A completely free WordPress plugin that handles customer support, recommends products, and tracks orders — across your website, WhatsApp, and Telegram — while you sleep.
Your WooCommerce store is live, products are stocked, and traffic is flowing. But the questions never stop. “Where is my order?” “Do you have this in blue?” “What’s your return policy?” Every one of these messages lands in your inbox, pulling you away from running your business.
Hiring a support team is expensive. Pre-written FAQ pages get ignored. Live chat tools require someone physically online. The result? Customers wait, get frustrated, and sometimes abandon their cart entirely — not because your product was wrong, but because no one answered fast enough.
Most WooCommerce store owners settle for slow response times and high support costs because they assume AI-powered customer support is complex, expensive, or requires a developer to set up. That assumption is about to change.
Aisk is a free, open-source WordPress plugin that puts an AI-powered chatbot directly on your WooCommerce store. It reads your existing content — product pages, FAQs, blog posts, PDFs — and uses that knowledge to answer customer questions instantly, 24 hours a day, 7 days a week.
No monthly subscription. No seat limits. No vendor lock-in. You bring your own OpenAI API key, and Aisk does the rest.
Install it, generate embeddings from your site content, and within minutes your store has an intelligent assistant that knows your catalog, handles order inquiries, suggests products, and escalates to a support ticket when it genuinely can’t help.
👉 Plugin page: wordpress.org/plugins/aisk-ai-chat
Aisk’s most powerful feature is its embedding-based knowledge system. During setup, you click “Generate Embeddings” and the plugin scans your entire site — pages, posts, product descriptions, FAQs, and custom content — then sends that content to OpenAI to generate vector embeddings.
The result is a searchable knowledge base that the chatbot queries in real time. When a customer asks a question, Aisk retrieves the most relevant chunk of your content and formulates a precise, context-aware answer.
Whenever you publish new content or update products, you simply regenerate embeddings. The AI updates immediately — no manual training required.
// How Aisk triggers embedding generation via WordPress admin
add_action( 'wp_ajax_aisk_generate_embeddings', [ $this, 'generate_embeddings' ] );
Customers can check their order status directly in the chat widget. They provide their order number and email, and Aisk queries WooCommerce in real time to return status, shipping updates, and tracking information.
This single feature alone eliminates a massive portion of routine support tickets — Aisk’s own documentation cites a potential 90% reduction in order-related support inquiries.
// Aisk integrates with WooCommerce order data natively
$order = wc_get_order( $order_id );
$status = $order->get_status();
$tracking = $order->get_meta( '_tracking_number' );
When a customer types something like “I need a waterproof jacket under $80,” Aisk searches your WooCommerce product catalog and returns matching products with images and direct purchase links — all inside the chat widget.
This transforms the chatbot from a passive support tool into an active sales assistant that increases average order value through personalized recommendations.
Aisk extends beyond your website. Using Twilio for WhatsApp and the Telegram Bot API, customers can browse products, track orders, and get AI-powered answers through the messaging apps they already use daily.
Setting up WhatsApp requires a Twilio account with a WhatsApp-enabled number. Telegram setup requires creating a bot via BotFather and entering the token in the Aisk dashboard. Both integrations are configured entirely from the WordPress admin panel.
Aisk isn’t limited to your published WordPress content. Through the External Knowledge section in the dashboard, you can upload PDFs, add external URLs, and paste custom text — product manuals, return policies, size guides, supplier documentation — and the AI learns from all of it.
All conversations — website, WhatsApp, and Telegram — appear in a single unified admin dashboard. You see every message, every user, and every support ticket in one place. An analytics panel tracks conversation volume, resolution rates, and common query types.
When Aisk cannot resolve a query from its knowledge base, it doesn’t just say “I don’t know.” It asks the customer if they’d like to submit a support ticket, collects the necessary details, and creates the ticket automatically in your WordPress admin.
aisk-ai-chat/
├── includes/
│ ├── class-aisk-admin.php — Admin dashboard, settings pages
│ ├── class-aisk-chat-handler.php — Core message routing + intent classification
│ ├── class-aisk-embeddings.php — OpenAI embedding generation + retrieval
│ ├── class-aisk-woocommerce.php — WooCommerce product + order integration
│ ├── class-aisk-whatsapp.php — Twilio WhatsApp webhook handler
│ └── class-aisk-telegram.php — Telegram Bot API handler
├── assets/
│ ├── js/chat-widget.js — Frontend chat bubble + conversation UI
│ └── css/aisk-style.css — Widget styles + theme variables
├── vendor/ — Composer dependencies (OpenAI PHP client)
└── aisk-ai-chat.php — Plugin entry point, constants, bootstrap
Before Aisk answers anything, it classifies the customer’s intent. This determines whether the query needs a knowledge base lookup, a WooCommerce product search, an order status check, or a support ticket flow.
// Intent classification dispatches to the correct handler
switch ( $intent ) {
case 'product_search':
return $this->handle_product_search( $message );
case 'order_inquiry':
return $this->handle_order_inquiry( $message );
case 'support_ticket':
return $this->handle_ticket_creation( $message );
default:
return $this->handle_knowledge_query( $message );
}
This routing pattern keeps each capability modular and testable.
When a knowledge query arrives, Aisk converts the customer’s message into a vector embedding via OpenAI, then performs a cosine similarity search against the stored embeddings from your site content. The most relevant content chunks are retrieved and injected into the OpenAI prompt as context.
