Speed & Search (SEO & GEO)

Last updated

Slow pages, junk code, and scripts nobody uses cost real money. Not in a theoretical Core Web Vitals sense — in the sense that someone clicked, waited, and left before your offer rendered.

This is the work of finding out why, fixing it, and then making sure both search engines and AI assistants can actually understand what your business does.

Speed

Site speed is usually not one big problem. It is a dozen small ones that nobody has measured.

What I actually look for:

  • Render-blocking resources — scripts and stylesheets that stop the page drawing while they load
  • Assets that cannot be cached — a surprisingly common own goal, where every visit re-downloads files that never change
  • Third-party scripts — analytics, chat widgets, pixels, and vendor telemetry, each costing time, several usually unused
  • Images — wrong format, wrong dimensions, no lazy loading below the fold
  • Fonts — blocking first paint, or loading weights the design never uses
  • Duplicate work — the same library loaded twice by two plugins that do not know about each other

One from my own site

The site you are on replaced a WordPress build I wrote myself. When I audited it before the rebuild, two things stood out.

The theme was generating a version string from the current timestamp, so every asset URL was unique on every request. Nothing could ever be cached by anyone — a one-line mistake that quietly defeated the entire caching layer.

The host was also injecting its own performance-monitoring script that blocked rendering for the better part of a second, on every page, for every visitor.

Neither was visible without measuring. Both cost real time on every single page view. That is what this work usually looks like — not exotic, just unmeasured.

Search (SEO)

The technical foundation, done properly at build time rather than bolted on:

  • Structured data that describes your business as a linked graph — not three disconnected blocks that search engines read as unrelated duplicates
  • Clean heading structure with no skipped levels, so the page outline makes sense to a crawler and a screen reader
  • Metadata that is actually per-page — titles, descriptions, canonicals, and social cards that differ between pages instead of every page inheriting the homepage's
  • Sitemap and robots generated from the site itself, so a page added next year appears without anyone remembering to update a list
  • Local search — consistent business information, and the schema to support it

GEO — can an AI assistant read your site?

An increasing share of "who should I hire for this" now starts with an AI assistant rather than a list of blue links. That is a different discovery surface with different requirements, and most sites have not been built for it at all.

Practically, that means:

  • Content that renders on the server, so a crawler or agent sees real text rather than an empty shell that requires JavaScript
  • An llms.txt file — a plain-language, link-bearing summary of what the site offers and where things are
  • Structured data that matches what the page actually says, rather than claiming services the visible content does not mention
  • A clean accessibility tree, since the structure that helps a screen reader is the same structure an agent reads

This site does all four, and it is a reasonable reference implementation if you want to see what it looks like in practice.

How the work runs

  1. Measure first. Real numbers from real page loads, not a plugin's self-assessment.
  2. Rank by impact. Most of the gain is usually in two or three fixes. The rest is cleanup.
  3. Fix, then measure again. A change that was not verified did not happen.
  4. Leave it maintainable, so the next person adding a script knows what it costs.

I also built a browser-based SEO audit tool that runs this kind of technical check with no server and no signup, and an open-source auditor for automotive dealer pages built to run across large dealer groups at scale. Both are in Selected Work.

Who this is for

Businesses whose site performance directly hits revenue. Anyone told their site is "optimised" without being shown numbers. Teams who rank for their own name and nothing else. Businesses wondering why AI assistants never mention them.

I work with clients in Cincinnati and nationwide.


Want to know what your site is actually costing you? Send me the URL and what you think is wrong. I reply within one business day.