Your site is already the knowledge base

A feature

No separate help centre to write. Helpie indexes the pages, products and posts you already have, including the content your page builder hides.

The reason most businesses never launch an AI assistant is the homework. Every tool wants a knowledge base, and building one means writing your entire site again in a format the tool likes.

You already wrote it. It is your site.

What it indexes

Pages and posts, out of the box. Products too, when you are running WooCommerce. Documentation, if you keep it in WeDocs.

Beyond that, Helpie discovers your custom post types automatically. A recipe type, a venue type, a case-study type, whatever your site is actually made of, all available to be indexed.

Every type is a switch you control. The obvious ones are on to begin with. Everything else is off until you decide, because you know which of your post types are content and which are plumbing.

Page-builder layout templates are excluded automatically, since nobody wants the assistant quoting a Bricks template.

Reviews and comments, if you want them

Two more sources sit available and switched off: your WooCommerce product reviews, and your Trustie testimonials.

Off by default because they change the character of the assistant. Turn them on and it can answer "what do people say about this?" from real customer words. Leave them off and it sticks to what you wrote yourself. Both are legitimate; the choice is yours rather than ours.

It reads the page, not the database

This is the part that makes it work on a real site.

Helpie loads each page the way a visitor would and reads what comes back, stripping out the navigation, header and footer so your menu is not indexed two hundred times.

That matters because most modern WordPress content is not in the field WordPress search reads. Bricks keeps its layout as JSON. Oxygen keeps shortcodes in postmeta. Divi writes bracketed markup. A tool reading the plain content field sees almost nothing on a builder-made page, which is to say almost nothing on a modern site.

Reading the rendered page sidesteps all of it. If a customer can see the text, Helpie can index it.

Where a page cannot be fetched, there are builder-specific fallbacks for Bricks, Divi and Oxygen, and the plain content field beneath that.

Custom fields come too

Helpie walks your ACF fields and folds their text in. Select fields index by their readable label rather than their stored value. Relationship fields index by what they point at. Repeaters, groups and flexible content are walked in full.

Spec tables, ingredient lists, tab content, delivery details: on plenty of sites the most answerable content lives in fields, not paragraphs.

Products carry their commercial detail

For a product, Helpie indexes more than the description. Price, including the sale split. Stock status. Weight and dimensions in your units. Categories, tags and attributes. Every variation’s SKU and options. Subscription terms where you sell subscriptions. Currency, stated explicitly.

That block is placed first, so it survives into the passage the assistant reads. "Is this in stock and what does it weigh" becomes answerable.

Structured data folded in

Where a page carries schema markup, Helpie reads the FAQ questions and answers, review text and ratings, and the meta description.

FAQ schema is particularly valuable, because it is already phrased as the exact question a customer would type.

Keeping it current

Publish or update a page and it is reindexed then. Delete or unpublish it and it leaves the index immediately, so the assistant cannot quote something you have taken down.

A nightly pass catches everything else: theme changes, plugin changes, a global block you edited once that appears everywhere. It compares a fingerprint of each page and skips anything unchanged, so the routine night is cheap.

Opting individual things out

Beyond per-type switches there is a per-item picker. Search for a specific page or product and exclude just that one.

Useful for the internal page that is technically public, the old campaign landing page, and the product you are phasing out but have not unpublished.

Reindexing on demand

There is a reindex control that runs in batches with a progress bar, rather than one request that dies halfway through on shared hosting. For after a bulk import or a migration.