Skip to content
Tutorials Center
My learning

Topics

Lesson 6 of 36 ยท How Search Engines Work

How Search Engines Rank Pages: Signals, Relevance and Links

Understand the main groups of ranking signals, how relevance scoring and link analysis work, and why updates change results. Includes runnable scoring and PageRank demos.

When a searcher presses enter, the engine has a few thousand or a few million candidate pages that match the words in the query. Ranking decides the order. Nobody outside the search company knows the full recipe, and the recipe changes constantly, but the public guidance and decades of research give us a reliable picture of what matters.

This lesson groups the signals, shows how two classic ideas work (text relevance scoring and link analysis) with code you can run, and explains how to think about algorithm updates without panic.

The groups of ranking signals

What ranking looks at

Relevance Does the page match the query?
Quality Is it helpful and trustworthy?
Usability Fast, mobile friendly, safe
Links Do other sites vouch for it?

Four broad families, hundreds of signals.

1 / 2
  • Relevance. Words on the page and in links pointing to it, headings, the match to meaning (not only exact words), and whether the page satisfies the likely goal behind the query (see search intent).
  • Quality. Original information, depth, accuracy, clear authorship and reputation. More in E-E-A-T and content quality.
  • Usability. Page experience: loading speed, stable layout, mobile design, HTTPS, no intrusive interstitials.
  • Links. Links from other sites act as votes, weighted by the quality and relevance of the source.
  • Context. Location, language, device and sometimes freshness change which results are best.

You may read lists of "200 ranking factors". Treat them as hypotheses. Search engines have said that some commonly claimed factors are not used and that signals interact in complex ways. Focus on the underlying goals: relevance, helpfulness, access and trust.

Text relevance: a worked example

A classic way to score a document against a query is TF-IDF: term frequency times inverse document frequency. A word that appears often in a page is a good sign (TF). A word that appears in almost every page, like "the", tells you little, so rare words get more weight (IDF). Modern engines use far more advanced language models, but TF-IDF explains why using the right specific terms matters and why repeating a common word does not help. We build on this in competitor and TF-IDF analysis.

Python Rank documents with TF-IDF

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The "guide" page wins because it contains all three query terms more often relative to its length. The "news" page scores zero. Try a query such as "the" and notice how little it separates the documents.

Early on, search engines realized that links are a signal of importance: a page linked by many important pages is probably important too. PageRank formalizes this. Imagine a random surfer who follows a random link most of the time and jumps to a random page sometimes (the damping factor). The share of time spent on each page is its score. Today link analysis is one signal among many, but understanding it explains why internal links and quality backlinks matter.

flowchart LR
    A["Home"] --> B["Guide"]
    A --> C["Shop"]
    B --> C
    D["Other site"] --> B
    C --> A

Pages that receive links from well-linked pages collect more of the surfer's attention.

Python Compute PageRank for a tiny site

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The home page ranks highest because every other page links to it, and "blog" is lowest because nothing links to it. This is a very simplified model of one signal. It shows, though, why a good internal linking structure sends attention to your important pages.

Machine learning and understanding meaning

Modern ranking uses machine learning to understand queries: synonyms, misspellings, questions, entities (people, places, things) and the relationships between them. Searching for "jaguar speed" should depend on whether the context is the animal or the car. For writers this means: cover a topic naturally and completely, use clear language and add context. You do not need to repeat an exact phrase in every sentence.

Algorithm updates

Search engines change their systems many times a year. Small changes happen daily; occasional core updates reassess how quality is evaluated broadly, and specific systems (spam, product reviews, helpful content) are improved over time. A drop after an update does not necessarily mean you broke a rule; it may mean other pages now serve the query better.

How to respond sensibly:

  1. Check dates in your performance report against announced updates, but do not assume correlation.
  2. Compare your page to those that now rank above it: depth, accuracy, clarity, freshness, experience.
  3. Improve the content for users, not for the update.
  4. Be patient; effects of improvements may take a later update cycle to show.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. In TF-IDF, why do rare words receive more weight than common ones?

    Inverse document frequency lowers the weight of words that appear everywhere.
  2. 2. In the PageRank idea, which page tends to score highest?

    Links act like votes, so pages that collect links from well-linked pages gain score.
  3. 3. A page drops after a core update. What is a sensible first reaction?

    Updates reassess quality broadly. Learn from the pages that now serve the query better.
  4. 4. Which is NOT one of the broad signal families named in this lesson?

    The families are relevance, quality, usability, links and context. Domain name length is not one of them.

Summary

  • Ranking blends relevance, quality, usability, links and context for each individual query.
  • TF-IDF captures the idea that specific, frequent terms signal relevance and ubiquitous words do not.
  • PageRank-style link analysis shows why internal and external links channel importance.
  • Understand meaning, not just words: cover topics completely and naturally.
  • Treat updates as a prompt to improve usefulness, not as a puzzle to outsmart.

Next: Search Intent and SERP Features