🔍

​Google releases the February 2026 Discover core update. The Google Discover documentation was also updated to recommend avoiding clickbait, sensationalism and artificially inflating engagement.

↪️

​Bing launches multi-turn search worldwide. Multi-turn search allows users to perform follow-up searches in a Copilot chat without leaving the search engine results page.

⚖️

​Google updates the documentation about default file size limits for Googlebot. While these file size limits have caused some worries for webmasters, Google also explained that there is "no need to weigh bytes".

🤖

​Microsoft launches a new initiative for publishers to get paid for content included in AI products. The Publisher Content Marketplace is an opt-in system that will allow content licensing for LLMs and crawlers.

🏨

​Google tests "Good to Know" AI-generated labels for local hotel listings. These AI-generated captions summarise information from reviews about what's featured in the photo can appear for both owner photos and public photos.

📉

​Sites using self-ranked listicles have seen significant visibility drops. Some brands are reporting 30-50% visibility loss. These kinds of articles are very common in SaaS spaces and it seems Google is punishing this tactic.

💰

​Google confirms tests for ads below the AI Mode answers. From the Alphabet earnings call, Google also confirmed that Search grew by 12% and AI Overviews reached 1.5 billion monthly users.

New experiment: ChatGPT ecommerce is built on Google

In our new experiment, we confirmed the theories that ChatGPT runs encoded fan-out queries through Google Shopping, using these results to generate product recommendations. In fact, 75% of the top products overlapped.

SEO is far from dead.

Brands that win in AI search will be the ones doubling down on their strong SEO foundations, backed by connected tools that help deliver visibility everywhere users search.

Your schema will make a fine addition to my collection

You hear all those GEO experts tell you that schema is super-important for AI systems? It's BS, and I have the receipts ⬇️🦆
​
Edit for clarity: This test is showing that schema is treated like any other text so does not “work” and is not specifically used as intended. Apologies for 🇬🇧 sarcasm.
​
I've commonly seen "proof" that LLMs are parsing structured data in real-time when people perform an experiment such as only including the company address in the schema, not the 'visible' content . The fact an LLM could return this address is 'proof' they are using schema.
​
So here is what I did:
​
I set up a page for a fictional company "DUCKYEA t-shirts" and did not include a company address on that page.
​
Instead, I completely made up my own JSON-LD schema:
​
<​script type="application/ld+json">
{
"@context": "https://lnkd.in/eN8KsnYG",
"@type": "MallardEnterprise",
"flockName": "DUCK YEA T-SHIRTS",
"waddleStyle": "Aggressive",
"nestingGrounds":
{
"@type": "LilyPadAddress",
"reedNumber": "77",
"puddle": "The Muddy Bank",
"region": "South Pondshire",
"featherCode": "DK99 YEA",
"country": "United Queendom" },
"migrationPattern": "Non-Migratory",
"quackVolume": "Loud" }
<​/script>
​
The result, when I ask ChatGPT or Perplexity: "what is the address of this company? [url]"
​
I got: "The address shown on that page (in its embedded structured data) is: 77 The Muddy Bank, South Pondshire, DK99 YEA, United Queendom."
​
My conclusions:
​
1️⃣ Firstly, I am *not* saying "schema is not worth it". Google specifically say they use their KG with their AI systems, and structured data helps reduce ambiguity (that is kinda your job). Still do schema.
​
2️⃣ I have long argued that because of how base LLMs are trained, the explicit nature of schema is lost, because the models do not retain training data. They are simply building a mesh of probabilities between tokens, your schema is lost within that.
​
3️⃣ In my opinion, this test shows that the LLM agent is simply picking up whatever you are listing in the HTML. It does not matter if it is valid schema. If the system interprets the text as relevant to the prompt, it is included. It would therefore indicate that schema is *not* being used in the explicit sense it was designed for with those systems.
​
The takeaway: Schema, good. Repackaging the basics as some magical new GEO formula, bad.
​
Keep testing and having fun🫡
​
​Richard Barrett told me he did this test first, and I wanted to repeat it. I don't think he drew a duck, though.

This is how markdown dies, with thunderous applause

"Should I create a new markdown version of my pages 'optimised' for AI?"

I was asked this yesterday on LinkedIn, and my answer was a definite "no".

​As luck would have it Jono Alderson went into the detail of why.​

Did you press the block button?

It's a common mistake to use robots.txt to block access to the CSS / theme files of your site. You should let Google access these, so it can accurately render your site and have a better understanding of the content.

He was deceived by a lie. We all were. It appears that the canonical is behind everything

Add a self referential canonical tag to all canonical pages. This means if someone scrapes your content or links to it with query strings, it's still clear to Google which version to give credit to.

I’m placing these Twiddlers in your care. Treat them well

Google use contained programs in a framework called 'Twiddlers' which re-rank results. Twiddlers focus on an individual metric, work in isolation of each other and try and improve SERP quality. This means a factor might not be in the 'core algorithm' but it might be affected by a conditionally run Twiddler that alters the result. For example, there is a Twidder that runs on YouTube queries that will improve the result of a matching video, if that channel as many videos that are also a close match to the search (implying the channel is specialist).

Top Core Updates referrer leaderboard

A big thank you to our top referrers, who have signed up over 10 people to the Core Updates newsletter, go follow them!

🥇 MJ Cachón Yáñez (LinkedIn / Bluesky)

🏅 Nikki Pilkington (LinkedIn / Bluesky)

🏅 Lidia Infante (LinkedIn / Bluesky)

🏅 Saijo George (LinkedIn / Bluesky)

🏅Alice Rowan (LinkedIn)

🏅Martijn Jeurissen (LinkedIn)

🏅Chester Beard (LinkedIn)