On-Page SEO and Content

By the end of this level you will be able to look at any page on your site and decide what to change inside it so that Google understands the page, indexes it against the right intent, and shows it in results with the wording you chose. On-page SEO is the part of search performance you control outright. It does not wait on a link from anyone, on a hosting provider approving a server change, or on a budget you do not have.

The mistake almost everyone makes is treating on-page SEO as a checklist. Put the keyword in the title tag, use exactly one H1, fill in the image alt attributes, push the keyword density up, call it done. That checklist takes an afternoon and usually moves nothing, because Google does not score the literal presence of a term. It judges whether the page resolves the query better than the pages already ranking above it.

What does move rankings is less photogenic. Deciding which pages should exist and which should not, writing down in one sentence the question each page answers, building the internal links that connect them, and deleting or merging whatever no longer earns its place. That work is slow, it shows up over weeks, and no plugin does it for you.

This level is built to be applied to a live site while you read it. Every lesson ends in something executable: a query to run in Google Search Console, a threshold to decide against, a review template. Work through all fourteen lessons on a single domain and you finish with a clean content inventory and a prioritised list of changes, not a pile of notes.

What this level covers

• Writing title tags and meta descriptions that describe the page accurately, survive Google’s rewriting, and earn the click on the results page.
• Building a heading hierarchy and a content architecture where every URL answers one intent and no two pages compete with each other.
• Treating internal linking as a system: which page links to which, with what anchor text, and for what purpose.
• Spotting keyword cannibalization in Google Search Console data and fixing it by merging, differentiating or redirecting.
• Making content citable: named entities, structured data, tables, clean definitions and first-party data a language model can lift.
• Auditing and pruning old content, rescuing pages in decline, and checking whether the traffic that arrives actually converts.

On-page SEO starts with search intent, not with the tag

The first on-page decision for any page is which search intent it answers, because the title, the structure, the length and the format all follow from that one choice. Google returns a fairly consistent result type for each query, and that pattern is the practical answer to what the searcher expects. If the first page of Google is full of comparison tables, a theory-heavy guide will not break in no matter how well written it is.

The way to work this out is to run the query, read the top ten results, and note what format they share, which subtopics all of them cover, and what none of them covers. That gap is your angle. The point is not to copy the competition’s outline. The point is to understand the contract Google has established between that query and the kind of page that satisfies it.

Once intent is fixed, the rest of on-page SEO stops being a matter of opinion. The title tag stops being an exercise in keyword placement and becomes the headline that carries the promise. Headings stop being decoration and become the index of questions the page answers. Length stops being a target: a short transactional page can outrank a four-thousand-word article when it resolves the query sooner.

Content architecture and internal linking are on-page work too

Content architecture belongs to on-page SEO because it determines what Google crawls, what it indexes, and which page it serves for each query. A site with thirty loose articles on one topic and no hierarchy forces Google to pick the main page itself, and that pick rarely matches the page you wanted to rank.

The pattern that works is easy to describe and expensive to apply. One pillar page covering the topic in breadth, a set of supporting pages covering each subtopic in depth, and links in both directions with descriptive anchor text. The pillar links out to every supporting page, each supporting page links back to the pillar, and supporting pages that complement each other link sideways. That is a topic cluster, and it is what turns a pile of articles into an asset.

Internal linking is by a distance the most underused lever available. It costs nothing, depends on no one outside your team, can be executed in hours, and does three jobs at once: it distributes authority inside the domain, it steers crawling toward the pages that matter, and it tells Google through anchor text what the destination page is about. Most sites have valuable pages reachable only from the navigation menu and the sitemap.

Direction of flow is worth checking as well. If the articles that attract the most external links do not link onward to the pages you want to rank, that authority is sitting still. An internal link map in a spreadsheet, with source, destination and anchor, usually exposes in half an hour problems that have been there for years.

Writing for humans and writing for language models is the same job

Writing for people and writing for answer engines such as Google’s AI Overviews or ChatGPT demand nearly identical things: self-contained sections, the answer first, and entities named explicitly. The reason is mechanical. These systems do not read the whole page, they retrieve chunks, and a chunk that starts at a heading and then uses a pronoun pointing back three sections is a chunk nobody can use.

