Vertex AI Search and Conversation has been renamed to Vertex AI Agent Builder. We are in the process of updating content to reflect the new branding.

Use structured data for advanced website indexing

If advanced website indexing is enabled in your data store, you can use the following types of structured data to enrich your indexing:

Predefined, Google-inferred page dates
Custom structured data attributes

This page introduces both these types of structured data for your web pages and describes how to add custom structured attributes to your data store schema.

About predefined, Google-inferred page dates

When crawling through the web pages in your website data store, Google infers page data using the properties that apply to your content. Vertex AI Search adds these inferred page data properties to your schema. This inferred data includes the following predefined date properties:

datePublished: the date and time when the page was first published
dateModified: the date and time when the page was most recently modified

These properties are indexed automatically. You can directly use these date properties to enrich your search without adding them to your schema. You can include these predefined date properties in your search requests, such as in filter expressions and boost specifications. For more information, see Example use case using a Google-inferred page date.

About custom structured data attributes

You can add structured data attributes as meta tags and PageMaps to your web pages and use these to enrich your indexing. To use custom structured attributes for indexing, you must update your schema.

Example use case for `meta` tags

Suppose you have a large number of web pages that are relevant to various departments in your organization. You can use meta tags to label the pages that are relevant for each department. You can then use the indexed tags as filters in your queries. This lets you to restrict search results to web pages containing a label that matches any of the specified departments.

This process can be summarized as follows:

Add the following meta tags to a subset of your web pages:
- Relevant to engineering and IT departments: <meta name="department" content="eng, infotech">
- Relevant to finance and HR departments: <meta name="department" content="finance, human resources">
Recrawl the updated pages.
Add department to your data store schema as an indexable array as described in the Add custom structured data attributes to the data store schema section.

After updating your schema, your data store is automatically reindexed. After the reindexing is complete, you can use the department filter in a filter expression to reorder or filter search results. For example, when users from the finance department issue queries, the search results can be made more relevant for them with the department filter set to finance.

Example use case for PageMaps

Suppose you have several web pages that contain food recipes. You can add PageMap data to each page's HTML content. You can then use the indexed PageMap attribute names as filters in your queries. For example, if you intend to boost or bury web pages depending on the recipe ratings, you can follow this process:

Add PageMap data similar to the following to your web pages:

<html>
<head>
...
<!--
<PageMap>
    <DataObject type="document">
        <Attribute name="title">Baked potatoes</Attribute>
        <Attribute name="author">Dana A.</Attribute>
        <Attribute name="description">Homestyle baked potatoes in oven. This
        recipe uses Russet potatoes.</Attribute>
        <Attribute name="rating">4.9</Attribute>
        <Attribute name="last_update">2015-01-01</Attribute>
    </DataObject>
</PageMap>
-->
</head>
...
</html>

Recrawl the updated pages.
Add rating to your data store schema as an indexable array as described in the Add custom structured data attributes to the data store schema section.

After updating your schema, your data store is automatically reindexed. After the reindexing is complete, you can use the rating attribute in a filter expression to reorder or filter search results. For example, when users search for recipes, boost the search results that are top-rated by using rating as a custom numerical attribute.

Example use case for schema.org data

Suppose you have a review website and its web pages are annotated with schema.org data in JSON-LD format within the HTML script tag. You can then use the indexed annotations as filters in your queries. For example, if you intend to boost or bury web pages depending on the aggregate ratings, you can follow this process:

Add the schema.org annotations for review content similar to the following to your web pages. To view other types of schema.org templates that are available, see Schemas:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Review",
  "aggregateRating": {
    "@type": "Average Rating",
    "ratingValue": 3.5,
    "reviewCount": 11
  },
  "description": "Published in 1843, this is the perfect depiction of the Victorian London. A Christmas Carol is the story of Ebenezer Scrooge's transformation.",
  "name": "A Christmas Carol",
  "image": "christmas-carol-first-ed.jpg",
  "review": [
    {
      "@type": "Review",
      "author": "Alex T.",
      "datePublished": "2000-01-01",
      "reviewBody": "Read this in middle school and have loved this ever since.",
      "name": "Worth all the adaptations",
      "reviewRating": {
        "@type": "Rating",
        "bestRating": 5,
        "ratingValue": 5,
        "worstRating": 1
      }
    }
  ]
}
</script>

Recrawl the updated pages.
Add the path to ratingValue to your data store schema. Use an identifier as the field name in the data store schema, such as rating_value as described in the Add custom structured data attributes to the data store schema section.

After updating your schema, your data store is automatically reindexed. After the reindexing is complete, you can use the rating_value attribute in a filter expression to reorder or filter search results. For example, when users search for books, boost the search results that are top-rated by using rating_value as a custom numerical attribute.

