
Most sites I audit do not have a content problem. They have a plumbing problem. Pages exist, they are useful, and Google either cannot reach them, will not keep them, or keeps a different version of them than the one the business cares about. Technical SEO is the work of removing that ambiguity.
This is the order I work in, and why the order matters.
1. Start with what Google already thinks
Before touching the site, I read the site the way Google reports it.
- Search Console Page indexing report: how many URLs are indexed, and the exact reasons given for the ones that are not.
- The URL Inspection tool on a handful of representative templates — home, category, service, blog post, and any paginated or filtered page.
- The Crawl stats report, to see whether crawl requests are being spent on real pages or on parameters, feeds and redirect chains.
Then I crawl the site myself with Screaming Frog and compare the two pictures. Where my crawl and Search Console disagree, that gap is usually the story: a template blocked in robots.txt, a canonical pointing somewhere unexpected, or a JavaScript-rendered navigation that a crawler never follows.
Rule of thumb: never fix a technical issue you cannot see reported somewhere. Otherwise you cannot prove it was worth doing.
2. Crawlability before anything else
Crawl issues silently cap everything downstream. What I check, in this order:
- robots.txt: no blanket disallow, no accidental blocking of CSS or JS (Google needs both to render), and no blocking of URLs you also expect to be indexed.
- Internal linking depth: every important page reachable within three clicks of the home page. Orphan pages get found by a crawl, not by intuition.
- Navigation implementation: links must be real anchor elements with href attributes. Buttons and click handlers are not links.
- Faceted and parameter URLs: decide deliberately whether each parameter creates a crawlable, indexable URL. Most should not.
- Redirect chains and loops: collapse every chain to a single hop.
- Server responses: soft 404s, 5xx spikes and slow responses under load all reduce how much of the site gets crawled.
3. Indexation is a decision, not an accident
For every template on the site I answer one question: should this URL be in the index, yes or no? Then I make the signals agree with the answer.
- Indexable: 200 status, self-referencing canonical, no noindex, present in the sitemap, internally linked.
- Not indexable: noindex via meta robots or the X-Robots-Tag header, excluded from the sitemap, and ideally not linked prominently.
The common mistake is mixing signals — a page that is noindexed but sitemapped, or canonicalised to another URL while also being linked as the main version. Google treats canonicals as a hint, so conflicting hints get resolved for you, and rarely the way you wanted.
Canonicals
Canonical tags exist to consolidate duplicates, not to hide thin pages. I check that:
- Every indexable page has a self-referencing canonical with an absolute URL.
- Parameterised, tracking and print variants canonicalise to the clean URL.
- Paginated pages self-canonicalise rather than pointing at page one.
- Canonical targets return 200 and are not themselves canonicalised elsewhere.
Sitemaps and robots
An XML sitemap should be a clean list of indexable, canonical, 200-status URLs, with accurate lastmod values. Nothing else. On larger WordPress and WooCommerce builds I split sitemaps by type so Search Console reports indexation per content type — that alone turns a vague number into an actionable one.
4. Core Web Vitals: LCP, INP, CLS
Performance is not a ranking silver bullet, but it is a real page experience signal and, more importantly, it decides whether the traffic you win converts. The three metrics that matter:
- LCP (Largest Contentful Paint) — how quickly the main content appears. Usually fixed by serving properly sized modern-format images, preloading the hero image, removing render-blocking CSS and JS, and putting a cache and CDN in front of the site.
- INP (Interaction to Next Paint) — how quickly the page responds to input. This replaced First Input Delay as the responsiveness metric, and it is almost always a JavaScript problem: too many third-party scripts, heavy event handlers, long tasks on the main thread.
- CLS (Cumulative Layout Shift) — visual stability. Reserve dimensions for images, ads and embeds, and load fonts with a strategy that does not reflow text.
On WordPress specifically, the biggest wins are usually unglamorous: fewer plugins doing overlapping jobs, a caching layer, image optimisation at upload time, and deferring anything that is not needed for first paint. Measure with field data in the Core Web Vitals report, not only with a single lab test.
5. Structured data that earns something
I add schema when it maps to a real feature or clarifies a real entity, not for its own sake:
- Organization or LocalBusiness on the site level, with consistent name, URL, logo and contact details.
- Product with genuine price and availability on commerce pages.
- Article on editorial content, with headline, author and dates.
- FAQPage only where the questions are genuinely on the page and useful.
- BreadcrumbList to reflect real site structure.
Every implementation gets validated with the Rich Results Test, and then monitored in the Search Console enhancement reports. Invented review markup or FAQ stuffing is a fast route to a manual action — do not do it.
6. Rendering and AI-era retrieval
Two things have made rendering matter more, not less. First, JavaScript-heavy builds still lose content when hydration fails or is slow. Second, the systems generating AI Overviews and answer engines need to fetch, parse and quote a page cleanly. That rewards the same fundamentals: server-rendered or pre-rendered HTML, semantic headings, one clear H1, self-contained answers near the top of the page, and text that is present in the raw HTML rather than assembled after the fact.
If you want to be quoted, be quotable in plain HTML.
7. Prove it landed
A technical fix is not finished until it shows up in data. After each change I:
- Re-crawl and confirm the specific issue count dropped to zero.
- Use URL Inspection with Request indexing on the key templates.
- Watch the Page indexing report for the reason category that was failing.
- Track impressions and average position for the affected URLs in the Performance report over the following weeks.
- Re-check field Core Web Vitals data, which lags by design.
The short checklist
- robots.txt allows what should be crawled, blocks what should not
- No orphan pages; important URLs within three clicks
- One canonical decision per template, signalled consistently
- Sitemap contains only indexable, canonical 200 URLs
- Redirects resolve in a single hop; no loops
- LCP, INP and CLS measured in field data and improving
- Structured data valid, honest and monitored
- Content present in server-rendered HTML
- Every fix verified in Search Console before it is called done
FAQ
How long does a technical SEO fix take to show results?
Crawling and reindexing can happen within days for important pages, but indexation and ranking shifts usually become readable over four to twelve weeks. Core Web Vitals field data moves slowest because it is based on a rolling window of real visits.
Is a technical audit worth it if my content is weak?
Yes, but sequence it. Fixing indexation first means the content work you do afterwards actually gets counted. Auditing costs a fraction of writing a content programme that Google never sees.
Do I need to fix every issue a crawler reports?
No. Crawlers report facts, not priorities. Fix what blocks crawling and indexation first, then what affects users, then cosmetic issues — and ignore anything that cannot be tied to either.
Want this run on your site?
Tell me the domain and the pages that matter. I will come back with what I would fix first and what it should move.
Start a project