You can guide Selleasy AI on what products to recommend using the "Tell the AI what to recommend" box in your campaign. It has two modes:
- Simple - Describe what you want in your own words.
- Advanced - Take full control over what the AI reads, how it searches, how it ranks the results and which model it uses.
Both modes are optional. If you leave them blank, the AI decides based on your store's product data.
1. Where to find it
- Open your campaign in the Selleasy app.
- In the Offer products section, select the Selleasy AI tab.
- Scroll down to the "Tell the AI what to recommend" box.
- Choose Simple or Advanced using the toggle on the right.
2. Using Simple mode
In Simple mode, type what you want the AI to recommend in plain words. The AI handles the rest.
Example
Recommend only suitable accessories based on product type. Do not recommend gift products.
More examples:
- "Recommend matching bottoms and footwear for this top."
- "Suggest refills and small add-ons under $20."
- "Only recommend products from the same brand."
Tips for better results
- Keep it short and specific. Say what to recommend and what to avoid.
- Use words that appear in your product titles, types, collections or tags. This helps the AI find the right products.
- Leave the box blank if you want the AI to decide on its own.
3. Using Advanced mode
Advanced mode gives you control over each step of how the AI finds recommendations. The AI reads the product fields you allow, searches within them, then ranks what it finds.
3.1 Using your Simple mode description
If you've already written a description in Simple mode, you'll see the message "Do you want to use the configuration inferred from simple mode?" when you switch to Advanced.
- Click "Use config" to have Selleasy convert your description into Advanced mode settings. You can then fine-tune them.
- Click × to ignore it and start from scratch.
3.2 Fields the AI can read
Choose which product details are sent to the AI for every product. The AI uses only these fields to understand and compare products.
Available fields: Title, Description, Product type, Category, Price range, Collections, Tags and Variants.
- Selected fields show a ✓ tick mark.
- Click a field to leave it out, or click it again to include it.
Example
If your product descriptions are long and full of marketing text, leave out Description so the AI focuses on Title, Product type and Tags.
3.3 Custom fields
Add a metafield or variant option that the AI should also read. This is useful when important product details, like brand or compatibility, are stored in a metafield.
- Click "+ metafield" or "+ variant option".
- Enter the field in the format shown:
- Metafield -
metafields.namespace.key, for examplemetafields.custom.brand - Variant option -
variants.options.option_name, for examplevariants.options.size
- Click "Add". The field appears as a tag above the box. Click × on the tag to remove it.
3.4 Search instructions (optional)
Tell the AI what kind of products to look for. You can use up to 1000 characters.
Use {{targetProduct}} to refer to the trigger product.
Example
Find compatible items, refills, parts, or extras that shoppers may want to buy with {{targetProduct}}.
3.5 Ranking instructions (optional)
Tell the AI how to sort and filter what it finds. You can use up to 1000 characters.
Example
Rank products by compatibility with {{targetProduct}}, usefulness, relevance, availability, and value. Remove any unrelated or duplicate items.
3.6 Filter results after the AI search (optional)
Enable "Filter results after the AI search" to control which products are finally shown, after the AI has finished its search.
- Recommendations can be filtered by Collection, Tag, Product type and Vendor.
- You can add more than one value in a condition. For Collection and Tag, choose how the values should match:
- "in any of the collections/tags" - products matching at least one value are shown.
- "in all collections/tags" - products matching all values are shown.
- Click "Add condition" to add more conditions, then choose:
- "All conditions must match" - a product must satisfy every condition.
- "Any condition must match" - a product must satisfy at least one condition.
Note
Filters are applied after the AI search. Strict filters may reduce the number of recommended products.
3.7 Deciding Recommendation Mode and Search Method
- Recommendation mode
- Standard (default) - Uses 1 credit per recommendation. Recommendations are generated quickly and are satisfactory for most use cases.
- Thinking - Uses 2 credits per recommendation. This model takes more time to generate recommendations, as it analyzes the products more thoroughly.
- Search method
- GraphQL (default) - The standard search method.
- Storefront MCP - If the GraphQL product search is not working for some reason, you can switch to the Storefront MCP option and try again.
FAQs
- Which mode should I use? Start with Simple mode. It works well for most stores. Switch to Advanced mode only if you want more precise control, for example to make the AI read a specific metafield or to rank products in a certain way.
- Do I have to fill in Simple or Advanced mode? No. Both are optional. If you leave them blank, the AI decides based on your product data.
- If I switch from Simple to Advanced, is my description lost? No. You can click "Use config" to convert it into Advanced mode settings.
- Why am I getting fewer products than the number I set? The AI shows only products it finds relevant. Strict instructions or filters can also reduce the count. Try loosening them.
If you need any assistance while setting up the above, please feel free to contact our customer support team.