LabraSEO

Universal Cart in Google Search 🛒, AI Mode link updates 🔗, ChatGPT traffic tracking bug 📉

Issue 16, 16 October 2026

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

Headlines & Launches

Google launches Universal Cart in Search and Gemini (3 minute read)

Google's Universal Cart, built on the Universal Commerce Protocol, lets US shoppers add products from multiple retailers, including Nike, Target, Walmart and many Shopify stores, into one basket and check out with Google Pay. It tracks price changes and applies promo codes automatically. If you sell online, Google is now a shopping destination in its own right, not just a referrer, so how your products appear there matters as much as your ranking.

Google changes AI Mode and AI Overview links again, still withholds click data (4 minute read)

Google says it is placing inline links higher more often, testing a block of official site links for some queries, expanding its timely article carousels globally, and testing linked creator names on AI Mode videos. Google also claims preferred sources usage is up more than sevenfold, but as Barry Schwartz notes, it still will not share click data to prove any of this drives traffic. Treat the claims as unverified until you see your own numbers move.

Bing tests related searches based on your browsing history (2 minute read)

Bing is testing a box labelled "Related Searches Based on Your Browsing," shown below sponsored results, mirroring a format Google trialled earlier this year. It is not yet confirmed whether this appears for organic results too. If it rolls out widely, personalised related searches could change which queries bring people to your site.

Google tests a redesigned search bar pushing AI Mode (2 minute read)

Google is testing a new search bar in its mobile app that reads "Ask Google" instead of "Search," with a plus button for image uploads and AI tools. It is a visual nudge toward AI Mode and multimodal search rather than typed queries. Expect more traffic to start as a photo or a conversational prompt rather than a keyword search.

Deep Dives & Analysis

ChatGPT traffic drop looks like a tracking bug, not a real decline (5 minute read)

Several analytics tools, including Graphite and LLM Pulse, logged ChatGPT referred sessions falling 22 to 57% from October 9, which Graphite's CEO ties to ChatGPT links now opening without the usual tracking tag, so visits land in GA4 as Direct instead. Adobe's Cedric Huesler found Direct traffic rising by roughly the same amount ChatGPT traffic fell on affected pages, with similar visitor behaviour. If your ChatGPT traffic looks like it collapsed this month, check your Direct channel before assuming you lost visibility.

A prediction model flags a Panda-style update as likely within a month (5 minute read)

SEO consultant Dan Taylor says a model he built puts the odds of a Panda-style, site-wide quality update at 96.8% within 30 days, aimed at scaled AI content and fake author bylines rather than any specific word count or template, as Panda originally was in 2011. This is one person's forecast from his own model, not a Google announcement. If you have been publishing AI-written content at scale with thin or fake bylines, now is the time to review it, since the original Panda hit entire domains rather than single pages.

One analyst's forecast: ranked results stop being the default between 2028 and 2030 (9 minute read)

Duane Forrester, who discloses that his company sells AI visibility measurement, predicts the classic Google results page will still exist but stop being where most people land by default sometime in that window, starting in the US. He is explicit this is a forecast built on adoption curves and regulatory pressure, not a confirmed Google plan, and notes Google has repeatedly denied making AI Mode the default. Worth tracking as a scenario for long term planning, not a near term change to act on.

Study of 76,450 shared Gemini chats shows how it picks sources (15 minute read)

Resoneo analysed over 222,000 Google searches run inside shared Gemini conversations and found Google classifies every question by topic, task, follow-up type and sentiment before deciding what to search for and cite, using a 106 category system invisible to users. This confirms that what Gemini searches for behind the scenes can differ meaningfully from your literal question, so matching the underlying intent categories, not just keywords, is what earns a citation.

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.

From X and LinkedIn

Nicholas Dulait: A practitioner's checklist for ranking inside AI answers in 2026 (2 minute read)

Nicholas Dulait, on X, lays out a GEO process: pick 5 to 10 target prompts, map each one's sub query fan out in ChatGPT or Claude, redo that weekly, and assign each sub query to a page, PR push or forum post, prioritising comparison, guide and review content that LLMs scrape most. He also argues for tracking share of voice across repeated runs rather than a single rank position. It is a concrete starting checklist your team could run by hand before paying for a tracking tool.