Somebody gets it ๐ง
Generate viral LinkedIn posts in your style for free.
Generate viral LinkedIn posts in your style for free.
Steve Toth
These are the best posts from Steve Toth.
4 viral posts with 2,050 likes, 207 comments, and 134 shares.
4 image posts, 0 carousel posts, 0 video posts, 0 text posts.
๐ Go deeper on Steve Toth's LinkedIn with the ContentIn Chrome extension ๐
Best Posts by Steve Toth on LinkedIn
Google just quietly reduced its crawl limit by 86.7%.
On Jan 31, 2026: Googlebot crawled the first 15MB of HTML files.
On Feb 5, 2026: That limit dropped to 2MB.
That's not a typo.
(H/T to Suresh Kumar Gondi Gondi for putting this on my radar)
Here's what this means for your site:
1. Every HTML file over 2MB gets is only partially indexed.
Google stops fetching and only sends what it already downloaded.
Your content below the cutoff? Invisible.
2. Every resource (CSS, JS, JSON) has the same limit.
Each file referenced in your HTML is fetched separately.
Heavy files? They're getting chopped.
3. PDFs get 64MB (the only exception).
Everything else, HTML, JS, JSON etc. now plays by the 2MB rule.
What to do now:
- Audit your page sizes with Screaming Frog
- Minify and compress all HTML, CSS, and JS
Note: Your HTML file size doesn't include linked resources (images, CSS, JS) - those are fetched separately. But each of those files also has its own 2MB limit.
At Notebook Agency, we've checked our client crawls and the largest client page we've identified is 1 MB (an infinite scroll category page).
And according to Screaming Frog, the average size of an HTML page is about 30 kb so most of us are probably good, but the message is clear: Google needs to save money and it's not going to waste it on bloated files.
P.S. I'll link the official Google documentation in the comments.
#seonotebook
On Jan 31, 2026: Googlebot crawled the first 15MB of HTML files.
On Feb 5, 2026: That limit dropped to 2MB.
That's not a typo.
(H/T to Suresh Kumar Gondi Gondi for putting this on my radar)
Here's what this means for your site:
1. Every HTML file over 2MB gets is only partially indexed.
Google stops fetching and only sends what it already downloaded.
Your content below the cutoff? Invisible.
2. Every resource (CSS, JS, JSON) has the same limit.
Each file referenced in your HTML is fetched separately.
Heavy files? They're getting chopped.
3. PDFs get 64MB (the only exception).
Everything else, HTML, JS, JSON etc. now plays by the 2MB rule.
What to do now:
- Audit your page sizes with Screaming Frog
- Minify and compress all HTML, CSS, and JS
Note: Your HTML file size doesn't include linked resources (images, CSS, JS) - those are fetched separately. But each of those files also has its own 2MB limit.
At Notebook Agency, we've checked our client crawls and the largest client page we've identified is 1 MB (an infinite scroll category page).
And according to Screaming Frog, the average size of an HTML page is about 30 kb so most of us are probably good, but the message is clear: Google needs to save money and it's not going to waste it on bloated files.
P.S. I'll link the official Google documentation in the comments.
#seonotebook
OK this is MASSIVELY useful for anyone optimizing for AI search. Notebook Agency Head of R&D Massimiliano Geraci shared 12 practical insights for getting cited by LLMs:
I recently hosted an AEO Masterclass in Chiang Mai with Max and Michaล Suski from Surfer. Max's talk was jam-packed with findings from his countless hours researching how ChatGPT and Google AI Mode actually retrieve content.
Here are the highest-leverage insights:
1. Put your most important answers in the first 30% of the page. LLMs often build their entire response from your title, meta, and first few paragraphs.
2. Structure content in listicles, tables, and comparisons. LLMs prefer pre-synthesized formats because they're easier to comprehend.
3. Avoid answering the same question twice in different wording. Redundancy creates competing chunks and lowers semantic confidence.
4. Do a "20 question fan-out" for each target page. Generate 20-30 subquestions someone could ask, then answer all of them clearly in one place.
5. Keep ONE idea per paragraph. If your paragraph switches topics or contains multiple ideas, the LLM cannot use it as a stable citation.
6. Ensure your page loads instantly. If Time to First Byte is above 1 second, agents time out and citations never happen.
7. Add unique insights, not generic fluff. LLMs downweight generic SEO text and look for specifics, numbers, tradeoffs, and real steps.
8. Verify your indexing in Bing Webmaster Tools. ChatGPT and Perplexity rely heavily on Bing's index.
If you only implement three things: Create a top section answering all fan-out questions in 150-300 words, convert the rest into clean chunks with lists and tables, and add schema with fast server-side rendering.
