📝 SilkGeo Blog

On Oct 2 Meta reported that mathematicians working with Muse Spark 1.1/1.2 produced six papers, five answering long-open problems in probability, differential equations, group theory and more, such as a GAP program finding a 384-element group-theory counterexample and proving finite-time wave collapse. This is not AI replacing mathematicians but a division of labor: humans choose problems, judge and take responsibility, while AI handles massive search, computation and cross-domain association. Mathematics fell first because its rules are tight and self-verifying.

TypeSafe launched the decision model Jev on Sept 15, and within 18 days OpenAI's Decisions API, Cloudflare Clef, and Amazon's Strands Decider followed. They can't write or chat; they only pick from finite options and return probabilities, yet run in tens of milliseconds at near-zero cost. AI is shifting from one all-purpose model to cheap small models making countless judgments plus expensive large models for complex thinking, a directional signal for how AI chooses and recommends content.

In late September 2026 the real numbers from Google's AI Contribution pilot surfaced: around 100 invited publishers, small sites earning under 0.1% of ad revenue, and an opaque formula. Crucially, only content that materially changes the AI's answer counts as a contribution; pages linked just to confirm a fact earn nothing. The math draws a clear line between load-bearing content and citation padding.

Google's fourth spam update of 2026 began September 24 and runs until around October 8. Behind it sits S-CTS, a cluster detection system using semantic embeddings to group templated content and action entire networks rather than single pages. Site networks, formulaic structures and burst publishing are now the riskiest signals.

A site published 1,200 AI pages in 60 days and lost 94% of traffic. Meanwhile, 86.5% of top-ranking pages use AI assistance. The difference is not whether you use AI, it is how you use it.

On August 6th, the China Advertising Association held its second round of public consultation for GEO group standards in Beijing. Attendees included C

China GEO market surges from $35M to $420M as national standards are drafted. 72.5% of commercial decisions now driven by AI search. The window for brands is closing fast.

When a user asks ChatGPT for a tool recommendation, the three to five names in the answer are becoming the new shelf space. This guide maps the 2026 AI visibility tool landscape, t

An AI recommendation list usually holds three to five slots. For a new brand without a big budget, earning one of those slots is not about spending more — it's about taking apart t

Three tech titans just hit the brakes on AI. Most brands aren't even in the race yet. As AI search replaces traditional search, GEO becomes the new battlefield for brand survival.

A University of Washington study estimates AI Overviews reduced Wikipedia search referrals by ~5%. BrightLocal shows 45% of consumers now use AI for local business recommendations, up from 6%. This article decodes what these numbers mean and outlines a three-step GEO action plan for brands.

AI Visibility is shifting from a marketing concept into a real brand metric. The true value of GEO is not getting AI to mention you, but getting AI to choose you when it matters.