Before you begin

Before you update the data store schema, do the following:

Turn on advanced website indexing for the data store. For more information, see Turn on advanced website indexing.
Understand how structured data works.
Understand how to use PageMaps. Review the list of recognized DataObjects that can be added to PageMap data.
Understand how to use meta tags. Ensure that you don't use any excluded or unsupported meta tags.
Ensure that the attribute that needs to be indexed doesn't have any of the following values:
- datePublished
- dateModified
- siteSearch
Understand that after you add structured data to your web pages, you must recrawl the pages. This might take several hours.
Understand that after you add structured data attributes to the data store schema, the web pages in your data store are reindexed automatically. Reindexing is a long-running operation that might take several hours.

Add custom structured data attributes to the data store schema

To add custom structured data attributes to the data store schema:

Add meta tags, PageMap data, and schema.org data to all the pages in your website that you want to enrich with structured data indexing:
- For meta tags:
  - Each meta tag must have its name attribute set to the field you want to index and its content attribute to a string of one or more comma-separated values.
  - Vertex AI Search supports meta tags with names that match the pattern [a-zA-Z0-9][a-zA-Z0-9-_]*. Ensure that you don't use any excluded or unsupported meta tags.
- For PageMap data:
  - PageMap data must consist of recognized DataObjects that contain Attribute names that you want to index. The Attribute names within the DataObjects must be set to the field you want to index.
- For schema.org data:
  - The annotations must be in valid JSON-LD, Microdata, or RDFa format. For more information, see Supported formats.
Recrawl the updated web pages.
View the schema definition for your data store over REST API.
Update the data store schema over REST API. For more information, see About providing your own schema as a JSON object.
1. Add objects for each custom attribute that you want to make searchable, retrievable, or indexable.
2. Add the custom attribute and set its type to array.
3. Add the data type of the custom attribute's value.
4. Specify the source of the custom attribute where it can be found in the siteSearchStructuredDataSources field.
5. For schema.org data: Specify the path of the attribute in the schema.org annotation starting with string _root in the siteSearchSchemaOrgPaths field.
The following is an example of a schema update for a website:
```
{
  "type": "object",
  "properties": [{
    "CUSTOM_ATTRIBUTE": {
      "type": "array",
      "items": {
        "type": "DATA_TYPE",
        "searchable": true,
        "retrievable": true,
        "indexable": true,
        "siteSearchStructuredDataSources": ["STRUCTURED_DATA_SOURCE_1", "STRUCTURED_DATA_SOURCE_2"]
      }
    }
  }
  {
    "IDENTIFIER_FOR_SCHEMA_ORG_FIELD": {
      "type": "array",
      "items": {
        "type": "DATA_TYPE_SCHEMA_ORG_FIELD",
        "searchable": true,
        "retrievable": true,
        "indexable": true,
        "siteSearchSchemaOrgPaths": ["_root.PATH_TO_THE_SCHEMA_ORG_FIELD"]
      }
    }
  }],
  "$schema": "https://json-schema.org/draft/2020-12/schema"
}
```
Replace the following:
- CUSTOM_ATTRIBUTE: the value of the name attribute. For example:
  - For a meta tag defined as <meta name="department" content="eng, infotech">, use department
  - For a PageMap Attribute defined as <Attribute name="rating">4.9</Attribute>, use rating
- DATA_TYPE: the data type of the name attribute. Must be either string, number, or datetime. For example:
  - For a meta tag defined as <meta name="department" content="eng, infotech">, use string
  - For a PageMap Attribute defined as <Attribute name="rating">4.9</Attribute>, use number
  - For a PageMap Attribute defined as <Attribute name="last_published">2015-01-01</Attribute>, use datetime
  For more information, see FieldType.
- STRUCTURED_DATA_SOURCE_N: an array consisting of one or both of the following structured data sources where the CUSTOM_ATTRIBUTE attribute can be found:
  - If the custom attribute can be found as a meta tag, specify METATAG
  - If the custom attribute can be found as a PageMap attribute, specify PAGEMAP
  - If the custom attribute can be found as a schema.org data, specify SCHEMA_ORG
  - If the siteSearchStructuredDataSources field is absent or left empty, the values from all three data sources are merged in an array.
- IDENTIFIER_FOR_SCHEMA_ORG_FIELD: a custom identifier to denote the schema.org field. It doesn't need to be same as the field name in the schema.org annotation on your web page. For example, if the path of the field is _root.nutrition.calories, the identifier can be calorific_value or nutrition_value
- DATA_TYPE_SCHEMA_ORG_FIELD: the data type of the schema.org field. Must be string, number, or datetime. For example:
  - For a schema.org field defined as "calories":"240 calories", use string
  - For a schema.org field defined as "calories": 240, use number
  - For a schema.org field defined as "foundingDate": "1991-05-01", use datetime
  For more information, see FieldType.
- PATH_TO_THE_SCHEMA_ORG_FIELD: the path to a single field in the schema.org field that needs to be accessed. It is specified using dot separators after each nested level. You must specify the complete path needed to access the required field. For example, if a field ratingValue is nested in the aggregateRating field, you can specify the path as _root.aggregateRating.ratingValue.

After you update the website schema, the website is reindexed automatically. This is a long-running operation that can take several hours.

What's next

Use the indexed metadata for the following:

Serving controls, such as boost, bury, and filter
Surfacing as facets in search results
Filter search results
Boost search results