Google has upgraded two key attention patterns to make document ranking faster, smarter, and more aligned with user intent:

1. Inter-Document Block Sparsity

Old:
Models compared documents unnecessarily, wasting compute.

New:
The model reviews each document individually but only compares them in relation to the query, reducing useless processing.

2. Query-to-Document Relevance

Old:
Not all words in the query were treated with the right importance.

New:
The model now learns which keywords, punctuation, and intent signals actually help identify the most relevant document, and focuses attention there.

What do you think? How should SEO/AEO/GEO community take this further?