Understand how AI discovers, evaluates, trusts and recommends online stores.

Persistent Merchant Identity measures how consistently and accurately AI systems understand and represent a brand across ChatGPT, Perplexity, Google AI Mode, Alexa for Shopping, and Apple Intelligence. A brand with low PMI gets understood differently by different systems, resulting in inconsistent recommendation outcomes.

Each AI system builds its understanding of your brand independently. If your signals are inconsistent across the web, your PMI breaks down. You might be recommended well on Perplexity but misunderstood on ChatGPT.

A supplement brand positions itself as performance nutrition for endurance athletes on its website, but Reddit mentions it mainly in casual fitness contexts and press covers it as a weight loss product. Low PMI. Inconsistent recommendations across platforms.

Merchant Interpretability is the degree to which AI systems can build an accurate, complete understanding of a merchant entity from all available signals. It combines Semantic Clarity, Persistent Merchant Identity, structured data completeness, and external signal consistency. High Merchant Interpretability means every AI system that encounters a brand builds the same accurate understanding of what it sells, to whom, and why.

AI recommendation decisions are only as good as AI understanding. A merchant that AI misinterprets, partially understands, or understands inconsistently across platforms will never achieve reliable recommendation share. Merchant Interpretability is the foundation of everything else in AI Commerce Intelligence.

An outdoor apparel brand has a clear brand description in Organization schema, specific product categories in JSON-LD, an llms.txt that explicitly defines its niche, and consistent external mentions as a premium hiking brand. High Merchant Interpretability. AI understands it the same way across ChatGPT, Perplexity, and Google AI Mode.

Machine-Readable Commerce is the complete architectural approach to making a store legible to AI agents. It goes beyond schema markup. It covers every layer where AI needs to extract, verify, or act on your data.

Most e-commerce stores are built for humans. AI agents cannot see beautiful design or emotional copy. They read structured signals. A machine-readable store gets recommended. One that is not gets skipped regardless of product quality.

A skincare brand has great photography and copy, but prices render via JavaScript, no Product schema exists, and the return policy is in a JavaScript modal. The store is not machine-readable. Its score reflects that.

AI Extractability measures how easily AI can chunk, parse, and extract answers from a store's content. High extractability means AI can pull answers effortlessly. Low extractability means AI skips the store even if the content is good.

AI does not read pages the way humans do. It extracts structured answers and evaluates them against buyer queries. Good content in a technically broken structure is still invisible.

A wellness brand has a detailed FAQ but it renders entirely in JavaScript. AI cannot read it. The same content in server-rendered HTML with FAQ schema would score highly on AI Extractability.

Prompt Visibility is whether a brand appears when buyers submit purchase-intent prompts to AI systems. A brand can have a high AI Commerce Score but zero Prompt Visibility if its positioning does not match buyer query language.

Every buyer prompt that AI answers without mentioning your brand is a potential sale that went to a competitor. Measuring Prompt Visibility reveals the exact scale of your AI visibility gap.

A pet food brand tests 50 prompts. It appears in 6 of 50. Prompt Visibility rate is 12 percent. The category leader appears in 38 of 50, or 76 percent.

AI Visibility is the foundational measure of whether an e-commerce store exists in the eyes of AI shopping systems. A store with high AI Visibility is discoverable, interpretable, and surfaceable by AI agents when relevant buyer queries are submitted. AI Visibility is necessary but not sufficient for recommendation. A store can be visible to AI and still not get recommended.

The shift from search-first to recommendation-first commerce means that visibility in AI systems is now the most important form of digital discoverability. Stores that are invisible to AI are invisible to a growing share of buyers.

A store with clean schema markup, server-rendered prices, and clear product descriptions has high AI Visibility. A store with JavaScript-rendered content, no schema, and generic copy is invisible to most AI shopping agents.

AI Discoverability measures the ease with which AI systems can find, parse, and surface a store in response to buyer queries. It is broader than Prompt Visibility, which measures appearance in specific prompts, and broader than AI Visibility, which measures whether AI can read the store. AI Discoverability encompasses the full path from AI crawling to query matching to surfacing in responses.

