AI Product Enrichment & SEO Publishing

Product imagery and catalog data contained descriptive information that could improve product discovery, but preparing consistent metadata across a large assortment required substantial manual effort. The opportunity was to use AI-assisted enrichment to produce structured information that downstream systems could consume.

The SEO application built on an existing Smart Tagging capability associated with internal search. Its purpose was to make selected enrichment useful in product-image alt text while preserving the product identity and attributes already maintained in the commerce platform.

SEO stakeholders considered opportunities across product titles, descriptions, headings, and image alt text. The initial proof of concept focused on image alt text for selected seasonal product categories, starting with English-language markets.

This required more than generating descriptions. The team needed to establish which products qualified, what seasonal enrichment they should receive, where that information would reside, and how SFCC would combine it with existing product attributes.

GPT-4 supported upstream interpretation and enrichment. SFCC controlled the composition of the published image alt text.

The original pattern contained:

Product name → Brand → Image index

The proposed pattern added color and an optional SEO enrichment value:

Color → Product name → Brand → SEO enrichment → Image index

The original requirements illustrated this with:

“White Belted Linen Shirtdress Polo Ralph Lauren for summer 1.”

Separating enrichment from formatting allowed the storefront to retain a predictable structure while incorporating additional context where available.

Color needed to appear whenever available, even when a product had no Smart Tag enrichment. The SEO value therefore remained optional rather than becoming a prerequisite for the base alt text.

The requirements addressed:

Dataiku provided the workflow environment for processing product information and invoking GPT-4. GPT-4 interpreted product imagery alongside catalog context to support descriptive enrichment and structured metadata preparation.

For the SEO application, selected products received relevant seasonal enrichment linked to their product records.

SFCC provided the product attributes and PDP rendering logic needed to assemble the final image alt text.

Implementation documentation refers to seoSmartTags and seoImgAlt in different requirements and ticket contexts. The essential publishing behavior was consistent: incorporate the optional SEO value alongside existing product information, with independent handling of color.

A current live product-page example shows the resulting structure:

“Moss Agate The Iconic Cotton Chino Ball Cap Polo Ralph Lauren for fall 1.”

The SFCC-to-Bloomreach catalog mapping included Smart Tags for internal-search consumption. That was a related use of product enrichment.

The SEO publishing path operated through SFCC product attributes and PDP rendering; Bloomreach was not a required intermediate step for publishing image alt text.

I led product prioritization and cross-functional requirements for the Global SEO Roadmap, working across regional SEO teams, data science, analytics, technology, and Publicis Sapient.

For this initiative, I coordinated the requirements connecting product selection, enrichment, SFCC field behavior, storefront rendering, and acceptance criteria. Within the broader roadmap, I aligned dependencies and delivery priorities across regions and workstreams.

The FY25 plan estimated $310K for the Sapient engagement. The delivery status report identified 10 Sapient contributors across product/project management and development/QA, supporting the wider SEO program.

Implementation and acceptance requirements covered:

The SEO measurement plan called for confirmation of updated alt tags followed by an eight-week pre/post analysis using Google Search Console and SEMrush. Planned measures included organic PDP traffic, keyword visibility, and image performance, with a separate UK analysis if that market participated.

Later roadmap documentation also identified a localization issue: color had appeared in English within non-English alt text. This highlighted the need to validate language behavior as well as the presence of the new attributes.

The enrichment workflow covered 10K+ catalog SKUs and reduced manual metadata-preparation effort by 80%. The efficiency measure applies to preparing metadata, rather than the complete development, testing, and publishing cycle.

The SEO extension connected that enrichment to a repeatable publishing process, with explicit rules for optional values, variants, product categories, and localization.

This initiative was part of the Global SEO Roadmap I led. North American SEO reporting recorded 10.9% year-over-year growth in visits, increasing from 19.06 million to 21.14 million during the April–January reporting period.

This was a combined US-and-Canada channel result across the broader program. The report did not separately quantify the traffic contribution of Smart Tags.

The central product decision was to separate AI-generated enrichment from the rules governing its publication. That made the enrichment reusable across discovery channels while giving each destination explicit requirements for product identity, optional fields, localization, and rendering.

Delivering that connection required alignment across data science, regional business priorities, commerce-platform implementation, and operational validation.