Skip to main content

Trustana AI Rules Configuration

A step-by-step guide to configuring Trustana AI Rules (via CSV) to control enrichment scope, data sources, search terms, and brand vocabulary for more accurate and consistent product attribute enrichment.

Updated this week

This guide explains how to use AI Rules in Trustana to control where enrichment data comes from, in what order it is used, and how Trustana searches for sources, using a simple CSV-based workflow.


1. Why AI Rules exist

  1. Default enrichment works without any configuration

    • You can leave an account without any AI Rule configuration and Trustana will still enrich attributes.

    • In the default setup, Trustana searches extensively and follows this source priority:

      • First Party Text → Images → Preferred URL → Third Party.

  2. AI Rules are for “specific requirements”

    • Use AI Rules when you need strict control, for example:

      • Only use official brand sites

      • Only use third-party sources

      • Use only a provided list of URLs (no broad search)

      • Use only images

      • Enforce brand wording (include/exclude words)


2. The mental model

AI Rules are a set of decisions you make for enrichment:

  1. Scope (who this rule applies to)

  2. Trusted websites (where Trustana is allowed to look)

  3. Source prioritization (which types of sources to use, and in what order)

  4. Search terms (how Trustana builds search queries)

  5. Vocabulary controls (how results should be phrased)

You can fill only what you need. The goal is to make enrichment more predictable and more aligned with a customer’s expectations.


3. Rule scope and how Trustana applies rules

Trustana supports multiple rule scopes. The more specific the scope, the more targeted the rule.

  1. Account level

    • A rule row with no Brand and no Category.

    • Applies to all products by default if no more specific rule exists.

  2. Category level

    • A rule row with Category only.

    • Applies to all products in that category.

  3. Brand level

    • A rule row with Brand only.

    • Applies to all products under that brand.

  4. Brand + Category level

    • A rule row with both Brand and Category.

    • Applies to products matching that Brand within that Category.

Important note: rules are “by field”

  • Different rule fields can come from different scopes.

  • Example: if you set Preferred URLs at Brand level, but Source prioritization at Account level, Trustana can consider both because they affect different fields.


4. End-to-end story: a realistic customer scenario

Below is a typical situation where AI Rules provide clear value.

Scenario

A retailer has 50,000 SKUs across several categories. They notice:

  1. Some brands have inconsistent attribute values because third-party pages conflict.

  2. Some niche brands have poor coverage because the official site is not being found reliably.

  3. Their marketing team wants consistent brand phrasing (avoid words like “cheap” and keep certain brand expressions).

What you would do with AI Rules

  1. Account-level baseline

    • Set a sensible default source strategy (for example: Official brand first, then third party).

  2. Brand-level precision for niche brands

    • Add official domains in Brand primary URLs so Trustana finds the correct site.

  3. Brand vocabulary governance

    • Add include/exclude vocabulary so generated text stays aligned with brand tone.

  4. Category constraints (only where needed)

    • If a category is frequently wrong (example: supplements or electronics), add category-level search terms or source priority that works better for that category.

This creates a controlled system:

  • Broad defaults for scale

  • Targeted overrides where accuracy or governance requires it


5. Step-by-step: How to configure AI Rules (CSV workflow)

Step 1 — Open AI Rule Configuration

  1. Log in to the Trustana PIM Portal at portal.trustana.com

  2. Navigate to "Enrichment Rules".

Step 2 — Download the CSV template

  1. You can start filling in the CSV by using 2 ways 1) Downlaoding template from "Import Rules" section (when you don't have any rules already imported) and 2) Exporting rules (when you already have rules)

    When you are starting new (download the template):

  1. When already have existing rules :

Step 3 — Fill the CSV (Refer to Section 6 in the article to understand how to fill each column)

  1. Add one row per rule you want (account / brand / category / brand+category).

  2. In any cell where multiple values are allowed, use a separator like | as shown in the examples.

Step 4 — Upload the CSV

  1. Upload the completed CSV.

Step 6 — Verify results and manage rules

  1. After upload, rules appear in the Current Rules tab.

  2. You can delete rules from this tab.

Optional: Clearing values (setting them to null)

  1. If you need to explicitly clear an existing rule field, use the toggle:

    • “Treat empty cells as null” during upload.