AI Discoverability is the upstream driver of Recommendation Share. A store that is hard to discover will never appear in AI responses regardless of how good its products are. Discoverability depends on crawlability, structured data completeness, semantic clarity, and external citation presence.

A store blocks GPTBot in robots.txt, has no schema markup, and uses JavaScript to render all product content. Its AI Discoverability is near zero. ChatGPT cannot find it. Perplexity cannot index it. It never appears in buyer queries regardless of product quality.

The AI Visibility Layer is the complete stack of technical, semantic, and trust signals that determine how visible and recommendable a merchant is within AI commerce ecosystems. It is the new optimization layer in e-commerce, sitting above SEO and paid acquisition. The AI Visibility Layer includes structured data, crawl accessibility, semantic commerce signals, trust verification, external citation networks, and agent-readable commerce flows.

Every era of e-commerce growth has had a dominant optimization layer. The search era had SEO. The social era had content and paid acquisition. The AI era has the AI Visibility Layer. Brands that master this layer first will compound an advantage that grows as AI agents handle more purchase decisions.

Building the AI Visibility Layer means optimizing schema markup, ensuring crawler accessibility, developing semantic commerce copy, building external authority signals, enabling agent-readable checkout flows, and maintaining consistent Persistent Merchant Identity across all AI platforms simultaneously.

AI Readability measures how easily AI crawlers can parse, chunk, and extract meaningful content from a store. It covers heading hierarchy, server-rendered text, FAQ schema, semantic HTML structure, and the absence of JavaScript-only content blocks. High AI Readability means AI agents can extract the store's full signal set on a single crawl pass.

AI agents do not render JavaScript. They do not experience visual design. They extract what is in the HTML source on first pass. If your content requires execution, interaction, or visual rendering to appear, it is invisible to every AI shopping agent.

A store with a clean H1, server-rendered product descriptions, and FAQ schema scores high on AI Readability. A store where the same content loads via JavaScript after page interaction scores near zero.

Recommendation Visibility is the binary and graded measure of whether a brand appears when AI systems process a relevant buyer query. Unlike Prompt Visibility which measures appearance rate, Recommendation Visibility focuses on the specific query-response moment. A brand with high Recommendation Visibility is consistently surfaced in the queries that matter most to its niche.

A brand can exist in AI training data and still have zero Recommendation Visibility at query time. The gap between existing and being surfaced is where most brands lose revenue. Recommendation Visibility measures that gap directly.

A skincare brand is known to AI but only appears in broad queries. For high-intent searches like fragrance-free moisturizer for eczema under $50, its Recommendation Visibility is near zero. Fixing semantic positioning raises it.

Trusted But Invisible describes a store that has built real-world trust, authority, and brand recognition but is systematically excluded from AI recommendations because its content cannot be extracted by AI crawlers. The problem is not trust. The problem is technical accessibility. It is the most frustrating AI visibility failure because the brand has done the hard work but cannot benefit from it.

This is the most common failure pattern among established brands. They have thousands of reviews, years of press coverage, and strong brand recognition. But their prices are JavaScript-rendered, their FAQ is in a modal, and their schema is missing. The fix is purely technical.

A wellness brand with 15,000 Trustpilot reviews, 200 press mentions, and strong DTC revenue scores 38 on AI Commerce Score. Its content is locked behind JavaScript. Fix the extractability and the authority immediately converts to recommendations.

The Semantic Commerce Layer is the structured layer of product copy, FAQ content, schema descriptions, and llms.txt positioning that translates a store's products into the semantic signals AI uses to match brands to buyer queries. Without it, a brand with perfect technical infrastructure still fails to appear in the queries that drive revenue.

AI does not match brands by reading product names or SKUs. It matches by semantic understanding of what a brand does, who it serves, and what problems it solves. Brands that build this layer using buyer query language get matched to more queries and appear more often.

A brand sells a product listed as Performance Trail Running Shoes Model XR7. Its Semantic Commerce Layer translates this into lightweight trail shoes for ultra marathon training, aggressive grip for muddy conditions, under $180. That gets matched to buyer searches for the best trail shoes for ultras.

