On-Page SEO

The title does two jobs at once: it tells the ranking system what the page answers, and it tells the searcher why to choose your result over nine others. For the first job, name the topic in plain words near the front — truncation after roughly sixty characters means words at the end may never be seen. For the second, add the differentiator a generic competitor lacks: the year, the format, the audience, or the concrete output the page delivers.

The failure modes are all forms of vagueness. Brand-first titles waste the strongest position on the name people were not searching for. Duplicate titles make your own pages compete with each other and signal a template nobody reviewed. And titles that promise a different page than the one delivered earn a click followed by an instant return — a signal that quietly unwinds rankings over time.

Meta descriptions are not a ranking signal, and Google rewrites them often — but when your description is shown, it is the entire sales pitch for the click. Write it as a two-sentence answer preview: state what the page gives the reader, with the specifics a rewrite would lose — the number, the audience, the deliverable.

Keep it between seventy and a hundred and sixty characters, make it unique per page, and never let it repeat the title. A description that duplicates the title wastes the only two lines of persuasion you get. Pages where Google rewrites a well-written description usually have a mismatch between the query and the description, which is itself a finding worth reading.

One H1 per page, naming the topic. H2s beneath it as the section structure, phrased as the questions a searcher would ask next. H3s only where a subsection genuinely needs its own label. Read in isolation, the headings should tell a stranger what the page covers — because that is precisely how parsers, passage-ranking systems and screen readers consume them.

Two rules protect the structure. Never use a heading for styling; use a styled paragraph instead, so the outline stays honest. And answer each heading's question in the first sentence beneath it — passage-level retrieval slices on headings, and a paragraph that opens with throat-clearing produces an extracted passage that opens with throat-clearing.

Search intent is the job the searcher is hiring the result to do, and it comes in recognisable shapes: learn (definitions, guides), compare (versus, best, alternatives), do (calculators, generators, templates) and buy (products, pricing). Before writing anything, search the target query and look at what already ranks. The format of the winners is Google's revealed judgement of what the intent is.

If the first page is all calculators, no essay will rank for that query. If it is all two-thousand-word guides, a three-hundred-word page will not compete. Matching intent is not copying the winners — it is delivering the same job better: more current, more specific, more honest about limitations, or served as a tool instead of prose where a tool serves the intent faster.

Structured data tells the parser what each part of the page is: this block is a question and its answer, this is a person and their role, this is a step in a sequence. Its direct reward is eligibility for rich results — FAQ dropdowns, breadcrumbs, article treatments — which raise click-through from the same position. Its indirect reward is disambiguation: a machine that knows what a block is extracts it more reliably.

The one hard rule is fidelity. Mark up only what is visibly present on the page, with values that match the visible content exactly. Mismatched schema — an FAQ block in markup that does not exist in the text, a rating nobody can see — risks losing all rich-result eligibility for the site, which is a heavy price for a decoration.