A technically flawless, fully-crawlable site with vague titles and thin content still won't rank. Technical SEO gets a page in front of Google. On-page SEO is what convinces Google, and increasingly an AI model deciding what to cite, that the page actually deserves to be there. Here's what that work covers in practice.
The Checklist
1. One Primary Keyword, Matched to Real Intent
Every page should target one primary keyword, mapped to a specific search intent, informational, commercial, or transactional, not just picked for volume. A page trying to rank for five unrelated keywords at once typically ranks weakly for all of them rather than strongly for one.
2. Title Tags and Meta Descriptions That Earn the Click
The title needs the primary keyword and a genuine reason to click, not just a keyword stuffed in front of the brand name. The meta description doesn't directly affect rankings but directly affects click-through rate once a page does rank, a wasted opportunity if it's left generic or auto-generated.
3. Header Structure That Mirrors How People Actually Ask
H1 through H3 should follow a logical hierarchy, and H2s in particular perform better when phrased close to how a real person would ask the question, since that's both good for readability and easier for an AI system to lift as a direct answer.
4. Content Depth That Actually Resolves the Query
Depth isn't word count, it's coverage. A page should answer the question completely enough that a reader doesn't need to open a second tab to finish understanding the topic. Thin pages that half-cover a subject consistently lose to competitors who went one level deeper.
5. Deliberate Internal Linking
Internal links should connect related content on purpose, building topical authority and passing link equity toward priority pages, not scattered in as an afterthought. A well-linked cluster of related content outperforms the same content sitting in isolation.
6. Image and Media Optimisation
Descriptive alt text (for accessibility and image search, not keyword stuffing), compressed file sizes that don't drag down page speed, and structured media that adds signal rather than just page weight.
7. Structured Data Where It's Genuinely Applicable
FAQPage schema on pages with real Q&A content, Article schema on editorial content, Product schema on commerce pages. This connects directly to the broader technical SEO foundation the on-page work sits on top of.
A quick self-test for any priority page: read only the H1 and H2s, ignoring the body copy entirely. Could someone understand what the page covers and in what order, just from the headings? If not, the structure isn't doing its job, for a human skimming or a system trying to extract a specific section.
What's Changed for 2026
The mechanics above haven't changed much in years. What's changed is the audience: content now gets parsed by AI crawlers deciding what to cite, not just by Googlebot deciding what to rank. Well-structured on-page SEO already does most of the work AEO needs, direct-answer framing, clean headers, genuine depth, which is exactly why the two disciplines overlap as much as they do rather than requiring entirely separate content.
Getting Started
Audit your five highest-traffic pages against this list first. Fixing on-page issues on pages that already have some traction typically moves faster than optimising a brand-new page from zero, since the domain and page already carry some existing signal to build on.
FAQ
How many keywords should one page target?
One primary keyword per page, matched to a specific search intent, supported by a cluster of closely related terms that naturally belong in comprehensive coverage of that topic. Targeting multiple unrelated primary keywords on one page usually means the page ranks weakly for all of them rather than strongly for one.
How long should a page be for good on-page SEO?
Long enough to fully answer the query and no longer for its own sake. Word count is a byproduct of coverage, not a target itself, a page that pads to hit an arbitrary length without adding real value usually performs worse than a tighter page that actually resolves the searcher's question.
Does keyword density still matter?
Not in the mechanical sense it once did. Modern search engines and AI systems understand topical relevance through language patterns and entity relationships, not keyword repetition counts. Write naturally for the topic and the relevant terms appear where they belong; forcing a specific density typically makes content worse and can look manipulative.
How often should on-page SEO be revisited on existing pages?
For priority pages, review every 6 months against current rankings, competitor content, and whether the search intent for that keyword has shifted. Search behaviour and competitor content both evolve, a page optimised well a year ago can quietly fall behind without any change on your end.