Three rules follow, and they serve the human reader who lands on a deep anchor link from Google just as well. First, name entities in every section: write Google, write Search Console, write the name of the format, never “the search engine” or “the tool”. Second, answer in the first sentence after the heading and elaborate afterwards. Third, keep paragraphs short, because a twelve-line paragraph chunks badly and gets quoted worse.

This matters more now because of zero-click behaviour: a share of searchers get their answer inside the results page and never visit your site. There are two complementary responses. One is to make the content citable, with tables, clean definitions and first-party data nobody else holds. The other is to keep on the page what cannot fit in a summary: the template, the calculator, the full case, the caveat that forces a visit.

Widespread on-page beliefs and what actually holds up

A large share of what circulates about on-page SEO comes from manuals nobody has revisited. The table below collects the beliefs that generate the most work for the least return, what happens in practice, and what to do instead. Each row maps to a lesson in this level.

Widespread belief	What happens in practice	What to do instead
A title tag must always open with the exact keyword.	Google rewrites the title when it does not describe the page well or does not match the query, and a keyword-forced title is exactly the kind that invites the rewrite.	Write a title that describes the page precisely and works as a headline in results.
The meta description influences rankings.	Google does not use it to rank, and often replaces it with a snippet from the body that matches the query better.	Treat it as ad copy. Its job is the click, not the position.
One extra H1 and Google penalises the page.	Google parses structure without penalising heading counts. What confuses it is an incoherent hierarchy.	Use one H1 for clarity and nest H2 and H3 to match the real content.
The more content you publish, the better.	Publishing without architecture creates cannibalization, dilutes internal linking, and leaves pages competing against your own.	One intent per URL, and merge what overlaps before writing anything new.
Structured data lifts rankings.	Structured data enables rich results and helps Google understand the page, but Google does not treat it as a ranking factor in itself.	Implement only the types Google documents for your content, and validate the markup.
The alt attribute is a place for keywords.	Alt text describes the image for accessibility and for Google Images. Stuffing it with terms adds nothing and degrades the experience.	Describe what is visible in one sentence, and handle format and file weight separately.
A published article stays as it is.	Content loses traffic on its own when the query shifts, something better appears, or the information ages.	Put your highest-contributing pages on a review calendar and update before you lose them.

The 14 lessons in this level

1. Title Tags: The Element Google Rewrites Most
2. Meta Descriptions: What They Actually Do
3. Heading Hierarchy: H1–H6 Without the Superstition
4. Content Architecture: Silos, Clusters and Why They Matter
5. Internal Linking: SEO’s Most Wasted Asset
6. Writing for Humans and Models at the Same Time
7. Keyword Cannibalization: Finding It and Fixing It
8. Image Optimization: Alt, Format, Size and Core Web Vitals
9. Semantic SEO: Entities, Not Keywords
10. Structured Data That’s Actually Worth Doing
11. Content Decay: Why Your Best Post Bleeds Traffic
12. Content Audit and Pruning: What to Delete, What to Merge
13. Formats That Get Cited: Tables, Definitions and First-Party Data
14. CRO for SEOs: Traffic That Converts

How to work through this level

The efficient way to work through this level is to apply it to one site as you go, in the order the lessons are listed. The first five build the skeleton, the next five handle semantics and format, and the last four close the loop with auditing, pruning and conversion.

1. Pick one domain, yours or a client’s, and export the full list of indexable URLs before you start.
2. Connect Google Search Console and pull the last twelve months of queries and pages. Without that baseline you cannot measure anything you change.
3. Do lessons 1 to 5 back to back and apply each one to ten real URLs. Ten is enough to see the pattern and few enough that you finish.
4. As you go, flag in your inventory the pages you suspect of cannibalization and the pages untouched for over a year.
5. Save lessons 12 and 13 until the inventory is complete. Pruning without an inventory is deleting blind.
6. Record the date of every change. Without dates you cannot attribute a recovery or rule out a Core Update.

Two to three hours per lesson, practical work included, is a realistic pace. Going faster usually means reading without applying, and on-page SEO is not learned by reading.

This level is part of the free SEO course at Doctor SEO, which also covers indexing, technical SEO, authority and measurement. If you arrived here from another level, head back to the course index to see where to go next.