See Gravton in action for your Store! 🚀 Launching Self Serve

See Gravton in action for your Store! 🚀 Launching Self Serve

See Gravton in action for your Store! 🚀 Launching Self Serve

Ecommerce AI & GEO Glossary: Which Terms Do Ecommerce Teams Actually Need?

Share this article

Share this article

Explore with AI

Ecommerce AI & GEO Glossary: Which Terms Do Ecommerce Teams Actually Need?

Ecommerce AI & GEO Glossary: Which Terms Do Ecommerce Teams Actually Need?

AI shopping has developed enough terminology that two teams can use the same word while measuring different things.

Tarun Babbar

Grounding in AI search
Grounding in AI search

This GEO glossary defines the terms ecommerce teams use for AI shopping, one short entry each, written for the people who run catalogs and report results. Most reporting mistakes start when two of these terms are treated as one, such as a mention and a recommendation.

Mixing terms changes the numbers. For one brand we track, we asked the consumer versions of ChatGPT and Google AI Mode 125 shopper questions in October 2026 and kept each engine's latest answer (three questions ran on AI Mode only). They named the brand in all 80 answers to the 40 questions that used its name, and in 16 of 167 answers (10%) to the 85 that did not. Blend them into one "AI visibility" score and you get 39%, a figure that describes neither situation.

Key takeaways

  • Define the question set and denominator before you read any AI shopping percentage.

  • Keep products, sources and sellers as separate units in every report.

  • Report AI-attributed revenue on its own, and never add modeled revenue to it.

What do the core GEO terms mean?

Generative engine optimization (GEO)

In ecommerce, the work of making your products easy for AI assistants to find, match to a shopper's request, trust and recommend, with a buy link that points to your store. Closely related terms include answer engine optimization (AEO) and LLM optimization, and our GEO guide covers the full model.

AI shopping engine

An AI assistant or search feature that helps shoppers find, compare or buy products, such as ChatGPT, Google AI Mode, AI Overviews, Gemini or Perplexity.

The four checks

Gravton's working model of how an AI shopping answer gets from question to buy link: understand the need, match products, verify the facts, route the buyer. Each check leans on different sources: explainers for the need, feeds and product pages for the match, reviews and independent sources for verification, seller data for routing.

What do shopper-demand terms mean?

Shopper question

The need, comparison or buying decision a shopper puts to an AI assistant in their own words. Our guide to what shoppers ask AI sorts them into types.

Branded and unbranded questions

A branded question names your brand; an unbranded question describes a need without naming any brand. Report them separately, because answers to branded questions mention you by default and inflate any blended score.

Intent

The goal a shopper's question implies, such as finding a gift under $100 for a runner. Intent is a reading of the question, not data the shopper gave you.

Journey stage

Where the shopper is in the decision: discovery, evaluation or ready to buy. Google Merchant Center's AI performance insights use those three stages for English queries in Australia, Canada, India, New Zealand and the US (Merchant Center Help). Two shoppers can share an intent while sitting at different stages.

Query fan-out

The way an AI system breaks one shopping question into smaller searches, such as capacity, laptop fit and price for a travel backpack. See query fan-out.

Question set

The fixed list of shopper questions a team runs on AI engines to measure results over time. No AI platform publishes shoppers' prompts, so a question set is written to mirror them, not copied from logs. Its size says nothing about how much demand sits behind it.

What do product, catalog and site-access terms mean?

SKU

Stock keeping unit: one sellable product or variant, such as a shirt in one size and color. Say whether a measure counts SKUs or product lines.

Product attribute

A fact that describes a product, such as size, material, capacity or battery life. Shoppers stack several attributes in one question, so a missing one can drop a product from the answer.

Catalog consistency

Whether a product's facts match everywhere it appears, from your own site to your feeds and marketplace listings. When two listings disagree, the assistant has no single fact to rely on.

Catalog enrichment

Adding missing product facts, or rewriting vague ones, to close a known gap. Start with the questions you lose.

Product feed

A structured file of your products, prices and stock that you send to a platform such as Google Merchant Center. Google says Merchant Center data can help products appear in its AI responses (Google Search Central).

Product schema

Structured data, usually schema.org Product and Offer markup, that states product facts in a format machines can read (schema.org). Google says it is not required for its AI features, and it can only state facts you already have; see does product schema help AI shopping.

Crawlability

Whether a crawler can find and fetch your pages and the product facts on them. An AI engine generally cannot show a page it is blocked from crawling; our guide to AI crawlability covers the checks.

robots.txt

A file at your site's root that tells crawlers which URLs they may access. Google notes it is not a way to keep a page out of Google (Google Search Central), and OpenAI says sites that block its search crawler, OAI-SearchBot, will not appear in ChatGPT search answers, though they can still show as navigational links (OpenAI).

llms.txt

A proposed file that gives AI agents a plain summary of a site and links to its key pages (llmstxt.org). It is voluntary, and Google says you do not need such files to appear in its AI features; see llms.txt vs robots.txt.

What do source and evidence terms mean?

Source

Any page an AI system draws on for an answer, including your site, retailers, marketplaces, review sites, editors and forums. See which sources shape AI product recommendations.

Citation

A link an AI answer shows as the source for part of what it says. A citation shows that the answer drew on a page, not why the assistant picked a product.

Citation share

The share of citations in a question set that point to a given site or type of source. Your site can earn citations while a rival's product wins the recommendation.

Product trust signal

Evidence that makes a specific product claim believable, such as an independent test, a certification or owner reviews. Each signal backs a different claim; see product trust signals.

Knowledge gap vs trust gap

A knowledge gap means the AI lacks the facts it needs about your product. A trust gap means it has them but relies on another source to judge the product.