Intent Misalignment describes a store that is credible and technically solid, but whose positioning, copy, and content are misaligned with how buyers actually ask AI for recommendations in that category. The fix is content strategy, not technical optimization.

AI matches stores to buyer queries semantically. If a buyer asks for the best collagen for runners and a supplement brand only talks about general wellness, AI will not match it to that query regardless of authority or technical quality.

A supplement brand has strong authority but focuses on general wellness messaging. Searches like best protein for marathon training and supplements for endurance athletes never match the brand. Rewriting content around buyer prompt language fixes this.

The Commerce Knowledge Graph is the semantic knowledge structure that AI systems construct to understand the e-commerce internet. It connects merchant entities to product categories, use cases, buyer intents, price points, trust signals, and competitive relationships. Unlike the AI Commerce Graph which maps recommendation flows, the Commerce Knowledge Graph maps semantic understanding. Atom Foundry measures each merchant's position and strength within this graph as part of the AI Commerce Score.

AI recommendation decisions are built on top of knowledge graph relationships. A merchant that is strongly represented in the Commerce Knowledge Graph will appear in more query matches, with higher confidence, across more buyer intent clusters. Building knowledge graph presence is the long-term strategy for AI commerce dominance.

In the Commerce Knowledge Graph, a premium pet food brand is connected to nodes for grain-free dog food, senior dog nutrition, vet-recommended, and subscription pet supplies. Each connection is a pathway to buyer queries. Strengthening these connections is the goal of semantic commerce optimization.

The Buyer Intent Graph is the structured map of the real purchase-intent prompts buyers submit to AI systems, organized by category and connected to the brands AI returns for each. Where the AI Commerce Graph maps how AI recommends, the Buyer Intent Graph maps what buyers actually ask for. Atom Foundry builds it by collecting high-intent prompts across categories and recording which brands surface for each one.

You cannot measure Recommendation Share without first knowing the full set of intents that define a category. The Buyer Intent Graph is that denominator. It turns the question of whether you are recommended into the more useful question of which of the queries that actually drive purchases you are recommended for.

In coffee, the Buyer Intent Graph holds intents like best espresso beans, best low acid coffee, and best coffee subscription. Each connects to the brands AI names for it. A roaster strong on one intent but absent from the rest has narrow Buyer Intent Coverage.

Hidden Authority describes a store that has real authority and trust signals but scores poorly on AI Extractability. AI may recognize the brand but cannot extract structured answers. The result: AI knows about the brand but cannot confidently recommend it.

This is the most fixable AI visibility problem. The brand has already done the hard work of building authority. The fix is purely technical. Improving heading structure and adding FAQ schema can move a Hidden Authority brand into recommendation flows quickly.

A wellness brand with 10 years of press coverage and 2,000 reviews still scores 41 because its FAQ is JavaScript-rendered with no schema. Fix the extractability and the authority converts to recommendations.

The AI Trust Graph is the network of external signals AI uses to validate a brand's trustworthiness. Reddit mentions, press coverage, review platform presence, and citations in articles all count. AI does not just trust what a brand says about itself.

AI systems are trained on the entire internet. A store with no AI Trust Graph presence is treated as unverified. Strong external presence creates a much higher trust baseline in AI recommendations.

Brand A has a beautiful website but no external presence. Brand B has an average website but 200 Reddit mentions, three press articles, and 500 Trustpilot reviews. AI recommends Brand B.

Machine Trust Signals are the structured, crawlable signals that AI systems verify before including a store in recommendation flows. They are distinct from visual trust cues that humans see. Machine Trust Signals include AggregateRating schema, return policy text in HTML, SSL certificate presence, legal identity in the footer, shipping information in crawlable text, and contact details. They must be machine-readable to count.

A trust badge image means nothing to AI. A star rating in a visual widget means nothing to AI. Machine Trust Signals are what AI can parse and verify programmatically. Without them, a store with excellent human-facing trust is still treated as unverified.

A store has a beautiful trust badge showing a 30 day return policy. AI cannot read the badge. The same store with a visible return policy paragraph in HTML and AggregateRating schema on its product pages has strong Machine Trust Signals.

