How a search algorithm works
An algorithm is a set of rules a system follows to turn an input into an ordered output. In search, it takes a query, works out what the person probably wants, retrieves candidate pages from the index, scores them against the inputs it trusts for that kind of query, and prints them in order. On a social platform the same idea decides which posts a feed shows and in which order.
There is no single algorithm. A results page is assembled by many systems working together — one for understanding the query, others for spam, for local results, for images, for freshness, for reviews — and which of them dominates depends on what was typed. Small changes ship continuously and are never announced. Larger core updates rebalance how those systems interact, are announced, and take time to roll out fully.
Why the algorithm matters
It is the reason SEO is a moving target rather than a checklist you complete once. When the balance shifts, sites that were doing the right things for the wrong reasons lose ground, and sites with genuine substance often gain without changing anything. Understanding that saves you rebuilding a website every time a competitor moves above you for a week.
It also explains why nobody can promise a position. The result you get is produced by systems the person selling to you cannot see, tuned for a searcher whose location, language and device you do not control. Anyone guaranteeing a place in the results either misunderstands this or is hoping that you do.
Common mistakes with algorithms
The first is treating rumour as diagnosis. Every fall in traffic gets blamed on an update within hours, usually before anyone has checked whether a page was accidentally set to noindex, a redirect broke during a release, tracking changed, or the season simply turned. Rule your own site out before you look outside it.
The second is optimising for the machine instead of the reader. Writing to hit a keyword density, spinning near-identical pages for every district, or filling structured data with claims the page does not support are all attempts to satisfy a scoring system that is explicitly designed to catch them.
The third is expecting recovery on your own timetable. A site affected by a core update usually has to wait for a later one before the improvements are fully reflected, however quickly the work gets done.
How to act on it
Keep a dated record of what you changed and when, so that when results move you can separate your own work from an outside shift. Watch a small set of pages that matter rather than a site-wide average, because an average hides which part of the site actually moved.
When an update lands, wait for it to finish rolling out before judging anything, then compare the pages that lost with the pages that held and look for the difference in substance rather than in settings. If the loss is broad and it persists, it is usually a content and trust problem, and the remedy is the unglamorous SEO work of making those pages genuinely better than whatever replaced them.