Mira AI is already answering visitor questions from your website and product feed — but sometimes you'll spot answers that need a human touch. The Training feature lets you add custom Q&A pairs, fix wrong answers directly from conversations, and decide which chatbot uses which custom responses.

We've broken the guide into the following sections:

Training fills the gaps that web scraping and the product feed can't cover — internal policies, special offers, brand-specific phrasing, or anything that isn't on your website yet. Watch the quick walkthrough, then jump into the steps below.

Add custom Q&A pairs in four short steps.

Go to the AI tab in your dashboard and select Training from the left menu.

Click the Add New button in the top-right corner (or from the list) to add a question that Mira AI couldn't previously answer based on web scraping.

In the modal, type your question and the answer Mira AI should give. Don't forget to click Save.

The Training page shows your usage against the 30-question limit. Keep your most-asked, hard-to-answer questions in the top 30 — and remove old ones as your needs change.

When Mira AI replies to a visitor and the answer misses the mark, you can fix it on the spot — without leaving the inbox.

In any conversation, click the Correct answer button next to Mira's reply. A modal opens where you add the question and the better answer. The new Q&A is saved straight to the Training page and Mira will use it from now on.

If you run more than one AI chatbot, you can decide which one uses your custom responses and which one doesn't. One chatbot can rely on the trained Q&A pairs while another sticks to web scraping and product feed only.

Open the chatbot you want to configure, go to Knowledge, and toggle the Custom responses switch.

Once custom responses are enabled, you can also pick specific Q&A pairs the chatbot will use. Select the ones you want and save.