Evidence grades

Our three-level scale for how much support a claim about AI systems has. Strong: stated by a platform and backed by research. Moderate: follows from how systems work or is proven only in general. Inferred: plausible, with no direct evidence.

What do seller and revenue terms mean?

Merchant

The business that sells the product to the shopper. The maker and the merchant may be different companies, as when a marketplace sells your product.

Offer

One seller's terms for a product, including price, availability and shipping (schema.org). One product can carry many offers.

Merchant routing

Which seller the AI sends the shopper to after recommending a product. OpenAI names availability, price and quality among the factors ChatGPT uses to rank merchants (OpenAI Help Center).

Product loss vs merchant loss

A product loss means the AI recommended a rival's product. A merchant loss means it recommended yours but sent the shopper to another seller.

AI-shopping session

A visit to your site that started from an identifiable AI source, such as a link in a ChatGPT answer. It is also called an AI referral. Google Analytics 4 groups visits from sources it recognizes, such as ChatGPT and Gemini, in its AI Assistant channel; others land in Referral, or in Direct when no referrer is passed (Google Analytics Help).

AI-attributed revenue

Revenue from orders placed in sessions with an identifiable AI source. Treat it as a floor. A shopper who researched in an AI assistant and returned another way does not count.

AI-assisted revenue (modeled)

An estimate of revenue AI influenced without producing the final identifiable session; always name the method. Never add it to AI-attributed revenue.

What do AI shopping measurement terms mean?

AI visibility

How often your brand appears in AI answers across a defined question set. On its own it blends branded and unbranded questions, which is why the 39% above describes neither.

Recommendation

An answer that presents a product as a good fit for the shopper's request. A mention is weaker: the brand appears, but the assistant may not suggest buying it.

Recommendation share

How often your products make the AI's shortlist compared with competitors. Our recommendation share guide explains what to count.

SKU coverage

The share of priority products that appear for the shopper questions they should win. It catches what a brand-level score hides: AI answers can name a brand often while most of its products never appear.

Share of voice

Your share of impressions against a set of competitors you define. In Merchant Center's AI performance insights, it is your AI impressions divided by the total for your brand and those competitors, so it counts appearances, not recommendations.

Evidence levels

The four labels we give a number by how we produced it: observed, calculated, attributed, modeled. Never add modeled value to measured value.

Demand at risk

Shopper demand where your product fits but shows up weakly in AI answers. It estimates opportunity, not lost revenue; see AI visibility and revenue loss.

Which terms are most often confused?

These six pairs account for the reporting errors we see most often:

Do not confuse

Why they differ

Mention and recommendation

An answer can name a brand without suggesting it

Recommendation share and citation share

One counts products, the other counts sources

Citation and cause

A link proves use, not the reason a product won

AI-attributed and AI-assisted revenue

One is measured, the other is modeled

Question count and demand

A longer list does not mean more shoppers

Product fit and offer availability

A product that fits can still lack a good seller

How does Gravton help teams measure these terms?

Gravton's free audit applies these definitions to your own category. It splits your buyer questions into branded and unbranded, tracks products, sources and sellers as separate measures, and returns a Demand Snapshot with the fixes ranked. Measurement lets you see if your recommendations have increased week over week.

Get your free AI visibility audit and see these measures for your own products.

Frequently asked questions

Is AI-referred revenue the same as AI-attributed revenue?

Yes, in practice. Both mean revenue from orders placed in sessions that came from an identifiable AI source, such as a ChatGPT link, and we use AI-attributed revenue across our guides. Whatever you call it, keep it apart from modeled estimates of AI influence, which try to count shoppers who never clicked through.

Why do GEO tools report different numbers for the same brand?

They rarely measure the same thing. Tools differ in the questions they run, the share of branded questions, the engines they query, how many runs they average, and whether they count mentions, recommendations or citations. Before you compare two reports, check those five choices, since any one of them can move the result.

Which of these terms belong on a leadership dashboard?

We suggest five: recommendation share, SKU coverage, merchant-routing share, AI-shopping sessions and AI-attributed revenue. The first three come from question runs and the last two from analytics. Our AI-shopping KPIs guide explains how to read them together and what each should change.

Where do these definitions come from?

Platform terms such as share of voice, journey stages and robots.txt, and the llms.txt proposal, follow the published documentation linked in each entry. Terms such as the four checks, knowledge gap, trust gap and evidence grades are Gravton's own, and we use them the same way in every guide on this site.

What should you read next?

Sources

  1. Google Merchant Center Help. "AI performance insights." Accessed October 22, 2026.

  2. Google Search Central. "Optimizing your website for generative AI features on Google Search." Updated July 10, 2026.

  3. Google Search Central. "Introduction to robots.txt." Updated December 10, 2025.

  4. OpenAI. "Overview of OpenAI crawlers." Accessed October 22, 2026.

  5. OpenAI Help Center. "Shopping with ChatGPT Search." Accessed October 22, 2026.

  6. Howard, J. "The /llms.txt file." Proposed September 3, 2024; version 2 updated August 10, 2026.

  7. Schema.org. "Product" and "Offer." Accessed October 22, 2026.

  8. Google Analytics Help. "Default channel group." Accessed October 22, 2026.

  9. Gravton Labs. Tracked-brand data for one ecommerce brand, 247 answers from ChatGPT and Google AI Mode, October 2026. Internal dataset, anonymized.



Related resources

Continue with these related Gravton resources:

See how AI shops your category.

Map demand, products considered, sale routing and the first gaps your team should fix.

See how AI shops your category.

Map demand, products considered, sale routing and the first gaps your team should fix.

See how AI shops your category.

Map demand, products considered, sale routing and the first gaps your team should fix.