Many WordPress sites were built ten years ago, when nobody planned for AI or machine learning. They still run the business: content, customers, orders. Machine learning does not replace that site. It lets the business use what the site already has to serve its clients better. The developer’s job is to bring it in without rebuilding everything.
What it is
Machine learning is software that learns patterns from data instead of following rules written by hand.
- Rules: “if the order total is over 100, show free delivery”.
- Machine learning: “customers who bought this usually also buy that”, learned from past orders.
For a business this means decisions based on its own history: what clients read, ask, buy and search for.
Why it matters for a business on WordPress
- The data already exists. Years of posts, orders, comments, searches and form submissions are the raw material. Most sites store it and never use it.
- Clients expect it. Search that understands meaning, relevant suggestions and fast answers are now the norm on large platforms.
- It changes the role of the site. From a place that publishes to a place that understands its visitors and responds to them.
- No rebuild needed. ML works next to the existing site: data stays in WordPress, models run as services, results come back into pages and the admin.
How it fits an existing WordPress project
flowchart LR
subgraph WP["Existing WordPress site"]
DATA["Data<br/>posts, orders, users, searches"]
UI["Pages and admin"]
end
JOB["Background job<br/>cron or queue"]
ML["ML service<br/>model or API"]
STORE["Results<br/>tables, meta, cache"]
DATA --> JOB --> ML --> STORE --> UI- Pick one use case with a clear business result.
- Prepare the data the site already has.
- Run the model outside the request path: in a background job or an external service.
- Store the results and show them on pages or in the admin.
- Measure whether clients actually use them, then take the next case.
Segments and domains
Five types of projects where machine learning gives value first.
| Domain | What the site does | Where ML helps first |
|---|---|---|
| Learn hub | Teaches through articles, videos, practice | Learning paths, search by meaning, content gaps |
| B2C service business | Sells a service through requests and consultations | Lead qualification, routing, demand by season |
| WooCommerce store | Sells products online | Recommendations, product search, stock planning |
| Culinary portal | Publishes recipes and food content | Recipe suggestions, search by ingredients, auto-tagging |
| Information site for lawyers | Publishes news and analysis by area of law | Classification by topic, related materials, alerts |
Learn hub
- Suggest the next material based on what the reader has already studied.
- Search by meaning, not exact words: “how to run code on a schedule” finds the WP-Cron article.
- Find content gaps from the questions readers ask.
- Compare what the market asks for with what the hub covers.
B2C service business
- Score incoming requests: which are ready to buy, which need information first.
- Route requests to the right person automatically.
- Forecast seasonal demand to plan capacity and campaigns.
- Answer repeated questions before they reach a manager.
WooCommerce store
- “Frequently bought together” and personal recommendations from real orders.
- Product search that understands intent and synonyms.
- Demand forecasting for stock and seasonal assortment.
- Detect unusual orders and returns.
Culinary portal
- Recipe suggestions by taste, history and season.
- Search by the ingredients a visitor has at home.
- Automatic tagging by cuisine, diet and cooking time, with editorial review.
- Collections built from what readers actually cook.
Information site for lawyers
- Classify new materials by area of law automatically.
- Show related materials across years of publications.
- Notify readers about new materials in their practice areas.
- Search by meaning across the archive.
Accuracy matters most here: ML suggests, an editor confirms.
What the developer should be able to do
- Translate a business goal into one measurable ML use case.
- Find and prepare the data the WordPress site already holds.
- Choose between a ready API and a custom model.
- Run ML outside the request path and store results in WordPress.
- Keep personal data under control and within the law.
- Measure the effect and explain it to the business.
Challenges
| Challenge | What to do |
|---|---|
| Messy data | Clean and structure it first; often the biggest part of the work |
| Personal data | Decide what may leave the site; anonymize where possible |
| Cost | Start with ready APIs; build custom models only when they pay off |
| Trust | Show suggestions, not decisions, where errors are expensive |
| No measurable effect | Define the metric before you start |
FAQ
Do we need to rebuild the site?
No. ML runs next to the existing site and returns results into it.
Do we need our own model?
Usually not at the start. Ready services cover most first use cases.
How much data is enough?
It depends on the case. Years of content or orders are usually enough to start.
Where is our data processed?
Where you decide. This is part of the design, not an afterthought.
How is ML different from an LLM?
An LLM is one kind of model that works with language. ML is the wider field: recommendations, forecasts, classification, search.