Frequently bought together, cross-sells, order bumps and post-purchase offers from your order history and DEV3 AI similarity, with A/B tests and revenue attribution.
DEV3 AI Recommendations & Upsells learns what your customers buy together from your own WooCommerce orders (HPOS-safe), adds semantic similarity from Google AI embeddings, and shows the right products in every place that sells: product pages, cart, checkout (classic and block), the thank-you page and order emails.
- Co-purchase statistics with confidence and lift, updated on every paid, cancelled or refunded order; full rebuild as a background job.
- Semantic similarity from product embeddings (optional), with category/tag similarity as the fallback.
- 12 strategies: frequently bought together, similar, goes well with (hybrid), upsells, cross-sells, best sellers, trending, new, on sale, top rated, recently viewed, pinned.
- Per-product pins, exclusions and "hide widgets"; in-stock, category and already-bought filters; blend weights.
- Precomputed neighbour lists and a fragment cache: no AI call and no heavy query on page load.
- Product page bundle box (checkboxes, live total, optional bundle discount), grid, list or carousel.
- Cart widgets (classic and block) and order bumps in the classic checkout and the block checkout summary.
- Thank-you page post-purchase offer with countdown: add to the existing order (offline payments) or express second checkout, with signed single-use tokens.
- "You may also like" in WooCommerce customer emails.
- Gutenberg block and [dev3ru_recommendations] shortcode.
- A/B tests per widget (strategy, layout, copy, count, discount) with z-test significance and one-click winner.
- Revenue attribution: widget adds, click-through window, order item source, counted when paid and reversed on refunds.
- Dashboard with attributed revenue, CTR, add rate, revenue per 1,000 views, per placement and per widget.
- Offer copy writer with alternative angles in your brand voice.
- Complementary picks for products without order history (single product or background job), drafts you pin.
- Bundle pitch and a performance insights report.
- Every AI result is a draft you accept, edit or reject. The plugin is fully usable without AI.
- WordPress 6.0 or newer (tested up to 6.8)
- PHP 7.4 or newer
- WooCommerce 8.0 or newer (required)
- Optional: your own Google AI API key (from Google AI Studio) to switch on the DEV3 AI features. The rest of the plugin works without one.
- Download
dev3-ai-recommendations-for-woocommerce.zipfrom the latest release. - In WordPress, go to Plugins → Add New → Upload Plugin, choose the zip and click Install Now.
- Click Activate.
git clone https://github.com/dev3tech/dev3-ai-recommendations-for-woocommerce.git
cd dev3-ai-recommendations-for-woocommerce
composer install --no-dev
npm ci
npm run buildThen copy the dev3-ai-recommendations-for-woocommerce folder into wp-content/plugins/ and activate it.
- Upload the zip in Plugins > Add New > Upload Plugin and activate (WooCommerce 8.0+ required).
- Open "DEV3 AI Recommendations & Upsells for WooCommerce" in the admin menu.
- Engine > Rebuild stats. Starter widgets are already active.
- Optional: API Config > add a Google AI API key, then Engine > Build index.
Full documentation: docs/index.html in the download package.
No. Order history, cross-sells, categories and popularity power every placement. The key adds semantic similarity and AI writing tools.
Yes, both are declared compatible and tested.
Never. All AI runs in the admin or in background jobs.
This plugin can send product data (names, descriptions, categories, prices) and aggregate widget statistics to Google AI when you use DEV3 AI features or build the similarity index. No customer data is sent. Storefront tracking stores anonymous daily counters and uses two first-party cookies (dev3ru_vid for A/B assignment, dev3ru_clk for click attribution).
The full HTML guide (installation, settings, hooks and FAQ) is in docs/index.html. Open it in a browser after cloning or downloading the repository.
- Initial release.
Released under the GNU General Public License v2.0 or later (GPL-2.0-or-later). See LICENSE.