Author: GeoParrot Team
-
Google’s AI Overview Citation Cards Now Show Text That Isn’t On Your Page. Nobody’s Checking If the Paraphrase Is Accurate.
SEO consultant Brodie Clark spotted Google AI Overview citation cards displaying AI-generated summaries instead of verbatim page text on October 23, 2026, corroborated by Barry Schwartz. Google hasn’t confirmed it’s a rollout, and it doesn’t appear on every query. We separate what this actually changes about GEO from what it doesn’t, and what to check…
-
AI Brand Recommendations Flipped 79% of the Time in 4 Weeks. Two of the Four Assistants Tested Never Touched the Web.
A new Envoyra study found 79% of AI assistant answers to the same buying questions changed their recommended brand at least once over 4 weekly tests. But Claude and Gemini were tested without web search — so some of that “instability” can’t be fresh search results. We separate the two different things the number is…
-
Adidas SEO & GEO Audit: Does Google AI Cite Adidas.com?
An evidence-first audit of adidas.com: richer per-review schema than Nike (AggregateRating plus authored, dated Review objects), but a slightly lower AI Mode citation rate (4/10 vs Nike’s 5/10), a robots.txt that names ClaudeBot by name, and a ProductGroup-vs-Product schema split between its US and Korea product pages for the same SKU.
-
A Model Gives a Google Update 96.8% Odds in 30 Days. It’s Not Confirmed, and Last Time Rankings Moved, Our AI Citations Barely Noticed.
SEO analyst Dan Taylor’s gradient-boosted prediction model puts 96.8% odds on a Google update within 30 days of October 13, 2026, with late October to mid-November as the likely window, theorized to target scaled AI content Panda-style. Google has confirmed nothing. We check what the model actually measures, and what our own two prior update-citation…
-
Nike SEO & GEO Audit: Does Google AI Cite Nike.com?
An evidence-first audit of nike.com: solid Offer schema and a 62-locale sitemap, but zero FAQPage schema anywhere on the site, no AggregateRating data, and AI Mode citations that split cleanly between brand queries (cited) and comparison queries (not cited).
-
Researchers Planted a Fake Protocol Into AI Search in a Week. The Rule Against It Doesn’t Start Until November 12.
A Fudan University paper posted October 8 shows 8 of 10 AI search tools cited a made-up “Cactus Gravity Protocol” within seven days of researchers planting it on nine sites those tools already cite often — Grok linked to it in 30 of 32 answers. The predictor wasn’t content quality, it was whether the platform…
-
We Searched “GeoParrot” on Google. Here’s Exactly What the AI Overview Cited.
We ran five brand-name queries for our own site through Google and logged what the AI Overview said and which pages it cited. Every query triggered an AI Overview, almost every citation pointed back to our own GEO Lab experiment posts specifically (not our guides or trend pieces), and the process surfaced one real external…
-
Google and Shopify Just Launched an Open Checkout Protocol for AI Agents. It Doesn’t Decide Who Gets Picked.
Google and Shopify’s Universal Commerce Protocol (UCP) launched October 15 with Universal Cart, backed by 20+ retailers and payment companies. It standardizes how an AI agent completes a purchase — not which merchant the AI recommends. We measured the same listed-isn’t-picked gap on OpenAI’s product feed seven weeks ago. Here’s what UCP actually controls, what…
-
IKEA SEO & GEO Audit: One Product, 91 Markets, and a Generative-AI FAQ That Doesn’t Travel
An evidence-first audit of IKEA’s BILLY bookcase across the US and Korean sites: a dense decision unit, task-based support, and a disclosed Generative AI FAQ feature that does not yet travel across locales.
-
Two Engines, Same Week, Same 10 Topics: Pre-Registering the Test for Last Week’s Reddit Split
Monday we found ChatGPT’s Reddit citations down 88% since March while our own Google AI Mode snapshot still sits at 20.5% — but the two numbers used different engines, units, and timeframes. This week we close those gaps: same 10 topics, same day, both engines, plus a locked rule for re-testing last week’s hygiene-score reversal…