Google may widen ranking pool, shaking up SEO competition

Google could be widening the pool it evaluates before ranking, and that would change how agencies value the pages they have been ignoring for years. On May 11, Search Engine Land reported that newer Google research and court testimony point to a system that may look at far more than the roughly 20 to 30 candidate pages SEOs have long assumed shaped final results.

That old 20-to-30-page frame matters because it was never just theory. In U.S. v. Google, Pandu Nayak, Google’s vice president of Search, testified that RankBrain looks at the top 20 or 30 documents and that Google waits until it is down to the final 20 or 30 before running it. He also said RankBrain is too expensive to run on hundreds or thousands of results, and described classical postings-list retrieval as the core mechanism that culls the corpus to tens of thousands of pages before ranking begins.

The newer signal comes from Google’s AI Search stack. In February, Jeff Dean described it as a staged system that first narrows the web to relevant pages, then ranks them, then generates an answer. Dean said such systems can identify around 30,000 documents and refine that set down to the final results. That does not mean Google has abandoned layered retrieval. It does suggest the system may now be capable of weighing a much larger field before the final cut.

For agencies sitting on massive, neglected inventories, that is the part worth auditing now. Archived blog posts that still answer real questions, faceted pages that surface useful combinations, support content buried three clicks deep, and thin but relevant URLs may no longer be dead weight if Google is evaluating a wider field. If more pages are in play, small differences in quality, structure, internal linking, and authority can decide whether a URL gets a real shot or disappears before the last pass.

The scale around this is hard to miss. Google said in 2021 that Search was updated thousands of times a year, handled billions of queries every day in 150 languages, and identified relevant information from hundreds of billions of pages in its index. It also said its evaluation processes had reduced irrelevant results on a search results page by over 40% in five years and that it sent visitors to more than 100 million sites every day. In March 2024, Google said its spam and quality changes would reduce low-quality, unoriginal content by 40%, later revising that estimate to 45%.

That combination is what makes this moment strategic rather than cosmetic. Google has spent years tightening quality while suppressing junk, and now the same machine may be capable of looking farther before it chooses. For agencies, the implication is straightforward: the content libraries that looked too messy, too old, or too marginal to matter may be the next competitive advantage.