// Simplified: retrieve top-matching embeddings for a user query
$query_embedding = $this->openai->create_embedding( $user_message );
$top_matches = $this->db->find_similar_embeddings( $query_embedding, $limit = 5 );
$context = implode( "\n\n", array_column( $top_matches, 'content' ) );
$response = $this->openai->chat_completion([
'system' => "You are a helpful assistant. Use only this context:\n" . $context,
'user' => $user_message,
]);
This retrieval-augmented generation (RAG) approach means the AI never hallucinates answers — it only draws from content you’ve explicitly approved.
Aisk stores all timestamps in UTC using gmdate('Y-m-d H:i:s') and returns ISO 8601 strings from its internal APIs. The frontend converts these to the user’s local timezone at render time. This ensures the admin dashboard and chat widget show correct times for both store owners and customers across different timezones.
Option 1 — WordPress Plugin Directory (Recommended)
Option 2 — Manual Upload
# Download from WordPress.org and upload via FTP
wp plugin install aisk-ai-chat --activate
Settings → Aisk → Integrations → WhatsApp
Twilio Account SID: [Your Twilio SID]
Twilio Auth Token: [Your Twilio Auth Token]
WhatsApp Number: whatsapp:+1234567890
Webhook URL: https://yourdomain.com/wp-json/aisk/v1/whatsapp
Settings → Aisk → Integrations → Telegram
Bot Token: [Token from BotFather]
Webhook URL: https://yourdomain.com/wp-json/aisk/v1/telegram
The technically impressive part of Aisk is its use of Retrieval-Augmented Generation — the same architectural pattern powering enterprise AI assistants — implemented entirely inside a WordPress plugin.
Here’s why this matters: a naive AI chatbot integration simply sends every customer message directly to OpenAI with no context. The AI guesses based on generic training data and frequently gives wrong or irrelevant answers about your specific store.
Aisk solves this by pre-computing vector embeddings of all your site content and storing them in a custom WordPress database table. When a query arrives, it retrieves only the relevant content chunks and injects them as context into the OpenAI prompt. The AI is then constrained to answer from your actual business data.
// Custom DB table schema for embedding storage
global $wpdb;
$table = $wpdb->prefix . 'aisk_embeddings';
$wpdb->query("
CREATE TABLE IF NOT EXISTS {$table} (
id BIGINT(20) UNSIGNED AUTO_INCREMENT PRIMARY KEY,
source_id BIGINT(20) UNSIGNED,
source_type VARCHAR(50),
content LONGTEXT,
embedding LONGTEXT, -- JSON-encoded float array
created_at DATETIME DEFAULT CURRENT_TIMESTAMP
) {$charset_collate};
");
The similarity search compares the query embedding against all stored embeddings, selects the top N matches, and returns their content field as context. This makes the AI’s answers specific, accurate, and grounded in your business — not generic internet knowledge.
| Setting | Type | Description |
|---|---|---|
openai_api_key |
string | Your OpenAI API key for embeddings and completions |
chat_model |
string | OpenAI model to use (e.g., gpt-4o-mini) |
embedding_model |
string | OpenAI embedding model (e.g., text-embedding-3-small) |
widget_header_logo |
file | Logo displayed in chat widget header |
chat_bubble_icon |
file | Custom icon for the chat bubble trigger |
excluded_posts |
array | Post/page IDs to exclude from embedding generation |
twilio_account_sid |
string | Twilio SID for WhatsApp integration |
twilio_auth_token |
string | Twilio Auth Token |
telegram_bot_token |
string | Bot token from Telegram BotFather |
contact_information |
textarea | Custom contact info injected into AI context |
custom_context |
textarea | Additional business instructions for the AI |
Cause: PHP execution time limit is too low for large sites, or the OpenAI API key is invalid. Fix: Verify your API key in OpenAI’s dashboard, then increase PHP time limit via WordPress filter:
// In your theme's functions.php
add_filter( 'aisk_max_execution_time', function() { return 300; } );
Cause: Webhook URL is not registered with Twilio/Telegram, or SSL is not active. Fix: Confirm your site uses HTTPS, then re-save integration settings to trigger webhook registration. Check the Aisk admin log for webhook errors.
Cause: A post or page had content that could not be extracted (JavaScript-rendered or bot-protected). Fix: This was fixed in v2.1.0. Update to the latest version. For persistent issues, use the Excluded Posts setting to skip problematic pages.
Aisk’s architecture is modular, making it straightforward to extend. Developers can hook into the intent classification system to add new query types, register custom knowledge sources beyond the built-in crawlers, or build integrations with additional messaging platforms. The plugin’s REST API endpoints (/wp-json/aisk/v1/) are available for external webhook receivers — opening possibilities for integrations with CRMs, helpdesks, or custom dashboards. Given the current trajectory, native integration with Facebook Messenger and a visual chatbot flow builder seem like natural next steps for the project.
WooCommerce store owners have always faced a painful trade-off: spend money on a support team, or sacrifice customer satisfaction with slow response times. Aisk breaks that trade-off entirely.
By bringing RAG-based AI directly into WordPress — free, self-hosted, and connected to your actual store data — Aisk delivers the kind of intelligent, context-aware customer support that used to require enterprise budgets. The WhatsApp and Telegram integrations mean you’re not just adding a chat widget; you’re building omnichannel support infrastructure on top of messaging apps your customers already use.
If you run a WooCommerce store and you’re still manually answering “where is my order?” — install Aisk today. Your OpenAI API key is the only cost you’ll ever pay.
👉 Install Aisk: wordpress.org/plugins/aisk-ai-chat 📖 Documentation & Support: aisk.chat
Built with ❤️ by the Aisk.chat Team.