How to run an SEO audit that changes something

Most SEO audits fail for the same reason: they produce a list of 340 issues sorted by tool category rather than by consequence, and the person who commissioned it reads the first page and does nothing.

A useful audit answers three questions. What is stopping pages being found at all, what is stopping the found ones from competing, and what should be fixed first given that nobody has unlimited time. Everything else is inventory.

Before you start: decide what you are auditing for

An audit with no question behind it produces a list with no priority. "Improve our SEO" is not a question. "Why did enquiries from search halve in March" is, and it points at a completely different set of checks from "we are about to redesign, what must survive the move".

That last row is worth taking seriously. If sessions are steady and enquiries are not, SEO is working and something after the click is not. Traffic but no customers is the diagnosis for that, and no amount of technical SEO will move it.

Stage 1: can Google reach and keep the pages?

Nothing else matters until this is true, and it is the fastest stage. You are answering one question: does the set of pages Google has match the set of pages you meant it to have?

• Search Console → Pages. Compare indexed against not-indexed, and read the reasons. This one report answers more than any crawler.
• robots.txt: fetch it and check what it blocks. Look specifically for a Disallow: / left over from staging.
• XML sitemap: does it exist, is it referenced in robots.txt, does it list live URLs only, and are those URLs the canonical versions?
• Canonical tags: does each page name itself, or is a template emitting the homepage on every page?
• noindex: check both the meta tag and the X-Robots-Tag header, which is invisible in the page source.
• Redirect chains: a URL that goes through three hops before arriving wastes crawl budget and loses a little at every step.

One number to write down at this stage: how many URLs Google reports as indexed. If you have 80 pages and Google has 4,000, you have a duplication problem generating URLs. If you have 80 and Google has 12, you have a discovery problem. Both are stage-one faults and both are invisible from the front of the site.

Stage 2: does each page declare itself properly?

This is the part most tools are good at, which is why most audits over-weight it. Get it right, then stop; the returns fall away quickly.

• Title: unique per page, describes the page rather than the company, and front-loads the meaningful words. Google rewrites titles it considers unhelpful, so a rewritten title is feedback.
• Meta description: not a ranking factor, but it is the advertisement for the click. A page with no description gets whatever Google chooses to pull.
• One H1 that matches what the page is about, and a heading structure that does not skip levels for styling reasons.
• Images: alt text describing the image, and dimensions on the tag so the layout does not jump when they load.
• Structured data that matches the visible content. Marking up a review that is not on the page is a policy violation, not a shortcut.
• Internal links using text that describes the destination.

Two things to actively not worry about. There is no ideal keyword density, and meta keywords have done nothing for over a decade. Both still appear in audit reports because they are easy to measure, which is not the same as being worth measuring.

Stage 3: is the content the right answer?

The hardest stage to automate and the one that decides most outcomes. For each page that matters commercially, search its target term and read what ranks.

1. 1Does the format match? If the results are all guides and your page is a product page, you are entered in the wrong race.
2. 2Does the page answer the question in the first screen, or does it warm up for four paragraphs first?
3. 3What do the top results cover that you do not? That list is your edit.
4. 4Is anything on the page actually wrong or out of date? Errors cost more than gaps.
5. 5Are several of your own pages targeting this same term? Pick one to win and point the others at it.

Stage 4: the experience of using the page

Performance and usability are real but secondary. They decide close contests; they do not enter you into one. Measure field data rather than lab scores, because Core Web Vitals are assessed from what real visitors experienced, at the 75th percentile over a rolling 28 days.

• Largest Contentful Paint at or under 2.5 seconds.
• Interaction to Next Paint at or under 200 milliseconds.
• Cumulative Layout Shift at or under 0.1.
• Usable on a phone without pinching, with tap targets that are not two pixels apart.
• HTTPS everywhere, with no mixed content warnings.

A PageSpeed Insights score of 68 is not a finding. A Largest Contentful Paint of 6 seconds for a quarter of your real visitors is. Why your website is slow covers the difference and why the two numbers disagree so often.

Stage 5: turn findings into an order of work

This is the stage that makes an audit worth commissioning, and the stage most audits skip. Every finding gets sorted by consequence and effort, not by which tool reported it.

An audit that lists 340 issues and does not say which twelve matter has moved the work onto the reader. That is the part they were paying somebody else to do.

How often is this worth repeating?

A full audit once or twice a year is plenty for most sites. What is worth doing monthly is much smaller: check Search Console for new indexing errors, confirm nothing important has started returning a 404, and look at whether the pages that earn enquiries are still doing so.

Audit immediately after a redesign, a platform migration, or a domain change. Those three events cause more sudden traffic losses than algorithm updates do, and the damage is usually done by a redirect map that was never finished.

Doing this without paid tools

You can get most of the way with Search Console, PageSpeed Insights and a handful of free checkers. What you cannot get free is reliable backlink data and search volume, because both need an index that costs real money to maintain. Any free tool claiming to give you those numbers is estimating, and it is worth knowing which of your numbers are measurements and which are guesses.

The free SEO check walkthrough is the thirty-minute version of this article using only free tools, and the tools list says plainly what each one can and cannot tell you.

Questions people actually ask

How long does a proper audit take?

For a 30-page business site, a focused day. For a large ecommerce site, a week or more, most of it spent on URL patterns and duplication rather than on individual pages. If a tool produced it in ninety seconds, you have a crawl report and not an audit.

Is an automated audit score worth anything?

As a way of tracking one site over time, yes. As a comparison between sites, not really, because every tool weights its checks differently and none of them are Google's weights. Treat the number as a thermometer rather than a diagnosis.

The audit found 200 issues. Where do I start?

With anything that stops a page being indexed or served. In practice that is usually fewer than ten items, and clearing them often moves more than the other 190 combined.

If the audit turns up problems that are structural rather than technical, the honest answer is sometimes that the site itself is the constraint. Ten signs a website needs rebuilding covers how to tell the difference between a site that needs fixes and one that needs replacing, and our website maintenance work exists for the first case.