On-Page SEO: Make the Pages You Already Have Work Harder
Most sites do not need more pages. They need the ones they have to be legible: to a reader deciding in five seconds, to Google deciding what the page is about, and to an AI engine deciding whether there is a sentence in it worth quoting. That is three audiences and one piece of work.
THE SERVICE
What on-page SEO covers
Content production creates pages that do not exist. On-page SEO fixes pages that already do.
On-page SEO is everything that lives on a page and affects whether it gets found: the title and meta description, the heading structure, the depth and shape of the body content, the entities it mentions, the schema attached to it, the internal links running out of it, and the images and their alt text. It is distinct from technical SEO, which decides whether a machine can reach and render the page at all, and from content production, which writes pages that do not yet exist. On-page work is applied to pages you already have. In 2026 it has two jobs rather than one: rank in search results, and be structured clearly enough that an AI engine can lift an answer out of it.
Why on-page matters more now, not less
The usual line is that on-page SEO is table stakes: fill in the title, write a meta description, use one H1, move on. That was roughly true when the only outcome was a blue link.
It stopped being true when engines started answering instead of listing. Around 60 per cent of searches now end without a click, and AI Overviews appear on a large share of queries. The page that gets quoted in those answers is not the one that stuffed the keyword in; it is the one where a clear claim sits near the top, supported immediately, in a structure a machine can take a clean passage from.
The good news is that this is not a second job competing with the first. Somewhere between 92 and 99 per cent of AI Overview citations come from pages already ranking in the top ten. Ranking and being cited are the same work done properly, which is why we do not sell them as two invoices.
The unglamorous version: your title tag is now a citation prerequisite, not just a click-through lever, because AI previews frequently reproduce it verbatim. And Google rewrites the majority of titles that are too long or do not match the page, which means a lot of sites are being represented in search by a sentence they did not write.
What is included
Titles and meta descriptions
Written to match what the page actually delivers, inside the length Google will display, and distinct from every other page on the site. This is the highest-leverage hour in on-page work and it is the one most often left to a plugin.
Heading structure and page shape
One H1 that states the subject, headings that describe sections rather than decorate them, and the main answer near the top rather than after four paragraphs of preamble.
Content depth and entity coverage
Keyword density stopped mattering; covering the concepts a subject genuinely involves did not. A page about a topic that never mentions the tools, standards and related ideas belonging to it reads as thin to a machine regardless of word count.
Answer shaping
Claims stated plainly and supported immediately, so there is something quotable. Comparison tables where a comparison is genuinely being made, because structured content gets cited noticeably more often than prose making the same point.
Internal links out of the page
Links that go where the reader would actually want to go next, with anchor text describing the destination rather than saying "click here". On a site above thirty pages this is usually the largest untapped gain available.
Per-page schema
Which structured data types the page genuinely supports, and whether the content backs the markup up. Article, FAQ, Service, Product and breadcrumbs, chosen rather than applied wholesale.
Images and media
Alt text that describes the image, file sizes that do not undo the performance work, and captions where they earn their place.
Content pruning and consolidation
Sometimes the fix is fewer pages. Two thin pages competing for one query beat neither, and merging them is usually faster than improving both.
THE BOUNDARY
Where on-page stops and something else starts
If a machine cannot render the page, this is not the service you need.
Content that only appears after JavaScript runs cannot be optimised into visibility. That is technical SEO and it comes first.
If the page should not exist, we will say so.
Some pages are better merged, and some are better deleted. Optimising a page nobody should land on is billable and pointless.
If the page does not exist yet, that is content production.
We do that too, but it is a different piece of work with a different price.
If you are chasing a term you cannot realistically win, on-page will not close the gap.
That is a keyword research conversation and an authority one, and we would rather have it before you commission anything.
OUR PROCESS
How we actually do it
Find the pages worth the effort
Not every page. The ones already ranking between five and twenty, where a page is close enough that structure decides the outcome, and the ones carrying commercial intent. Effort follows the money, not the alphabet.
Read the page against the query it is meant to serve
Usually the problem is visible in ninety seconds: the page answers a different question from the one being asked, or answers the right one too far down.
Rewrite what is on the page
Title, headings, the opening, the structure, the entity coverage, the schema and the links out. Delivered as changes you can review rather than a report you have to action.
Measure both channels
Rankings and clicks as usual, plus whether the page is now being cited in AI answers. Being ranked and being quoted are different outcomes and both are checkable.
What you get
We can hand this over for your team to implement, or implement it ourselves if you are on a platform we can reach. Either way it arrives as changes rather than as advice.
Who this works for
You rank on page two for things you should own.
The highest-return situation in SEO and the reason on-page still earns its place.
You publish plenty and rank for little.
Usually a structure and intent problem rather than a volume one, and more content will not fix it.
Your rankings held and your traffic fell.
The answer is being given without the click. The fix is being the page that gets quoted.
You inherited a site with hundreds of pages and no system.
Titles written by three different people over six years, half of them auto-generated.
You have a content budget and want it spent on the right pages.
Frequently the cheapest answer is improving twenty existing pages rather than writing twenty new ones.
Questions, answered straight
No fluff, as promised.
What is the difference between on-page and technical SEO?
Technical SEO decides whether a machine can reach, render and index the page. On-page decides whether it is worth ranking once it gets there. Technical comes first: no amount of on-page work rescues a page a crawler cannot read.
Is on-page SEO still worth doing in 2026?
More than before. When engines answer instead of listing, the page that gets quoted is the one with a clear claim near the top in a structure a machine can lift from. That is on-page work, and it did not exist as a discipline three years ago.
How long do titles and meta descriptions need to be?
Titles under about sixty characters, because Google truncates beyond that and rewrites the majority of titles that are too long or do not match the page. Meta descriptions are not a ranking factor but they decide click-through, and they are worth writing rather than generating.
Does word count matter?
Less than people think. Longer pages are cited somewhat more often, but structure and relevance matter considerably more than length. A focused 900-word page that answers the question beats a padded 2,500-word one that circles it.
Can we just use an SEO plugin or a tool score?
For basics, yes, and you should. A tool will tell you a title is missing. It will not tell you the title is present, technically fine and describing a page nobody wanted. Scores out of a hundred measure compliance, not whether the page answers the question.
Will you rewrite our content?
Where the fix is structural, yes: opening, headings, order, the addition of an answer near the top. Where the page needs to be written from scratch, that is content production and priced separately, and we will tell you which one you are looking at.
Can AI do this?
Parts of it, well. Drafting title variants and spotting coverage gaps are good uses of it and we use it for both. What it does badly is deciding which pages are worth the effort and whether a claim is actually true, and the March 2026 core update specifically penalised AI-generated content shipped without human oversight. We use it as a tool, not as the deliverable.
How many pages should we start with?
Fewer than you think. Twenty pages done properly will teach you more, and usually earn more, than two hundred touched lightly. We pick the ones where ranking position and commercial value say a change is worth making.
How quickly does it show?
Faster than most SEO work, because the pages are already indexed and already have history. Weeks rather than months for title and structure changes, longer where the fix is depth. It is still search, so nobody honest promises a date.
Do you implement the changes or just recommend them?
Either. On a platform we can reach, we implement. Otherwise you get the changes written out and ready to paste, rather than a PDF describing them. An on-page audit nobody actions is a cost with no outcome, which is the usual fate of this work.
Ranking on page two for something you should own?
Send us three URLs and the queries they are meant to serve. We will tell you what we would change and whether it is worth doing, before anyone talks about a contract.