ProductKnowledgeGraph vs. Salsify: Two Approaches to Product Data for AI Commerce

Salsify and ProductKnowledgeGraph both address a rapidly emerging enterprise problem: AI systems are becoming part of how products are discovered, compared, recommended, and purchased. But the two platforms approach that problem from different architectural starting points.

Salsify extends Product Experience Management (PXM) into the agentic commerce era.

ProductKnowledgeGraph starts with a different question:

That distinction matters as enterprise product data moves beyond websites, retailer feeds, and traditional search into AI assistants and autonomous commerce agents.

The shift from the digital shelf to the agentic shelf

For years, enterprise product infrastructure was primarily designed around a familiar flow:

AI commerce introduces another decision layer:

An AI system may need to answer questions such as:

• Is this the correct product?
• Which variant is compatible?
• Is this specification current?
• Does the manufacturer make this claim?
• Does a retailer describe the product differently?
• Is there regulatory or certification evidence?
• Which purchase destination represents the current offer?

This changes the product-data problem.

Structured product content remains essential, but identity, provenance, authority, conflicts, and evidence become increasingly important when software is interpreting product information on behalf of a buyer.

What Salsify is building

Salsify is an established enterprise Product Experience Management platform for manufacturers, retailers, and distributors.

Its platform covers product information management, syndication, enhanced content, supplier onboarding, GDSN, digital shelf analytics, automation, and integrations across the commerce ecosystem. Salsify describes PIM as a central system of record and syndication as a way to distribute product content across consumer touchpoints.

In 2026, Salsify moved significantly further into AI commerce with SalsifyIQ.

Salsify describes SalsifyIQ as an intelligence layer built around a context-rich product knowledge graph containing several forms of knowledge, including shelf, PXM, brand, and ecosystem knowledge.

Salsify is also introducing an MCP layer that allows AI agents and third-party systems to access approved product attributes and contextual product data.

And the platform now goes beyond preparing data for conventional ecommerce channels.

Salsify has an OpenAI syndication channel that maps and validates customer product information against OpenAI’s Product Feed specification and publishes brand-approved product data to OpenAI for agentic shopping experiences.

This makes Salsify increasingly relevant not only to the digital shelf, but also to what it calls the agentic shelf.

What ProductKnowledgeGraph is building

ProductKnowledgeGraph starts further upstream in the trust problem.

PKG publishes canonical, source-backed product records designed so that AI systems can identify a product, inspect where product facts came from, and reach the manufacturer’s own source.

Its public infrastructure separates several states that are often incorrectly treated as equivalent:

A crawler visiting a product record does not prove that an AI system cited it.

A citation does not prove that a particular fact was used.

And retrieval does not prove that publishing the information caused a product to be selected.

PKG records those states separately.

This reflects a broader architectural objective:

Salsify vs. ProductKnowledgeGraph

The biggest difference is not whether both systems can contain product information.

They can.

The difference is what problem each system treats as primary.

This is why the products can appear similar from an AI-commerce perspective while representing different architectural philosophies.

Product data is not the same thing as product truth

Consider a simple example.

A retailer describes a backpack as:

Capacity: 48 L

The manufacturer’s current product documentation says:

Capacity: 45 L

A historical document contains:

Capacity: 48 L

All three values are product data.

But simply putting them into a knowledge graph does not make one of them universally true.

An AI system needs more context:

• Who made each claim?
• What product and variant does it refer to?
• When was the source observed?
• Is the source still current?
• Which issuer has authority for this particular claim?

PKG’s approach is therefore not to flatten every observation into a single undifferentiated fact.

Instead, the architecture preserves the relationship:

Conflicts can then be identified and resolved according to their context.

This distinction becomes particularly important when product information originates from multiple enterprise systems, manufacturers, retailers, marketplaces, certification bodies, or regulatory sources.

Manufacturer truth and retailer truth can both matter

AI commerce also makes the traditional concept of a single product “system of record” more complicated.

