI surfaces are now explicitly part of that distribution.

Google Ads lets you preview AI generated ad text before it runs (3 minute read)

A new Google Ads help section lets advertisers preview sample AI generated headlines and descriptions for Search campaigns before opting into text customization, with most previews ready in under 30 seconds across several languages. It addresses a common worry about not knowing what the AI will write before it goes live on your account.

Google says to factcheck AI content before publishing (5 minute read)

Barry Schwartz's weekly video recap notes Google's advice to factcheck AI generated content before it goes live, alongside warnings against fake authorship and news that Google may be issuing more manual action penalties. Worth a reminder to anyone using AI drafting tools on the content team.

From X and LinkedIn

Jake Ward: Mine your positions 4 to 20 keywords before writing anything new (1 minute read)

Jake Ward's suggested workflow: export 28 days of Search Console data, filter for keywords ranking 4 to 20, prioritise by business impact, then update the existing ranking page's headings and sections rather than publishing a new page, checking what's beating you and what AI Overviews cover for ideas. His claim is that doubling down on pages already ranking produces faster results than creating new content for new topics, a reasonable low cost check to run on your own site next week.

Lily Ray: flags the new Anthropic policy as a turning point (1 minute read)

Lily Ray called out the updated Claude usage policy banning fake source networks meant to sway AI answers, noting the previous September 2025 version did not mention search engines or AI answers at all. She compared it to "the early days of Google's spam policies," suggesting AI companies are starting to police manipulation the way search engines once did.

Cyrus Shepard: sums up the mess in AI traffic attribution (1 minute read)

Cyrus Shepard summed up 2026 AI SEO attribution bluntly: direct traffic is rising and nobody can explain it, some legitimate AI referral traffic gets attributed correctly and some does not, rank trackers are struggling to scrape Google, and in his view 97% of web analytics data is "a jumbled mess." A blunt reminder to treat any single analytics number on AI traffic with scepticism.