On-page SEO is everything on the page itself that affects whether it can be found and understood: what it says, what it’s called, how it’s arranged, and how directly it answers the question behind the search. It’s the layer of SEO you control most completely, and the one that needs the least specialist help. Most of it is writing and editing decisions rather than technical ones.
In plain English
Think of a printed application on a hiring manager’s desk, in a stack of two hundred, with about four minutes each.
You can be the strongest candidate in the pile and still lose, because the one thing the manager needed to see is on page three, the headings don’t match the job that was advertised, and the first paragraph describes the company rather than the work. Nothing in the document is false; it’s just not organized for the person reading it.
On-page SEO is organizing the document for the reader. Search engines are reading it the same way, and for the same reason: to work out quickly what this is and whether it fits.
Why this matters for your site
This is the only layer you can improve today, without a budget, a developer, or anyone’s permission. It’s also where the fit between your page and the search intent behind a query lives, meaning the actual goal someone had when they searched. A page that misses that fit doesn’t rank badly. It doesn’t compete at all.
The other reason it matters is that it’s cheap to get wrong at scale. One badly built page template repeats the same mistake across four hundred pages, and fixing the template fixes all of them at once.
How it actually works
The title. Google generates the headline shown in results using several sources, including the title element in your page’s code and the headings on the page. Its advice is that a good title is unique to the page, clear, concise, and an accurate description of the contents. Most content management systems turn the title you write into that element automatically.
The description. Below the headline, results show a short description. Google says this comes from the page’s own content, and occasionally from the meta description you wrote, a one or two sentence summary in the page’s code.
The structure. Headings break a page into sections and help people navigate. Here Google is more relaxed than the industry: it says heading order doesn’t matter from Search’s perspective, and there’s no ideal number of headings. It also notes that semantic order helps screen readers, which is the real argument for doing it properly.
The content. Google’s published attributes for content that performs are worth reading literally: text that’s easy to read and well organized, content that’s unique rather than a rehash of what others published, content kept up to date, and content written for people first. Its guide to generative AI features sharpens this into a distinction worth internalizing. Commodity content is the sort of piece anyone could have written from general knowledge. Non-commodity content is the piece only you could write, because you did the thing. Google’s own contrast is between a generic list of tips for first-time homebuyers and a specific account of waiving an inspection and the money it saved.
The words your readers use. People search in their own vocabulary, not yours. Google’s example is that some people look for charcuterie and others for a cheese board. Same food, two audiences. A page written entirely in industry language is invisible to everyone outside it.
The links on the page. Link text, meaning the visible words you link on, tells both people and search engines what’s on the other side. Google says the vast majority of new pages it finds every day come through links, so this is a discovery mechanism as well as a description.
The images. Alt text is a short description of an image, added in the code, that explains the image’s relationship to your content. Google recommends placing images near the text they relate to, which helps it understand what the image means in context.
The web address. Google says parts of the address can appear in results as a breadcrumb trail, so words a person would recognize are more useful than random identifiers.
What this looks like in practice
A hotel has a page titled “Rooms” containing a photo gallery and about forty words of copy.
Everyone who searches for a room wants to know four things before booking: what it costs, how big it is, what the view is, and whether they can cancel. Three competing hotels answer all four on the page. This one answers none of them, and it loses without anyone doing anything clever.
The fix involves no technical work and no new software. Split the page by room type, title each one the way a guest would describe it, answer the four questions in the first screen, and let the gallery support the text instead of replacing it.
What people usually get wrong
“Exactly one H1 and headings in strict order.” Google says order doesn’t matter for Search and there’s no ideal number. Do it for accessibility, which is a real reason, not for rankings, which isn’t.
“Longer pages rank better.” Google states that length alone doesn’t matter and there’s no magic word count. Long pages often win because thorough answers win, not because they’re long.
“Write the perfect meta description and Google will show it.” It might. Snippets come mostly from the page’s actual content.
“Repeat the phrase a few more times.” Excessive repetition of terms is against Google’s spam policies. Google also says its language systems can connect a page to many searches even when the exact words are absent, so the practice is both risky and unnecessary.
What you can do about it
Write a title that names the thing, on every page. If two pages could share a title, at least one of them is wrong.
Answer the question in the first two sentences. Not after the company introduction. The reader who leaves in eight seconds never reaches the good part.
Turn your headings into the questions your customers actually ask. This forces the page into the reader’s order rather than your organization’s order.
Add alt text to images that carry meaning, describing what the image shows and why it’s there.
Give each page one subject. Two subjects on one page means the page competes with itself, the problem usually called keyword cannibalization, and satisfies neither reader.
Read the finished page out loud. Anything you stumble over was written for a search engine, and it stopped working on those years ago.
Words you will hear
Above the fold. The part of a page visible before scrolling. Where your answer belongs.
Page template. The reusable layout a content management system applies to a group of pages. Fix it once, fix them all.
Alt text. A short written description of an image, in the page’s code.
Word count. The number of words on a page. Not a target, despite being treated as one.
For search console, query and SERP, see What is SEO? For crawler, index and canonical, see How search works.
