2 min read
|
Saved February 14, 2026
|
Copied!
Do you care about this?
This article analyzes how Google’s Gemini AI selects content from web pages for grounding. It reveals a fixed budget of around 2,000 words per query, with the amount allocated based on source relevance. Longer pages see diminishing returns, making concise, relevant content more effective.
If you do, here's more
Google’s Gemini AI system relies on a fixed grounding budget of about 2,000 words per query, which is allocated among various sources based on their relevance ranking. In a study analyzing over 7,000 queries and more than 2,275 tokenized pages, it was found that the top-ranked source receives an average of 531 words, while the fifth-ranked source gets only 266 words. Essentially, the higher your rank, the more content you have selected for grounding, with the top source getting about double the allocation of the lowest-ranked source.
The analysis reveals diminishing returns for longer pages. For instance, a typical page under 1,000 words sees around 370 words selected, giving a coverage of 61%. In contrast, pages exceeding 2,000 words drop to around 22% coverage. The trend indicates that after a certain point, adding more words does not enhance the amount of content selected; instead, it dilutes overall coverage. The study emphasizes that concise, relevant content is more effective for grounding than lengthy articles, suggesting that content creators should prioritize density and relevance to enhance their visibility in search results.
Questions about this article
No questions yet.