WordPress · Construction

FoxBud

A lead-generation site for a fit-out company, readable for search engines and AI

FoxBud fits out and finishes offices in Warsaw, Wrocław and Kraków. The new site generates leads from two audiences, property managers and office managers, on a custom WordPress theme and plugin with complete structured data for search engines and language models.

Role: Sole developer: 124 of 125 commits, 28 pull requests
Period: Jul - Sep 2026
Stack: WordPress
PHP
Custom theme and plugin
ACF
schema.org
llms.txt
Cloudflare

custom post types: 4
schema.org types in the graph: 18
AI bots with an explicit policy: 17
months until leads / consents are deleted: 24 / 48

Context

The site was rebuilt from scratch on a custom theme, with two service paths written in different tones: fit-out for property managers and office finishing for office managers, plus a free cost-estimate audit.

Old service addresses redirect permanently to the new service pages.

What I did

Classic PHP theme and a custom plugin, no page builder and no build step
Four custom post types: services, projects, case studies and a knowledge base, with ACF fields registered in code
Own form handling: leads saved to a database table, with the consent record stored next to each lead
schema.org graph extended from ACF data: services with offers and prices, FAQ, articles, breadcrumbs, places and audiences
llms.txt generated from the database on every request
An AI bot policy: 8 search bots and 8 training bots allowed by the client's decision, one bot blocked
Daily retention job: leads and their attachments deleted after 24 months, consent records after 48 months

How it fits together

From content to search engines and AI

Content
1. Services
2. Projects
3. Case studies
4. Knowledge base
Output
1. Pages
2. schema.org graph
3. llms.txt
Readers
1. Visitors
2. Search engines
3. AI assistants

Leads and consents are stored together and deleted on schedule.

Hard parts and how I solved them

Problem
After each deploy visitors saw stale pages, because the server cache and Cloudflare each kept files for 7 days.
Solution
The cache is purged automatically when the theme version changes, and asset URLs include the file's modification time.
Problem
On iOS Safari some words ran together. Text pasted from Figma carried an invisible U+2028 line separator.
Solution
Cleaned 27 fields in one batch and strip the character on every save, so it cannot come back.
Problem
Content edits made by the client kept coming back after deploys, because content migrations re-ran every time.
Solution
Each content batch now runs once per install and records that it ran.

Results

Content model for 26 projects, 23 knowledge-base articles, 3 case studies and 3 service pages
16 content batches delivered as versioned migrations that run once per install
Structured data with 18 schema.org types, built from the same fields editors fill in

The site is waiting for launch, so there is no live link yet.