6. CSV fields explained (what each column means)

6.1 Targeting fields (Scope)

  1. Brand name

    • Any brand name in your PIM (example: Adidas, Nike).

  2. Product category

    • A leaf node in your custom category tree (example: Electronics → Headphones).

  3. Google category

    • A leaf node in the Google category tree if your setup uses it.

6.2 Trusted website fields

  1. Brand primary URLs

    • Domains or URLs you want Trustana to use as trusted sources.

    • You can enter many URLs.

    • You do not need to enter this for every brand because Trustana can identify brand sites automatically, but it is useful for less popular brands or when the brand name and domain do not match clearly.

  2. Brand secondary URLs

    • Same behavior as primary URLs (provided separately for clarity).

6.3 Brand vocabulary fields

  1. Brand vocabulary include

    • Words or phrases that should be included when relevant to the attribute and context.

  2. Brand vocabulary exclude

    • Words or phrases that should be excluded when relevant to the attribute and context.

6.4 Source prioritization

  1. Source prioritization

    • The order in which enrichment should happen across source types.

    • Supported values:

      • IMAGE, FIRST_PARTY, THIRD_PARTY, OFFICIAL_BRAND, PREFERRED URL

    • You can choose one or more and order them to match your requirement (example: only images, or only brand site, or brand site first then third party).

6.5 Search terms

  1. Search terms

    • Field keys used to build search queries (example shown: product model, brand, name).

    • These attributes are used for search in Google and Gemini to find the right sources.


7. Worked examples (copy-paste ready)

The exact column order in your template should be followed. Below is a practical example layout using the fields described in the guide. Adjust column names to match your downloaded CSV template.

How to read these examples

  1. Row 1: Account-level default

    • Applies to everything in the account (no brand, no category).

    • Uses the default-like order (First Party, Images, Preferred URL, Third Party).

  2. Row 2: Brand-level rule

    • Forces Official Brand first, then Third Party.

  3. Row 3: Category-level rule

    • Applies to all products in that category.

  4. Row 4: Brand vocabulary governance

    • Adds include/exclude phrasing controls.

  5. Row 5: Brand + Category precision

    • Adds trusted brand domains and a specific source order only for that brand in that category.


8. Common “recipes” (what to configure for typical requirements)

These map directly to the supported use cases in the guide.

  1. Only use official brand sources

    • Source prioritization = OFFICIAL_BRAND

  2. Only use third-party sources

    • Source prioritization = THIRD_PARTY

  3. you already have the URLs, so no broad search is needed

    • Fill Brand primary urls (and secondary if needed)

    • Set Source prioritization = PREFERRED URL (optionally add fallback sources after it)

  4. Image-only enrichment

    • Source prioritization = IMAGE

  5. You need very specific vocabulary

    • Fill Brand vocabulary include and Brand vocabulary exclude at the correct scope.


9. Best-practice guidance (recommended approach)

These are practical recommendations to keep configurations stable and maintainable.

  1. Start broad, then narrow

    • Use an account-level baseline only when it is truly safe for all products.

    • Use brand/category overrides only where the baseline fails.

  2. Avoid overly strict rules unless necessary

    • If you set only one source type, enrichment can become fragile when that source is missing key details.

  3. Use Brand primary URLs for control and accuracy

    • Add them especially when:

      • The brand is niche or hard to find

      • The brand’s official site is not the obvious domain

      • The brand has multiple domains per region

  4. Treat vocabulary as governance

    • Use include/exclude lists to improve consistency and tone.

    • Do not use vocabulary rules to “force” facts that are not supported by sources.


10. Verification and operational checklist

  1. After every CSV upload

    • Confirm the rules appear in Current Rules.

    • Confirm the scope looks correct (brand/category filled as intended).

  2. When you want to roll back

    • Delete the rule from Current Rules.

  3. When you want to clear a field

    • Upload a CSV with that field empty and enable Treat empty cells as null.


Did this answer your question?