On-page SEO

On-page SEO is everything you change on your own pages to make them rank: the words on the page, the headings, the title, the internal links, the images and the way each page answers one specific search. It is the middle of the three parts of SEO. The technical side decides whether Google can read the page at all, and the off-page side decides how much your site is trusted. On-page is what the page itself says and how clearly it says it.

In short

• On-page SEO is what is on the page, and it is the part you fully control.
• Matching the search intent matters more than any individual element.
• One page per thing you sell. Two pages for one search means neither wins.
• The writing that helps a reader skim is the writing an AI answer quotes.

The elements on a page, and what each does

Almost every on-page checklist is a version of this list. The order roughly matches how much each one matters.

• The main content. What the page actually says. Everything below is packaging. No amount of packaging rescues a page that does not answer the question.
• The title tag. The headline shown in search results and in the browser tab. It is the strongest single signal about what a page is for, and it is also the line that decides whether anyone clicks. It should describe the page plainly and be different on every page of the site.
• The H1. The visible heading at the top. One per page. It should agree with the title tag without being an exact copy of it.
• Subheadings. H2 and H3, used to structure the page rather than to decorate it. Skipping from H2 to H4 confuses the outline for anyone reading with assistive technology, and the outline is how a machine understands the shape of your page.
• The URL. Short, readable, describing the page. A URL full of numbers tells nobody anything, and it is the part people copy into messages.
• The meta description. The summary under the title in the results. It does not move rankings. It moves clicks.
• Internal links. Links from your pages to your other pages, using words that describe the destination. This is how Google learns which of your pages matter and how they relate.
• Images and alt text. Alt text describes the image for people who cannot see it, and tells search engines what the image shows. It should describe the image, not repeat a keyword.
• Structured data. Machine-readable markup describing what the page is: a product, a service, a local business.

Search intent, which decides everything above

You can do every item on that list correctly and still not rank, because the page answered a different question from the one being asked.

Somebody typing “aircon cleaning” might want to hire someone, learn how to do it themselves, or find out how often it should be done. Those are three different pages, and matching the right one is not a refinement you add at the end. It is the decision that determines whether the page has any chance at all.

The full account of how to tell which is which is on the search intent entry, and the wider reading of the results page is SERP analysis.

One page per thing, and why

The most common structural fault is a business with six services and one page listing all six.

That page competes for six different searches at once, says little about any of them, and loses to businesses with a dedicated page each. It also gives Google nothing to distinguish, so it tends to rank for the most generic version of the search, which is the one with the least buying intent.

The opposite fault is nearly as common. A business creates ten near-identical pages, one per city, with the place name swapped and nothing else changed. Google recognises the pattern and discounts it, and increasingly ignores the pages entirely.

The rule that holds up: one page per thing you actually sell, and one page per place you genuinely operate, with something real and specific on each. Once there are enough pages to group, arranging them as a topic cluster around a broader page is what keeps them from competing, and deciding what gets written next is a content marketing question rather than an optimisation one.

Writing for a person and a machine at once

These used to be different jobs. They are converging, which is convenient.

• Answer first. Say the thing in the first sentence under the heading, then expand. Readers scanning on a phone get what they came for, and a machine extracting an answer has something clean to take. This is most of what AEO asks for.
• Plain words. Write the words your customers use, not the words your industry uses. If customers say “aircon” and you say “HVAC solutions”, you will not be found and you will not be understood.
• Headings that ask real questions. A heading reading “How long does it take?” matches how somebody searches. A heading reading “Timelines” does not.
• Self-contained facts. A sentence that only makes sense after reading the two above it cannot be quoted anywhere. State things whole.
• Something worth reading. The pages that hold their position over years are the ones that say something the other results do not.

What does not work any more

A great deal of advice still circulating describes an internet from fifteen years ago.

• Keyword density. Repeating a phrase a fixed number of times. It never worked as well as claimed and it now reads badly to both readers and search engines.
• Hidden text. White text on white backgrounds, keywords stuffed into footers. This is a policy breach, not a tactic.
• Meta keywords. A tag listing your keywords. Google has ignored it for well over a decade.
• Exact-match repetition. Writing “aircon cleaning Cebu” awkwardly into every paragraph. Search engines understand synonyms and context now.
• Spun or generated filler. Publishing many low-value pages to cover more terms. The volume is easy to detect and it drags the whole site down rather than only the weak pages.

Words you will hear, and what they mean

• Title tag. The clickable headline in search results.
• Meta description. The summary beneath it.
• H1 to H6. The heading levels that structure a page.
• Slug. The readable last part of a URL.
• Alt text. The written description of an image.
• Anchor text. The visible words a link is wrapped around.
• Canonical. A tag naming the preferred version of a page when several addresses show the same content.
• Cannibalisation. Two of your own pages competing for one search.
• Above the fold. What is visible before scrolling, which on a phone is a very small space.
• Content pruning. Removing or merging pages that add nothing, which often lifts the pages that remain.

How to tell whether a page is doing its job

Rankings are the least useful of the available numbers. These are better.

• Impressions and clicks for that page in Google Search Console, which also shows the searches it appeared for. Those searches are often not the ones you intended, and that is the most useful thing on the screen.
• Click-through rate. A page with many impressions and few clicks is ranking but not persuading. That is a title tag problem, not a content problem.
• What happens after the click. Whether the visit produces an inquiry, which requires tracking to be set up first.
• Which pages produce nothing. Every site has pages that have never produced anything. Knowing which they are tells you where to merge, improve or remove.