Hire Me

How search works

How search works, from a query to an AI answer

Five diagrams that explain how I think about search: how a page gets ranked, how AI builds an answer, how an entity becomes a topical map, what a strong page contains, and how I test ideas before trusting them.

01 Chapter

From a search to a ranked page

Each stage can fail on its own. Knowing which one is broken decides the fix.

  1. 01

    The query

    Someone types or speaks a search. The engine first works out the entity, the intent and the kind of result they expect.

  2. 02

    The crawl

    The page is found through links and sitemaps, then fetched. Pages that are hard to reach, slow or blocked may never be read.

  3. 03

    The understanding

    The page is rendered and read. Entities, attributes and their relations are pulled out, and the page is filed under what it really covers.

  4. 04

    The ranking

    Candidate pages are compared on relevance, quality and trust, including how much the whole site knows about the topic.

  5. 05

    The result

    The winner appears as a blue link, a map pack listing, a featured snippet or a source inside an AI answer.

Crawl problems
Technical SEO
Understanding problems
Semantic SEO
Ranking problems
Topical authority & links

02 Chapter

Inside an AI-generated answer

This is the stage GEO and AEO work on: being one of the sources an AI answer quotes.

01 One question

How often should I service my air conditioner?

02 Split into smaller searches

  • AC service frequency
  • AC maintenance checklist
  • How often to change an AC filter
  • Signs an AC needs service
  • AC tune-up cost

03 Passages read and judged

  • HVAC company maintenance guide
  • Manufacturer care manual
  • Discussion thread on Reddit
  • Page with no clear answer
  • Outdated forum post

04 One answer, a few sources

Once a year, ideally before summer, with filter checks every one to three months. 123
  • Answer in quotable sentences. AI reads passages, not whole pages, so each section should answer its question on its own.
  • Cover the follow-up questions. The smaller searches are the questions your page should already answer.
  • Be consistent everywhere. The same facts about the brand on the site, profiles and communities like Reddit make it a safer source to cite.

03 Chapter

From an entity to a topical map

An attribute earns its own page when it has its own demand, its own intent and enough unique substance. Otherwise it becomes a section.

Source context

A local HVAC company that earns from repairs and installations

Central entity

Air conditioning

  • Types

    • Central AC
    • Ductless mini-split
    • Heat pump
  • Problems

    • AC not cooling
    • AC leaking water
    • Strange noises
  • Cost

    • Installation cost
    • Repair cost
  • Maintenance

    • AC tune-up
    • Filter changes
  • Service area

    • Main city page
    • Nearby towns, until each has unique substance
  • Brands

    • Brands installed

Own pageSection on its parent page

  • The source context decides the weight. A repair company and a review site can share an entity but need different pages.
  • Core first, outer second. Money pages form the core section. Supporting guides form the outer section and link up to the core.
  • No page without a job. Pages that split the same intent compete with each other, so they get merged into one.

04 Chapter

What a semantic page contains

One page, one main question. Everything on it either answers that question or leads to the next one.

  1. Title and H1 that match the query

    They name the page's single main topic, in the words people search with.

  2. The answer first

    The opening lines answer the main question directly, before any background.

  3. Entity facts a machine can read

    Attributes such as types, prices, steps and service areas, stated plainly in lists or tables.

  4. Subheadings in the order people ask

    Each H2 answers the next question in the query network.

  5. The follow-up questions

    A short FAQ for what people ask after the main answer.

  6. Internal links with a purpose

    Up to the parent topic, across to siblings, and on to the page that earns.

  7. Structured data

    Schema that names the entity, the business and the type of page.

  8. One clear next step

    A single call to action that fits the intent of the search.

05 Chapter

How I test an SEO idea

Much of what I know comes from tests like this, not from rules repeated online.

  1. 01

    Write the hypothesis

    State the idea as a claim that can fail: this change will lift this metric on these pages.

  2. 02

    Pick test and control

    Choose pages to change and similar pages that stay untouched for comparison.

  3. 03

    Change one variable

    Only one change at a time, so the result can be traced back to it.

  4. 04

    Measure long enough

    Track impressions, clicks and positions until the difference rises above normal movement.

  5. 05

    Decide and record

    Roll it out, refine it or drop it, and write the result down either way.

Variables I've tested

  • AI-generated content
  • Exact match domains (EMDs)
  • Technical variables

06 Next

Want to talk through your site?

Request a review

Start your semantic SEO project

Share where your site is now and what you want from search. I read every message and reply myself.

Prefer email? [email protected]
I only use your details to reply.