How to Audit a Page for On-Page SEO in 15 Minutes

Most people who worry about SEO worry about it in the aggregate. The site isn’t ranking, traffic is flat, something must be wrong somewhere. This kind of worry is useless because it has no unit of work. The trick that makes SEO tractable is to shrink the question to a single page, because a single page can be audited completely, by hand, in about fifteen minutes, with nothing but a browser. I do a mechanized version of this for a living — I built GazeSite to run these checks automatically across whole sites — but I still think everyone should do it manually at least once, the way you should do long division at least once before trusting a calculator. The manual pass teaches you what the checks mean.

So pick one page. Not your homepage — homepages are weird, committee-designed things. Pick a page that exists to answer a question or sell one thing. Then start where the crawler starts, which is not where you start: with the source, not the pixels. View source, or better, fetch the page with curl, because that shows you what the server sends before any JavaScript runs. The single most clarifying moment in a page audit is discovering how different the raw HTML is from what you see in the browser. If your main content isn’t in that raw HTML — if the body is an empty div waiting for a script to fill it — you’ve already found the most important thing about the page, and you can spend your remaining minutes confirming whether search engines are managing to render it anyway. Usually the answer is: partially, slowly, and you shouldn’t depend on it.

Now the head of the document, three checks. First the title tag. It should exist, be unique to this page, describe this page, and lead with the words a searcher would use. Read it out of context, as a line on a results page, because that’s what it is: your ad, written for a slot you don’t control. Second, the meta description. It doesn’t influence ranking, and it’s still worth two of your fifteen minutes, because it’s the sentence under your title in the results, and if you don’t write it, a machine excerpts something arbitrary. Missing meta descriptions are among the most common findings my crawler emits, not because people don’t know about them but because nobody owns them. Third, the canonical tag. Check that it points where you think it points. A canonical aimed at the wrong URL — the staging domain, the paginated variant, some other page entirely — silently tells search engines to credit all this page’s work to a different address. Thirty seconds to check, catastrophic when wrong.

Then the body’s skeleton. There should be exactly one H1, and it should say what the page is about, which is not the same as being clever. Heading levels below it should nest like an outline, H2s under the H1, H3s under H2s, because headings are the page’s table of contents to both crawlers and screen readers, and a page whose headings were chosen for font size reads as noise to both. A page with no H1 at all — another finding I see constantly — usually means the visual design has a big headline styled as a div, which is a way of whispering your main point to machines while shouting it at humans.

Next, intent, which is the check that separates mechanical audits from useful ones. Ask: what question does someone have when they arrive here, and does the first screen of the page answer it or stall? Search engines have gotten steadily better at measuring whether a page satisfies the query that brought the visitor, which means the old game of matching keywords is increasingly the game of actually answering the question. Read your opening paragraph as a stranger. If it’s three sentences of throat-clearing before the substance, the throat-clearing is an SEO problem now, not just a writing problem.

Then links, both directions. Outbound: click a few, especially old ones; dead references date a page instantly. Inbound and internal: ask what on your own site links here, with what anchor text. A page you care about that your own site barely links to is a page you’ve told crawlers not to care about either. And check the links out of this page — do they use descriptive text, do they lead where a reader would want to go next? A page that dead-ends, with no onward path, wastes whatever attention it earned.

Last, the two-minute mechanical sweep. Load the page on your phone, or narrow the browser window: does anything overflow, is the text readable without zooming? Check that images have alt text that says what the image shows — this is an accessibility obligation first, but it’s also the only text a crawler gets from your images. Confirm the page loads over https and that the http version redirects. Watch how long the page takes to become readable on a slow connection; you don’t need instruments to notice five seconds of blank.

Fifteen minutes, and you’ll end with a short list of concrete defects instead of a vague dread. Fix them, then do the next page. The compounding effect is real but it isn’t the main point. The main point is that auditing one page completely changes your defaults, so the next page you create doesn’t have these problems to begin with. Tools like mine exist because you can’t hand-audit five hundred pages. But the reason to hand-audit one is that afterward, you understand what the tool is telling you — and more importantly, you stop producing the findings in the first place.