Your usage is measured on two separate meters. They don't share a pool, so running\
out of one doesn't affect the other. Both reset every 30 days with your billing\
cycle.

\
![](https://support.aruslabs.co/rails/active_storage/blobs/redirect/eyJfcmFpbHMiOnsibWVzc2FnZSI6IkJBaHBIUT09IiwiZXhwIjpudWxsLCJwdXIiOiJibG9iX2lkIn19--a97c8d6f35f9f2568e282ea8ca60adcda8356bdf/screenshot%206.jpg?cw_image_width=671px)

> 📷 The stats row on the app home page: Impressions, AI generations and AI replies with their progress bars.

## AI generations

These get spent when the app writes something for you.

<!--cw-colwidths:355,316-->
| Action                                                  | Cost |
| ------------------------------------------------------- | ---- |
| Index a product (knowledge base and questions together) | 1    |
| Regenerate just the knowledge base                      | 1    |
| Regenerate just the predefined questions                | 1    |
| Edit a knowledge base by hand                           | 0    |
| Restore storefront data                                 | 0    |

It's mostly a setup cost. Indexing 300 products uses 300, and the month after you\
might use five, on products you've since edited.

## AI replies

These get spent when Arus AI answers a shopper. One reply, one credit.

These are free:

* Sending a shopper off to ChatGPT, Claude, Perplexity or Grok, since the answer is\
  generated on their side.

* The copy-prompt option.

* Widget impressions. Shoppers seeing the widget costs nothing — only answering does.

## Keeping an eye on them

Both are on the app home page as live counters. Generations show as used out of your\
limit, and replies do the same, or show a running total if you're on Pro's unlimited\
allowance.

Check generations before a bulk index. Check replies if you've just made Arus AI your\
default provider on a busy store, since that's the change that actually moves the\
number.