Consider price.

The manufacturer may be authoritative about:

• model, dimensions, materials, technical specifications

while a retailer may be authoritative about:

• current price, inventory, promotion, pickup availability

A certification body may be authoritative about a certification.

A regulator may be authoritative about a regulatory record.

Those sources should not necessarily overwrite one another.

They represent different claims from different issuers.

PKG is being designed around this issuer-aware model rather than assuming that every product attribute should originate from one universal authority.

Where Salsify is stronger

For enterprises evaluating these platforms, there are areas where Salsify is clearly the more mature choice.

Salsify already operates a broad enterprise PXM and syndication infrastructure.

It supports product information management, retailer connections, GDSN workflows, enhanced content, supplier onboarding, digital shelf operations, and a substantial technology and solution-partner ecosystem. Its technology partner directory includes companies such as Adobe, SAP, Salesforce, Shopify, commercetools, and Profitero.

Salsify reported that its customers published more than 5 billion products across the global commerce ecosystem in 2025.

An enterprise looking primarily for:

PIM + PXM + retailer syndication + digital shelf operations

should evaluate Salsify seriously.

ProductKnowledgeGraph is not attempting to reproduce that entire stack.

Where ProductKnowledgeGraph takes a different approach

PKG becomes relevant when the enterprise question changes from:

to:

The PKG model focuses on:

Product Identity

Resolving the exact product and its identifiers.

Issuer-aware Product Truth

Keeping manufacturer, retailer, certification, regulatory, and other claims attributable to their respective issuers.

Evidence and Provenance

Preserving where a claim came from and when the underlying source was observed.

Conflict Resolution

Detecting when authoritative or operational sources disagree instead of silently flattening them.

AI-native Publication

Making approved product records accessible through machine-readable product surfaces.

Retrieval Evidence

Distinguishing publication, crawling, retrieval, citation, fact use, and answer influence.

The last distinction is particularly important.

ProductKnowledgeGraph’s current public evidence supports the claim that AI systems can discover, retrieve, cite, and use PKG product records. It does not yet establish that publishing through PKG causes a repeatable increase in product selection or purchase routing.

That boundary is intentional.

Are Salsify and ProductKnowledgeGraph competitors?

Increasingly, yes.

The overlap is no longer theoretical.

Both companies are operating around enterprise product knowledge and the emerging agentic commerce layer.

Salsify now explicitly describes SalsifyIQ as using a product knowledge graph, provides MCP connectivity, positions itself for the agentic shelf, and provides an OpenAI syndication channel.

ProductKnowledgeGraph is building product identity, source-backed product truth, evidence, AI-readable publication, and retrieval measurement for the same emerging AI-commerce environment.

So it would be misleading to describe the companies as operating in completely separate categories.

But competitive overlap does not mean architectural equivalence.

Salsify is extending a mature PXM platform toward the agentic shelf.

ProductKnowledgeGraph is approaching the same transition from the product identity, verification, evidence, and AI-retrieval side.

That creates overlap — and potentially very different answers to the same enterprise problem.

Do enterprises have to choose one?

Not necessarily.

An enterprise architecture could theoretically contain both:

But that is an architectural possibility, not a currently claimed Salsify-PKG integration or partnership.

There is no reason to pretend otherwise.

Salsify is both a useful benchmark for ProductKnowledgeGraph and a company whose expansion into product knowledge and agentic commerce creates genuine competitive overlap.

For enterprise buyers, the more useful question is therefore not simply:

It is:

If the problem is:

Salsify is built around that problem.

If the problem is:

That is the problem ProductKnowledgeGraph is being built to solve.

And increasingly, large enterprises may have both problems at the same time.

The emerging enterprise stack

The transition to AI commerce may ultimately require enterprises to separate several functions that were previously treated as one product-data problem:

The important question is no longer merely whether product data exists.

It is whether an AI system can determine:

That is the problem ProductKnowledgeGraph is working on.