AI Engine plugin review and common issues

Configure a detailed system prompt in AI Engine → Chatbot settings. The system prompt tells the AI what role it plays and what it knows. For accurate business information: (1) provide specific facts about your business, products, and policies directly in the system prompt; (2) enable knowledge context (Embeddings) — this allows the chatbot to retrieve information from your actual pages and posts rather than relying on AI general knowledge; (3) restrict the chatbot to only answer questions in specific domains by adding a directive like “Only answer questions about [company name] products. If asked about unrelated topics, politely redirect.” The more specific the system prompt, the more accurate and on-brand the responses.

AI Engine adds content generation tools to the Gutenberg sidebar and toolbar. If the options are not visible: (1) check AI Engine → Settings → Editor and verify “Enable AI writing” is enabled for the block editor; (2) ensure your API key is correctly configured and tested — if the API connection fails, editor tools may not initialize; (3) clear the Gutenberg editor cache (Gutenberg → Settings → Advanced → Reset editor preferences); (4) check for JavaScript conflicts — an Elementor or page builder script conflict may prevent AI Engine’s editor module from loading.

AI Engine allows configuring token limits per request in Settings → Usage. Set a maximum tokens per request to control costs. For the chatbot: (1) enable context window management to limit conversation history sent with each request — longer histories cost more tokens; (2) use a smaller, less expensive model (GPT-3.5-turbo instead of GPT-4) for routine chatbot responses; (3) monitor usage in your AI provider’s dashboard (OpenAI Platform) for a breakdown by feature and date; (4) for image generation, limit image size and frequency. Also review whether your chatbot system prompt is unnecessarily long — every request includes the system prompt in its token count.

AI Engine supports Embeddings — a process that encodes your pages and posts as vector representations the AI can search contextually. Go to AI Engine → Embeddings and add your posts, pages, and custom post types. AI Engine sends these to OpenAI (or your configured provider) as embeddings. Configure the chatbot to use embeddings as context. When a user asks a question, AI Engine searches the embeddings for relevant content and includes it in the AI prompt, allowing the chatbot to answer based on your actual site content. Rebuild embeddings whenever you significantly update your content.

With AI Engine Pro, go to WooCommerce → Products → select multiple products using checkboxes → Bulk Action → Generate Descriptions. Configure the generation prompt and style before running. AI Engine generates descriptions based on product name, existing attributes, and your prompt template. Review generated descriptions before publishing — AI-generated content benefits from human review for accuracy and brand voice. For very large catalogs, generate in batches to manage API costs and rate limits.

AI Engine itself is free for core features. You pay the AI provider (OpenAI, Google, Anthropic) directly based on API usage — AI Engine does not mark up or bundle API costs. OpenAI API costs vary by model: GPT-3.5-turbo is very affordable (fractions of a cent per request), while GPT-4 is more expensive. For a chatbot with moderate traffic (100 conversations/day), monthly API costs might be $5–15 on GPT-3.5-turbo. Monitor your OpenAI usage dashboard for actual cost tracking.

Yes — AI Engine supports local AI models compatible with the OpenAI API format (Ollama, LM Studio, and similar local inference servers). In AI Engine → Settings → Models, add a custom model with your local server’s endpoint. This is useful for organizations with data privacy requirements where sending content to external AI APIs is not acceptable. Performance depends on your local hardware — consumer GPU hardware typically runs smaller models with lower quality than cloud-based GPT-4.

Yes, that can happen, especially on older sites with many plugins. This usually happens when the plugin, theme, and add-ons are updated out of sequence. In most cases, testing on staging catches the issue before it reaches the live site. From experience, backups and changelog reviews save a lot of cleanup time.

Start by checking whether another plugin already does the same job. In most cases, overlap is what creates avoidable conflicts and performance issues. A common issue is installing a plugin because it looks convenient without checking the stack first. From experience, a short compatibility review avoids most of the pain later.