Attribute Guide

structured_title Attribute in Google Shopping: AI-Generated Title Guide

Q: What values are allowed for digital_source_type?

Google currently accepts two values for the digital_source_type sub-attribute: trained_algorithmic_media — the title was generated using a trained AI/ML model (e.g., a language model, a rule-based title generator using structured product data); and default — the title was assembled using standard template or rule-based logic without a trained model. If you are using a large language model (LLM) to generate titles, use trained_algorithmic_media.

Q: When should I use structured_title instead of title?

Use structured_title when your product titles are generated or substantially enhanced by an AI or algorithmic system rather than manually written. Common scenarios: using an LLM (GPT, Gemini, Claude, etc.) to write titles from product attributes; using a feed tool or PIM system that generates titles programmatically from product data fields; using template-based title construction with rule logic across large catalogues. Manual titles — even if based on a formula — are typically submitted as standard title.

Q: Does Google penalise or treat AI-generated titles differently?

No — structured_title is designed precisely to allow AI-generated titles in a transparent, policy-compliant way. Google introduced this attribute as part of its AI-disclosure framework for product data. The same quality standards apply: titles must be accurate, relevant, not promotional, and follow the same character limits and formatting rules as standard title. The disclosure via digital_source_type satisfies Google's transparency requirement for algorithmically generated content.

Q: If I submit both title and structured_title, which one is used?

When both attributes are present for the same product, structured_title takes precedence. The title attribute is effectively overridden. This means you can safely transition from title to structured_title incrementally — populate both during migration, and structured_title will be used where it is present while title serves as a fallback for products not yet migrated.

Complete guide to the structured_title attribute in Google Merchant Center. When to use it over the standard title attribute, the digital_source_type…

Updated 31 July 20264 min readBoostora Team

structured_title — Standard Title vs. AI-Enhanced Structured Title

IKEA MARKUS Ergonomic Office Chair — Vissle Dark Grey, Adjustable Lumbar Support, Swivel Base

€229.00furniture-store.com

AI-generateddigital_source_type: trained_algorithmic_media

Richer title with brand, product type, material, colour, key features. Better keyword coverage. Generated at scale by AI.

MARKUS Office Chair

€229.00furniture-store.com

Title = brand name + product name only. Missing material, colour, key features. Short, low keyword density. Submitted manually, hard to maintain at scale across thousands of SKUs.

structured_title is a 2024 attribute that allows merchants to submit AI-generated or algorithmically assembled product titles with explicit disclosure of the generation method via digital_source_type. It uses the same 150-character limit and same display placement as title.

What is the structured_title attribute?

Introduced in 2024, structured_title is a variant of the standard title attribute that includes a mandatory digital_source_type sub-attribute. This sub-attribute discloses how the title was generated — specifically, whether it was produced by a trained AI model or a rule-based algorithm. The attribute was introduced as part of Google’s broader framework for transparent AI-generated content in product feeds.

Functionally, structured_title and title display identically in Shopping ads and listings. The distinction is purely in how the title was produced and how that production method is disclosed. When both are submitted, structured_title takes precedence.

Technical specifications

Feed name	`structured_title`
Required	Optional — use instead of (or alongside) `title` for algorithmically generated titles
Max length	150 characters (same as `title`)
Sub-attribute	`digital_source_type` — required when submitting `structured_title`
digital_source_type values	`trained_algorithmic_media` (LLM/ML-generated) or `default` (template/rule-based)
Priority vs. title	`structured_title` takes precedence when both are submitted
Display	Identical to `title` — same Shopping ad placement, same character display limit (~70 visible)

Format: how to submit structured_title

In XML feeds, structured_title uses nested sub-attributes:


  trained_algorithmic_media
  IKEA MARKUS Ergonomic Office Chair — Vissle Dark Grey, Adjustable Lumbar Support

In tab-delimited feeds, use the format: digital_source_type:trained_algorithmic_media [structured_title] with the title text as the value, or follow the column-header convention in your feed template.

