Pagination and Result Freshness
How search APIs handle paging through results and how 'fresh' those results really are.
What it means
Most SERP APIs expose pagination via a `page` or `start` offset parameter, mirroring how a human would click 'next page' on a real results page — page 1 returns positions 1-10, page 2 returns 11-20, and so on. What isn't obvious is that the underlying index isn't frozen between those calls: the live results a query returns can shift slightly between page 1 and page 2 if enough time passes, or if the query happens to straddle a re-crawl.
In practice
An agent paginating through 3 pages of results for a fast-moving topic (say, breaking news) fetches page 1, waits a few seconds while processing, fetches page 2 — but by the time page 2 comes back, a newly published article has bumped everything down one position, so what the agent expected to be position 11 on page 2 might actually be a repeat of something already seen on page 1, or a gap where an expected result silently disappeared.
Tradeoffs
For stable, non-time-sensitive queries this drift is negligible and pagination behaves predictably. For anything volatile, an agent that paginates across multiple calls should deduplicate by URL rather than trusting position numbers to stay stable, and should treat a multi-page pull as a best-effort snapshot rather than a guaranteed consistent view.