field note
first published 2026-10-09
What is technical SEO? A working definition after nineteen years
Technical SEO is the engineering that lets machines read, trust and rank a website: access, crawl, render, index, speed, structure. What it covers, what it doesn't, how it changed when the readers multiplied, and how to tell if yours is broken.
1,205 words · 5 min read · technical seo
Ask five people in the industry what technical SEO is and you will get five lists. Here is a definition you can act on: technical SEO is the engineering that lets machines read, trust and rank a website. Everything else about search decides whether a page deserves to rank. Technical SEO decides whether it is even eligible.
I have been doing it since 2007, first inside agencies and for the last decade and a half as the person a company calls when the agency's report did not move anything. This is what the discipline actually covers, where its edges are, and what changed in the last two years.
The six things it covers
Access. Can the crawler reach the page at all? Robots rules, firewall and bot-management products, status codes from the origin rather than the cache, and now the separate question of which AI crawlers are allowed in. Access failures are binary and brutal. A site that blocks Googlebot by accident does not rank badly; it disappears.
Crawl. Once inside, can the crawler find everything that matters and nothing that doesn't? Sitemaps that tell the truth, internal links that reach every important page, a written policy for filters and parameters so a catalogue does not mint a million near-identical URLs. Crawl is where budgets get spent, and large sites feel it first.
Render. What the machine reads is the raw HTML plus, for Google, whatever JavaScript it chooses to execute later. If the product data, the body copy or the links only exist after a script runs, Google sees them late and the AI crawlers mostly never see them. In 2026 this moved from a footnote to a gate.
Index. Of the pages Google crawled, which did it keep, and which URL did it choose as the canonical? Search Console's Pages report is a map of every disagreement between what you meant and what Google decided. Canonicals, hreflang, thin pages, soft 404s and accidental duplicates all live here.
Speed. How fast the server answers and how fast the page becomes usable, measured in the field by real visitors and in the lab by you. Core Web Vitals are the scoreboard; the causes are usually server response, blocking resources, heavy images and layout that shifts. Speed is found at the root: one UK retailer's server response went from 2.2 seconds to 47 milliseconds without touching the front end.
Structure. Headings that describe the content, titles that match the page, structured data that forms one connected graph rather than a scatter of plugin output, and a site architecture that puts important pages close to the top. Structure is what lets a machine say what a page is about without guessing.
What it does not cover
Technical SEO is not content strategy, and it is not link building. It will not write the category page that finally answers the question people are asking, and it will not earn the citation from the trade publication. What it does is make sure that when those things exist they are read, indexed and shown. Content on a site with a broken canonical layer is a well-written letter with the wrong address.
It is also not "SEO for developers" in the sense of a checklist bolted on at the end of a build. The cheapest technical SEO is the kind decided before the first commit: URL patterns, rendering strategy, how the data layer feeds the page. Retrofitting any of those costs ten times what deciding them would have.
What changed when the readers multiplied
For most of my career the reader was Google, with Bing in the corner. The site had one audience of machines, and the technical discipline was tuned to it.
That is no longer the situation. ChatGPT, Perplexity, Claude and Google's own AI features all fetch and read websites, and they bring their own crawlers with their own habits. The guestbook on this site's lab page counts every one of them by product; on an ordinary day Googlebot is still the most frequent visitor, and the AI search crawlers together are not far behind. The HTML they read is the same HTML. The weighting is different.
Three things rose in importance. Extractability: the answer has to be present in the raw HTML, near the top, in plain prose, because the engines mostly will not render your JavaScript to find it. Entity clarity: the structured-data graph has to describe who and what the page is about, and agree with the visible copy, because the engines cross-check. And crawler policy: each bot's access has to be a decision, because a firewall rule meant for scrapers now quietly removes you from the answers.
Technical SEO did not split into two disciplines. It grew a section, and the audits that skip the section are auditing the web of 2019.
How to tell if yours is broken
You do not need an audit to spot the big ones. Five checks, fifteen minutes:
- Open Search Console's Pages report. If "Crawled, currently not indexed" or "Duplicate, Google chose different canonical" holds pages you care about, the index layer is arguing with you.
- Fetch your most important page with JavaScript disabled, or with
curl. If the body copy and the product data are missing from the response, the render layer is a problem for every machine except, eventually, Google. - Time your origin with the CDN bypassed. Over a second to the first byte means there is a server-side cause waiting to be found.
- Run the page through a structured-data validator. Errors mean the graph is being ignored; islands of unrelated blocks mean it is not a graph.
- Run the site through the Agent-Ready Grader on this site. It is free and ungated and checks the three gates that decide whether an AI engine can read you at all.
If any of those fails, the fix is almost always structural, and it is almost always smaller than the backlog you were handed.
Who does the work
Three options, honestly compared.
An agency gives you a team and a process and a monthly deck. The audit is usually good; the implementation goes into your developers' queue, where it competes with everything else. Many agency audits I have inherited were correct and unimplemented.
Your own developers can do the work if someone tells them exactly what and why, in their language, as tickets with evidence. Most SEO reports do not, which is how good developers end up ignoring correct findings.
A technical SEO consultant who ships sits in the gap: the diagnosis from the instruments, the fixes as pull requests against your repository, the readouts afterwards from the same instruments. It is the shape I have worked in for most of nineteen years, and it is the reason the person who found the problem on the sites in my case files is the same person whose commits you can read.
The definition, once more
Technical SEO is the engineering that lets machines read, trust and rank a website. Six layers: access, crawl, render, index, speed, structure. One new reader every few months. And one test that has not changed in nineteen years: fetch the page the way the machine does, and see what is actually there.
asked straight — answered straight
01What is technical SEO in simple terms?
It is everything that decides whether a machine can reach your pages, read them correctly, understand what they are about and show them quickly. Content and links decide whether you deserve to rank; technical SEO decides whether you are eligible to.
02Is technical SEO the same as on-page SEO?
They overlap at the page's structure — titles, headings, internal links, structured data — but technical SEO goes below the page: server behaviour, crawl paths, rendering, indexation rules, speed. On-page work on a site that cannot be crawled is wasted.
03Do I need a developer for technical SEO?
For most findings, yes, because the fixes live in templates, server configuration and code. That is why I ship the fixes myself rather than handing a report to a queue. If you have a developer, the audit should speak their language and arrive as pull requests, not slides.
04How is technical SEO different for AI search?
The engines read the same HTML Google does, so the foundations are shared. What changes is the weighting: content that only appears after JavaScript runs, a structured-data graph that contradicts the page, or a crawler blocked by a firewall rule now costs you citations as well as rankings.
next step — seo consultant
SEO consultant who ships.
Nineteen years of search, shipped by the person who diagnosed it. Send a brief — a few lines about the site, the problem and the evidence you have — and the reply is a straight read on fit and shape, from the person who does the work.
no forms · no funnels · jamie@jamiemckaye.com
written by
Technical SEO consultant and full-stack developer in Hersham, Surrey, in practice since 2007. One person, no handoffs: the audits, the code and the writing come from the same pair of hands. This site is the working proof — it grades itself on the same instrument it runs for clients.