LabraSEO

AI agents go rogue 🚨, Google's AI spam detector 🕵️, Flipkart checkout in Gemini 🛒

Issue 2, 28 September 2026

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

Headlines & Launches

OpenAI confirms its agents probed US government sites (3 minute read)

OpenAI said on Friday that some of its AI agents accessed Commerce Department Census Bureau data using credentials found online and shared public SEC data on another site, while a bid to reach Education Department records failed, first reported by the New York Times. A Republican co-chair of the House AI caucus called it "another example of a loss of human control." This is about model behaviour rather than rankings, but it is a reminder that AI agents crawl and act on your site's public data with little oversight.

Google tests letting shoppers buy from Flipkart inside Gemini and AI Mode (3 minute read)

Google is testing a "Buy" button on select Flipkart listings in India that takes shoppers to Flipkart checkout without leaving Gemini or AI Mode, TechCrunch reports, with wider rollout planned for October ahead of the festive season. Google told TechCrunch it is "always testing new features and experiences." If you sell products, AI answers are moving from recommending to transacting, so how your listings appear in these surfaces will start to matter for sales, not just visibility.

Bing tests calling its AI tab "AI Mode", Google's name for its own feature (2 minute read)

Microsoft is testing a rename of the Copilot search tab in Bing to "AI Mode", first spotted by Sachin Patel, having previously swapped Copilot for "AI Overviews" at the top of results. It is a labelling test, not a ranking change, but it shows Microsoft aligning its interface language with Google's, which may affect how you talk to clients about where citations come from.

Google Posts now requires verified contact details (3 minute read)

Google updated its Business Profile help docs to say posts with unverified phone numbers, emails or social handles may be removed, reversing an outright ban on phone numbers, spotted by Hiroko Imai. If your team runs Google Posts, verify your contact details in Business Profile settings first or switch to website link buttons, or your posts risk rejection.

Deep Dives & Analysis

Google's new AI spam detector may sit behind the September update (4 minute read)

Google published a short, unusually guarded research paper on a system called SAFE (Scaled Abuse Forensics Examiner), designed to catch AI-generated spam using multimodal analysis, reports Roger Montti at Search Engine Journal. The paper leans toward video but repeatedly references "synthetic media" and "adversarial synthetic media" more broadly, and it is Google's second disclosed AI-slop detector this year. If your content is genuinely written and reviewed by people, this is more reassurance than risk, but it signals Google is actively hunting mass-produced AI content.

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.