There is a reliable way to tell whether an SEO engagement is doing anything: ask what shipped. Not what was recommended, prioritised, or presented — what changed in the codebase. A great many retainers produce a monthly document and nothing else, and the document is not the work. It is a description of work that somebody else is expected to do, and usually nobody does.

Technical SEO is unusual among marketing disciplines in that it is almost entirely engineering. That makes it easy to verify, which is exactly why it is worth being precise about what it actually fixes.

It fixes whether you can be found at all

Before ranking is a question, indexing has to happen, and a surprising number of sites quietly fail here. The recurring findings:

  • Content rendered entirely client-side, so what a crawler receives is an empty shell. Google can execute JavaScript, but it does so on a delay and a budget, and every other crawler — including the ones now feeding AI answers — is far less patient.
  • A noindex left over from staging. This is more common than it should be and it is completely invisible from the front end.
  • Robots rules blocking the CSS and JavaScript needed to render the page, so the crawler sees something broken and judges it accordingly.
  • Pagination and faceted navigation generating thousands of near-duplicate URLs, consuming crawl budget that should be spent on pages that matter.

None of these are subtle once you look. All of them are code changes, not recommendations.

It fixes which page ranks for what

The most common structural problem we find is not a missing page. It is two pages competing for the same query — a service page and a blog post, or three near-identical location pages generated years ago by somebody chasing a local SEO tactic.

Search engines have to pick one, and the one they pick is frequently not the one that converts. The fix is a keyword-to-URL map: one page owns each query, competing pages are consolidated or redirected, and internal links point deliberately at the owner. It is unglamorous and it routinely produces larger movement than months of new content.

Canonical tags are a hint, not an instruction. Two pages genuinely targeting the same intent is an architecture problem, and no tag fixes an architecture problem.

It fixes how fast the page becomes useful

Core Web Vitals get discussed as a ranking factor, which undersells them. As a ranking signal they are modest. As a conversion variable they are decisive, and the correlation between a slow page and a lost session needs no algorithm to explain it.

The work is specific and measurable:

  • Getting the largest above-the-fold element to load first rather than last.
  • Reserving space for images, embeds, and fonts so the layout stops moving under the reader.
  • Removing or deferring third-party scripts, which are almost always the single largest cost.
  • Serving modern image formats at the size actually displayed rather than the size uploaded.

Every one of those has a number attached before and after. If your SEO provider cannot produce both numbers, the work either did not happen or was not measured, and neither is acceptable.

It fixes what machines understand about you

Structured data is where technical SEO overlaps with something newer. Search results increasingly include answers assembled by systems that need to know what an entity is — not merely what words appear on a page. An organisation with a properly described identity, service catalog, and location is legible to those systems. One without is a collection of text that has to be inferred.

This is genuinely cheap to fix and widely skipped. It is also the part most likely to be "recommended" in a document and never implemented, because it requires touching the code rather than the content.

What technical SEO does not fix

Being honest about the boundary matters, because the discipline gets oversold and then blamed.

  • It does not create demand. If nobody searches for what you do, perfect technical execution ranks you first for nothing.
  • It does not substitute for content. A fast, crawlable, beautifully marked-up page that fails to answer the question will not hold a position.
  • It does not overcome an authority gap on its own. Against an established competitor on a head term, technical work narrows the gap; it rarely closes it alone.
  • It is not permanent. Sites decay. New pages, new plugins, new redirects. Without periodic re-checking, most of it comes back within eighteen months.

How to tell if yours is real

Three questions, and they are all answerable in a single email:

  1. What shipped last month? Not what was recommended. What is different in production, and where can I see the commit or the change record?
  2. Which page owns which query? If there is no map, nobody is managing cannibalisation, and cannibalisation is probably your largest single loss.
  3. What are the before and after numbers? For performance, for indexation, for the queries that matter. Impressions and average position are diagnostics, not outcomes.

If the answers are uncomfortable, you are paying for a document. Our technical SEO work is built the other way round — the deliverable is the change, and the report is just how you find out it happened. If you want the three questions above answered against your own site, send us the URL.