Ken Bingham Full-stack developer · websites, CMS, and search Portfolio llms.txt Contact

Demonstration builds

Three small-business sites, built to be found and to be quoted.

Each of these is a complete website for a business I made up, built the way I would build it for a paying client: hand-written HTML and CSS, no page builder, no theme, no framework. They load instantly, work on a phone, read correctly in a screen reader, and carry structured data that matches what a human sees.

Ranking in a list of links and being named inside an answer are now two different jobs. These sites do both: conventional technical SEO, plus the entity clarity and self-contained answers an assistant needs before it will cite a business by name.

Built for HFB Technologies. The three businesses are invented, but the categories are not: a service trade, a hospitality operator and a nonprofit, because that is the work you do most. The AI-recognition pieces are here because it is the newest thing on your services list, and the hardest part to show with a screenshot.

The work

Open any one and view source

Every business here is fictional, including names, phone numbers, licenses and review counts. Each site says so in its own header and in its llms.txt, so nothing misleads a visitor or a crawler.

What is actually in them

The checklist each build passes

Found by search engines

  • One h1, headings in order, landmark regions, descriptive link text
  • Canonical URL and Open Graph metadata on every page
  • JSON-LD that matches the visible text, never contradicts it
  • robots.txt and sitemap.xml served from the domain root
  • No layout shift: nothing moves as fonts and assets arrive
  • Real text, never facts baked into an image or a PDF
  • Local signals: identical name, address and phone everywhere

Quoted by AI assistants

  • Answer blocks that still identify the business when lifted alone
  • Numbers, not hedges: prices, hours and radii as concrete figures
  • llms.txt per site, a plain-text brief of the facts worth citing. Google has said Search ignores the file and no major assistant has committed to reading it, so it is a cheap bet and a tidy fact sheet, not a ranking tactic. Claiming otherwise is how agencies lose trust.
  • Entity clarity: what it is, where it is, who it serves, in the first paragraph
  • Questions phrased as people ask them, with the answer directly beneath
  • Assistant crawlers allowed explicitly in robots.txt
  • Honest labeling so a model never mistakes a demo for a real business

How these were built

AI as leverage, with a human accountable for every line

These went from blank folder to deployed in a working day. That speed comes from using AI where it is strong, and from checking everything it produces, because a confident wrong answer costs more than it saves.

01 · Direction

Decide the design before generating anything

Palette, type pairing and layout are chosen first, per business. That is why an HVAC dispatch board and a dawn balloon sky do not look like the same template with different colors.

02 · Build

Generate fast, then edit hard

AI writes markup and copy at speed. I cut what is generic, replace filler with specifics a real owner would say, and make sure every number in the schema matches the number on the page.

03 · Verify

Check it rather than trust it

Rendered and read at phone width, structured data validated, contrast and keyboard paths checked, links followed. Nothing ships on the assumption that it was written correctly the first time.

Built by Ken Bingham

Full-stack: Java and Spring Boot, C# and .NET, Angular and Node, WordPress and CMS work, with Linux and VPS administration. This showcase is hand-built and self-hosted on my own server.

See the full portfolio