The CEI Benchmark

What we measure when we say a catalog is AI-ready

The Commerce Eligibility Index is grounded in an ongoing benchmark: a recurring measurement of how today’s AI shopping assistants actually read, trust, and recommend real product catalogs. It is the reference set behind every CEI score, and it grows every month.

Benchmark panelUpdated monthly
FOUNDATIONDIFFERENTIATIONRETRIEVALINTEGRITYAUTHORITY
At a glance

The reference set behind every CEI score.

An ongoing measurement of AI catalog readiness that grows every month.

Brands
100

brands tracked across the panel.

Verticals
10

consumer-retail categories.

Signals
16

readiness signals scored per brand.

Query volume
100,000+

AI shopping queries each cycle, and growing.

Cadence
Monthly

refreshed every cycle, with coverage expanding over time.

The verticals

Ten retail categories, so readiness reads in competitive context.

A brand’s readiness means more against its own category than in isolation. The benchmark spans ten verticals today, and the list expands as we add coverage:

  • Beauty and Cosmetics
  • Consumer Electronics
  • Drugstore and Discount Retail
  • Fashion and Apparel
  • General Merchandise and Marketplaces
  • Home and Furniture
  • Home Improvement
  • Jewelry and Accessories
  • Pet Supplies
  • Sporting Goods and Outdoor
The assistants we test

We ask the systems shoppers actually use.

We do not model what an assistant might do. We query the live systems directly and observe how they respond: ChatGPT, Claude, Gemini, and Perplexity.

How a brand is scored

Over 1,000 real shopping queries per brand.

For each brand we run over 1,000 LLM-generated shopping queries, drawn from the brand’s own product vocabulary (categories, attributes, use cases, and price and fit constraints), against the live assistants. We record whether the brand is found, described accurately, and recommended, and where it is missed, misrepresented, or substituted.

Those observations roll up through the 16 signals and five pillars (Foundation, Differentiation, Retrieval, Integrity, Authority) into a single 0 to 100 Commerce Eligibility Index. CEI is weighted so a strong pillar cannot paper over a weak one: the lowest pillar pulls the score, because an assistant only has to fail one check to leave a product off the shortlist.

What we report

Directional patterns, not shopper-level guarantees.

The benchmark is a panel, not a census of the retail sector, so we report tendencies rather than guarantees. It surfaces patterns such as:

  • How readiness varies by vertical and by assistant.
  • Where catalogs are most often misread: missing attributes, boilerplate descriptions, and claims that do not reconcile.
  • That high familiarity does not equal high accuracy. Well-known brands can still land in hallucination-risk positions, while cleaner brands can be under-discovered.
What is public, and what is yours

We publish the patterns. We protect the method, and your results.

We share the patterns and the benchmark itself so commerce leaders can see where the shelf is moving. We do not publish any individual brand’s score. We keep the exact signal weighting and scoring implementation as our own method, so results stay comparable across cycles and clients control their own findings. Your assessment is private to you.

Measure, do not guess

See where your catalog sits against the benchmark.

Request a CEI assessment and we will show you your score in competitive context.

Request a CEI assessment