On-page: matching the job to the query

On-page work is the part people think SEO is, and it is the part most often done backwards — as decoration applied after writing, rather than as decisions about what the page is for.

Start from the job

You already know the query and its intent. The page's job is to be the best available completion of that job. Practically: does it answer the actual question quickly, does it cover what the competing pages cover, and does it add something they do not have?

That last part matters more than length. Information gain — something genuinely new, a first-hand test, real data, a clearer explanation — is what makes a page worth ranking above equally competent ones. Word count is not a strategy. Neither is covering every subtopic a tool suggested.

The elements that carry weight

• Title element. The single most important on-page signal, and also your headline in results. Say what the page is, in the words a searcher would use.
• Headings. Structure for the reader; they help a machine understand the shape too.
• The opening. Answer the query early. Burying the answer under preamble loses both readers and snippets.
• Internal links. How you tell a search engine which of your pages matter, and how readers move between them. Descriptive anchor text, pointing at genuinely related pages.

First-hand experience

Google's guidance leans hard on people-first content and, in its quality documentation, on experience — evidence the author has actually used the thing. For product, medical, financial and review content this is not decoration; it is the difference between ranking and not.

Pruning and consolidation

Sites accumulate thin pages that fragment their own relevance. Merging five weak posts into one strong page usually beats keeping five. Deleting genuinely worthless pages is a legitimate optimisation. This feels counter-intuitive and is repeatedly one of the highest returning things a mature site can do.

AI-generated content

Google's stated position is that it does not care how content is produced; it cares whether the content is helpful and reliable. In practice, generating at scale without editorial judgement produces exactly the commodity content its systems are built to discount — and "scaled content abuse" is a named spam policy.

What trips people up

• Keyword density. Not a thing. Has not been for a very long time.
• Writing for a word count.
• Publishing more instead of publishing better on a site already carrying dead weight.

You have got this when

You can look at a draft and say what it offers that the current top three do not — and if the answer is "nothing", you rewrite rather than publish.