#search-algorithms

#search-algorithms

Grokipedia might be dropping more with the March 2026 broad core update. The decline seems to be continuing with the latest broad core update.

Buyers no longer open ten tabs, skim through blog posts, and slowly form an opinion over weeks. Instead, they ask a single question to an AI system and receive a shortlist in return, usually two or three companies that feel familiar, credible, and safe enough to justify internally. That shortlist often becomes the entire market in the buyer's mind.

Google had a brief serving issue with Google Search. Google is testing showing vertical search services next to its own in the EU. Google said a spike in impressions do not cause problems for Google Search. Google won't use a sitemap file if it is not convinced of new or important content.

For most of the last two decades, SEO mostly meant one thing, where you ranked on Google (and occasionally Bing). The customer journey was familiar. Someone searched, scanned a list of links, clicked and explored. Now the journey is increasingly 'ask, get an answer, take action.' And the platforms shaping that journey include ChatGPT, Claude, Gemini, Perplexity and Google itself, which is inserting AI summaries, what Google calls AI Overviews, into search results.

Several weeks after Google rolled out support for Preferred Sources globally, Google added official help documentation for site owners to use to help them understand what it is all about and how to encourage their readers to subscribe to your site as a preferred source. In December, Google rolled out Preferred sources globally after rolling it out in the US and India in August and beta testing it in June. Now the new help documentation is available here if you need it.

What happens when the AI companies (inevitably) encounter spam and attempts at SEO/GEO manipulation in the markdown files targeted to bots? What happens when the .md files no longer provide an equivalent experience to what users are seeing? What happens if they continue crawling those pages but actually toss them out before using the content to form a response? ...And we keep conflating "bot crawling activity" with "the bots are using/liking my markdown content?" How will we know if they're actually using the .md files or not?