Quality rules (same as standard title)

Maximum 150 characters; first ~70 are visible in most Shopping placements — put critical keywords first
Must accurately describe the product — AI-generated does not mean promotional language is acceptable
No ALL CAPS (other than brand names/acronyms that are officially capitalised)
No promotional phrases: “Best price”, “Free shipping”, “Sale”, ”% off”
No HTML tags or special encoding
Include: brand, product type, key attributes (color, material, size where relevant)
Recommended structure: Brand + Product Type + Key Attribute(s) + Model/Variant

When to use structured_title vs. title

Use title	Manually written titles; titles based on simple string concatenation of existing fields with no AI involvement
Use structured_title	Titles generated by an LLM; titles substantially rewritten or enhanced by an AI system; titles produced by a trained product-data ML model
Use both (migration)	During transition: `structured_title` for AI-enhanced SKUs, `title` as fallback for unprocessed SKUs

Best practices

Apply the same title optimisation best practices — being AI-generated does not exempt a title from the brand-first, attribute-rich structure that maximises Shopping impression share. Brand + Product Type + Key Attributes remains the optimal formula.
Review AI-generated titles before deployment — validate a sample of generated titles for accuracy, hallucinations (invented specifications), and policy compliance before submitting to the feed at scale. AI models can generate plausible-sounding but factually incorrect product attributes.
Use structured_title as your primary for AI-generated at scale — for large catalogues where AI title generation is systematised, structured_title is the correct vehicle and ensures policy compliance.

Common mistakes

Submitting AI-generated titles via the standard title attribute without using structured_title — technically works but does not satisfy Google’s AI content disclosure framework
Omitting digital_source_type when submitting structured_title — sub-attribute is mandatory
Using structured_title for manually written titles — not wrong, but adds unnecessary complexity; title is simpler for human-authored content
Allowing AI-generated titles to include promotional language or inaccurate specs — same quality rules apply as to standard titles
Not reviewing AI-generated output for factual accuracy — LLMs can hallucinate product specifications

Frequently asked questions

What is the difference between title and structured_title?

Both attributes serve the same function — they provide the product title shown in Shopping ads and listings, with the same 150-character limit and the same display placement. The key difference is that structured_title includes a mandatory digital_source_type sub-attribute that discloses how the title was generated. This disclosure requirement makes structured_title the correct attribute to use when your title was generated or significantly enhanced by an AI system rather than written by a human. If you submit both attributes, structured_title takes precedence over title.

What values are allowed for digital_source_type?

Google currently accepts two values for the digital_source_type sub-attribute: trained_algorithmic_media — the title was generated using a trained AI/ML model (e.g., a language model, a rule-based title generator using structured product data); and default — the title was assembled using standard template or rule-based logic without a trained model. If you are using a large language model (LLM) to generate titles, use trained_algorithmic_media.

When should I use structured_title instead of title?

Use structured_title when your product titles are generated or substantially enhanced by an AI or algorithmic system rather than manually written. Common scenarios: using an LLM (GPT, Gemini, Claude, etc.) to write titles from product attributes; using a feed tool or PIM system that generates titles programmatically from product data fields; using template-based title construction with rule logic across large catalogues. Manual titles — even if based on a formula — are typically submitted as standard title.

Does Google penalise or treat AI-generated titles differently?

No — structured_title is designed precisely to allow AI-generated titles in a transparent, policy-compliant way. Google introduced this attribute as part of its AI-disclosure framework for product data. The same quality standards apply: titles must be accurate, relevant, not promotional, and follow the same character limits and formatting rules as standard title. The disclosure via digital_source_type satisfies Google's transparency requirement for algorithmically generated content.

If I submit both title and structured_title, which one is used?

When both attributes are present for the same product, structured_title takes precedence. The title attribute is effectively overridden. This means you can safely transition from title to structured_title incrementally — populate both during migration, and structured_title will be used where it is present while title serves as a fallback for products not yet migrated.

Can Boostora generate structured_title content?

Boostora reads structured product attributes from your feed and can assemble or enrich titles using product data signals. When Boostora uses AI-assisted enrichment to generate or significantly enhance a title, it outputs this via structured_title with the appropriate digital_source_type disclosure. Boostora does not verify product specifications it cannot observe from the feed data — title content accuracy for attributes like material, dimensions, or features depends on the quality of the source product data.