Max says these three moves alone put you ahead of 99% of websites in LLM visibility. Time will tell how this evolves, but the directional thinking here is solid.
To get tips like this straight to your inbox, check out ainotebook.com (Comments > Sort By Most Recent)
#ainotebook
I recently hosted an AEO Masterclass in Chiang Mai with Max and Michaล Suski from Surfer. Max's talk was jam-packed with findings from his countless hours researching how ChatGPT and Google AI Mode actually retrieve content.
Here are the highest-leverage insights:
1. Put your most important answers in the first 30% of the page. LLMs often build their entire response from your title, meta, and first few paragraphs.
2. Structure content in listicles, tables, and comparisons. LLMs prefer pre-synthesized formats because they're easier to comprehend.
3. Avoid answering the same question twice in different wording. Redundancy creates competing chunks and lowers semantic confidence.
4. Do a "20 question fan-out" for each target page. Generate 20-30 subquestions someone could ask, then answer all of them clearly in one place.
5. Keep ONE idea per paragraph. If your paragraph switches topics or contains multiple ideas, the LLM cannot use it as a stable citation.
6. Ensure your page loads instantly. If Time to First Byte is above 1 second, agents time out and citations never happen.
7. Add unique insights, not generic fluff. LLMs downweight generic SEO text and look for specifics, numbers, tradeoffs, and real steps.
8. Verify your indexing in Bing Webmaster Tools. ChatGPT and Perplexity rely heavily on Bing's index.
If you only implement three things: Create a top section answering all fan-out questions in 150-300 words, convert the rest into clean chunks with lists and tables, and add schema with fast server-side rendering.
Max says these three moves alone put you ahead of 99% of websites in LLM visibility. Time will tell how this evolves, but the directional thinking here is solid.
To get tips like this straight to your inbox, check out ainotebook.com (Comments > Sort By Most Recent)
#ainotebook
Google has been telling SEOs ๐๐ต๐ฎ๐ ๐ป๐ผ๐ ๐๐ผ ๐ฑ๐ผ since 1997.
Some notable "don'ts":
- Don't build links
- Don't create pages for SEO
- Don't guest post
- Don't use exact-match anchor text
- Don't optimize for featured snippets
- "Just write good content"
- And now: don't actively seed mentions
- Don't optimize for LLMs differently
Here's the reality:
- Every time Google said "don't," the cutting-edge SEOs tested it and won the channel
- Brand mentions correlate with AI Overview visibility at 0.664, over 3x stronger than backlinks (per Ahrefs' 75K brand study)
- Clients are looking at LLMs as a distinct channel, with different priorities. They're saying things like, "I don't like how ChatGPT describes our product, help us change that"
- "Optimizing" for LLMs goes way beyond visibility for non-branded keywords. It's about perception, narrative, and how your company is recommended
- And Google shouldn't speak for ChatGPT, Claude, or Perplexity, though this guidance reads like it does
Sure, the "it's just SEO" camp will show prospects their LLM citations and call it a win. Fair, you can get 100% cited by doing solid SEO.
But you'll get dramatically better results by prioritizing generative AI in the work you do... without sacrificing SEO (which is another de-posititioning tactic from the "it's just SEO" camp.
Remember: Google's job is to protect Google.
Your job is to test.
#seonotebook #ainotebook
Some notable "don'ts":
- Don't build links
- Don't create pages for SEO
- Don't guest post
- Don't use exact-match anchor text
- Don't optimize for featured snippets
- "Just write good content"
- And now: don't actively seed mentions
- Don't optimize for LLMs differently
Here's the reality:
- Every time Google said "don't," the cutting-edge SEOs tested it and won the channel
- Brand mentions correlate with AI Overview visibility at 0.664, over 3x stronger than backlinks (per Ahrefs' 75K brand study)
- Clients are looking at LLMs as a distinct channel, with different priorities. They're saying things like, "I don't like how ChatGPT describes our product, help us change that"
- "Optimizing" for LLMs goes way beyond visibility for non-branded keywords. It's about perception, narrative, and how your company is recommended
- And Google shouldn't speak for ChatGPT, Claude, or Perplexity, though this guidance reads like it does
Sure, the "it's just SEO" camp will show prospects their LLM citations and call it a win. Fair, you can get 100% cited by doing solid SEO.
But you'll get dramatically better results by prioritizing generative AI in the work you do... without sacrificing SEO (which is another de-posititioning tactic from the "it's just SEO" camp.
Remember: Google's job is to protect Google.
Your job is to test.
#seonotebook #ainotebook