Where on-page ends and technical begins
People argue about this boundary far more than it deserves. A title tag is on-page. Page speed is technical. robots.txt is technical. Schema is both, depending on who you ask.
The distinction earns its keep for one reason: staffing. On-page is an editor's job, technical is a developer's job, off-page is an outreach job. Three people, three calendars — and misfiling a task means it sits in the wrong queue for a month.
| Bucket | Examples | Who owns it |
|---|---|---|
| On-page | Content, title, headings, internal links, URL, alt text, page-level schema | Editor or content lead |
| Technical | Crawling, indexing, site speed, Core Web Vitals, redirects, sitemaps, rendering | Developer |
| Off-page | Earned links, digital PR, brand mentions, citations, reviews | Outreach or PR |
The 12-point checklist, in the order we actually run it
The order is the value here. Most checklists list twelve things alphabetically; running them in the wrong sequence means you polish tags on a page whose premise was wrong. Steps 1 to 3 happen before anyone writes a sentence.
- Confirm the intent behind the query. Search it, look at what ranks. If the results are all comparison pages and you planned a service page, you've already lost — see search intent.
- Check nothing on the site already targets it. Filter the query in Search Console. If an existing page has impressions, update it rather than build a rival — that's where keyword cannibalization starts.
- Set the [URL slug](/glossary/url-slug) before publishing. Lowercase, hyphenated, three to five words. Changing it later costs more than getting it right now.
- Write the [H1](/glossary/h1-tag) for the reader who already clicked — plain language, one per page.
- Build the heading outline before the prose. H2s as real sub-questions, H3s nested properly, no skipped levels. If the outline doesn't hold, the article won't either.
- Write the body to answer the question fully, including the sub-questions the ranking pages forgot. Topic coverage beats keyword repetition.
- Add the specifics. Numbers, ranges, prices, timeframes. Real figures outperform "it varies" every time, because figures are what gets quoted.
- Write the [title tag](/glossary/title-tag) last, to pixel width, front-loading the query wording. Last means it describes what you actually wrote.
- Write the meta description for the query, not the company — or leave it blank on long-tail pages and let Google pick per query.
- [Alt text](/glossary/alt-text) on every meaningful image, empty
alt=""on the decorative ones. - Add three to eight contextual links out, and link into the new page from two or three pages that already rank. See internal linking.
- Add the schema type that matches the page — Article, FAQPage, Product, BreadcrumbList. Only mark up what's visibly there.
Which on-page changes move rankings fastest
Not all twelve are equal. If you have one afternoon and a page stuck on page two, this is the order of expected return — and it's very different from the order most checklists imply.
- Fixing an intent mismatch. Rewriting a page that answers a different question than the searcher asked produces the largest single jumps we see. It also requires admitting the original page was wrong.
- Adding internal links into a page that had none. Often moves a page several positions within weeks, and costs half an hour.
- Covering the sub-topics the page skipped. Not more words — more of the specific questions the top results answer and yours doesn't.
- Merging two cannibalising pages. Slower, but it raises a ceiling rather than nudging a position.
- Rewriting the title tag on a page already between positions 5 and 15. Improves clicks first, relevance second.
- Everything else. Alt text tweaks, keyword density, URL changes. Hygiene, not levers. Do them because the page should be well built, not because you expect a jump.
How often to re-optimise a page
Pages decay. Competitors publish better versions, facts go stale, the query itself shifts. That's content decay, and it's normal rather than a failure.
A workable cadence: review any page that has lost roughly a fifth of its clicks quarter-on-quarter, plus your top twenty pages by impressions once a quarter regardless. Everything else waits until the data says otherwise.
The rule that saves the most pain: don't touch a page that's climbing. Rewriting mid-ascent resets whatever Google was working out, and you'll spend six weeks getting back to where you were. Wait for the plateau — the full case is in when to update old content instead of writing new.