Guide
Technical SEO The checklist has grown a section
The classic list still holds: crawling, speed, markup, redirects, security. What changed is that a second kind of machine now reads your pages, and it asks for permission separately.
The list, and the part that is new
Nothing below has been retired. The point is the last row, which almost no audit template includes yet and which decides whether your page can be used in an answer at all.
- Crawling and indexing
- Can a crawler reach the page, is it allowed to, and does it end up in the index. Robots directives, sitemaps, canonicals, redirect chains.
- Speed and stability
- How fast the page answers and whether it moves about while loading. Core Web Vitals live here, and so does plain server response time.
- Markup and structure
- HTML that parses, one heading outline that makes sense, and structured data that validates rather than merely exists.
- Security and serving
- HTTPS without mixed content, certificates that are current, redirects that resolve in one hop rather than four.
- AI crawler access
- The new row. GPTBot, ClaudeBot, PerplexityBot and Google-Extended ask separately from Googlebot, and a robots.txt written before 2024 has no opinion about any of them. Blocking them is a decision; doing it by accident is the common case.
- Machine readability
- Whether a model reading the page without a browser can still work out what it says. This is where sites that pass every other row still fail.
Schema SEO, and why engines lean on it harder now
Structured data used to be worth doing for rich results: stars under a product, a recipe time, an FAQ accordion in the listing. That is still true and it is no longer the main reason.
An AI engine answering a question has to decide what your page claims. Prose requires interpretation, and interpretation is where it gets you wrong: the wrong price, last year's opening hours, a service you stopped offering. A labelled fact needs no interpretation. Schema is the difference between an engine inferring what you charge and an engine reading what you charge.
Which is why the common failure is not missing schema but schema that does not validate, or that disagrees with the visible page. Both are worse than none, because both get trusted and then get it wrong.
The detail sits on structured data for AI, and the schema checker reads what is on your page now and says what parsed.
What we check, and what we leave alone
Worth being plain about, because technical SEO is a large field and the free check on this site reads one page rather than crawling a domain.
It covers crawler access, markup and schema validity, response speed, security, and whether a model can state what your business does from the page alone. It does not cover redirect chains across a site, log file analysis, crawl budget, index bloat, international hreflang, or anything that needs a full crawl to see. For those, Search Console and a crawler such as Screaming Frog or Ahrefs do the job properly.
Questions about technical SEO
What is technical SEO?
Everything about a site that decides whether a machine can get to the content and make sense of it, as opposed to whether the content is any good. Crawling, indexing, response speed, redirects, canonicals, markup and security. It is the layer underneath the writing, and nothing above it works if this layer is broken.
What is schema in SEO?
Schema, or structured data, is machine-readable labelling added to a page saying what each part of it is: this is a product, this is its price, this is the author, this is the date. Search engines use it for rich results. AI engines use it more heavily than that, because a labelled fact needs no interpretation, and an engine that does not have to interpret you is far less likely to get you wrong.
Is technical SEO different from on-page SEO?
They overlap and the line moves depending on who is drawing it. The useful split is by who fixes it. Technical SEO is mostly the developer's list: server responses, redirects, markup, speed. On-page is mostly the writer's list: titles, headings, what the words actually say. Schema sits awkwardly in the middle, which is why it gets argued about.
Does technical SEO still matter if AI answers the query?
More than before, and that is the part people get backwards. A model cannot quote a page it was not allowed to fetch, cannot parse markup that does not close, and will not wait for a page that takes nine seconds. The classic technical list got shorter in importance for ranking and longer in importance for being used at all.
What does technical SEO cover that it did not five years ago?
Crawler access for AI agents, which is a separate permission from Googlebot and defaults differently. There are 12 named AI crawlers across 6 engines, and a robots.txt written in 2021 has an opinion about none of them. Add llms.txt, agent-readable files, and markup that has to survive being read without a browser.
Do I need a technical SEO agency?
For a small site, rarely. The list is finite, most items are one-time fixes, and a developer with the codebase in front of them will work faster than anyone you brief. Agencies earn their place on large sites where the same fault repeats across thousands of templated pages and somebody has to own the rollout.
Check the new row first
Crawler access and machine readability are the two most sites have never looked at. The free check reads both in under a minute.