Recommendation Authority is the aggregate strength of a merchant's AI-verifiable signals across trust, semantic clarity, external validation, and commerce accuracy. A merchant with high Recommendation Authority is not just visible to AI but deeply credible to it. AI systems recommend high-authority merchants first, most often, and with the highest confidence. Recommendation Authority compounds over time as external signals accumulate.

Domain Authority was the currency of SEO. Recommendation Authority is the currency of AI commerce. Brands that build it early compound a structural advantage that becomes increasingly difficult for competitors to close. It is the long-term moat in the Recommendation Economy.

Brand A has perfect schema markup, reviews averaging 4.9 stars with AggregateRating, 500 Reddit mentions, 20 press citations, and a high-quality llms.txt. Its Recommendation Authority is high. Brand B has equivalent products but none of these signals. AI recommends Brand A first across nearly every relevant query.

AI Recommendation Eligibility is the binary state of whether a store has met the minimum requirements to enter AI recommendation flows. It is not a score or a ranking. It is a gate. A store that fails the AI Trust Layer, has no structured data, or is blocked to AI crawlers is ineligible for recommendation regardless of product quality, brand recognition, or marketing spend. Eligibility must be established before any optimization is meaningful.

The most important insight in AI Commerce Intelligence is that recommendation is not a spectrum for every store. For stores below a minimum threshold, it is binary. You are either in the recommendation pool or you are not. Understanding and achieving eligibility is the first objective before any advanced optimization.

A store with no schema markup, JavaScript-rendered prices, a robots.txt that blocks GPTBot, and zero external mentions is ineligible for AI recommendations. No amount of content optimization will change that until the eligibility blockers are removed first.

The Recommendation Eligibility Score quantifies how close a store is to the minimum threshold required to enter AI recommendation flows. Unlike the AI Commerce Score which measures overall recommendation quality, the Recommendation Eligibility Score focuses specifically on the gate-level requirements. A store below eligibility threshold gets zero recommendations regardless of its score on other dimensions. The Recommendation Eligibility Score shows exactly how far away eligibility is and what is blocking it.

For stores below the eligibility threshold, optimizing content or positioning is wasted effort. The first priority is always eligibility. The Recommendation Eligibility Score makes that priority measurable and actionable.

A store scores 38 on AI Commerce Score overall but its Recommendation Eligibility Score is 2 out of 5 because it blocks GPTBot, has no schema, and has no return policy in HTML. Fixing all three moves the eligibility score to 5 and unlocks the recommendation pool.

The AI Commerce Graph is the complete network of relationships that AI systems build between brands, products, trust signals, citations, buyer intents, and semantic relevance. It is the emergent structure across ChatGPT, Perplexity, Google AI Mode, and Alexa for Shopping that determines which brands get recommended, in which contexts, to which buyers.

Brands that understand their position in the AI Commerce Graph can proactively build a moat. Understanding which competitors are adjacent, which signals strengthen your position, and where your Recommendation Routing flows is the foundation of long-term AI visibility strategy.

A skincare brand's position connects it to nodes for sensitive skin, organic ingredients, and under $60 pricing. Each node connects to buyer queries. Strengthening any node connection increases Recommendation Share for that query cluster.

Recommendation Routing describes the dynamic process by which AI systems direct specific buyer queries to specific brands. Two brands can both be present in the AI Commerce Graph but receive completely different query traffic based on their positioning signals, trust scores, and semantic alignment.

Optimizing for Recommendation Routing means targeting the specific query flows with the highest buyer intent for your category. A brand routed to budget skincare queries drives different revenue than one routed to premium natural skincare queries.

Brand A gets routed to budget skincare queries. Brand B gets routed to premium natural skincare queries. Same category, completely different buyer audiences and average order values.

Recommendation Confidence measures how strongly AI endorses a store when it appears. A weak mention and a first confident recommendation are very different commercial outcomes. Confidence is built through verified reviews, clear policies, and accurate commerce data.

Appearing in AI responses is not enough. A weak mention converts at a fraction of the rate of a strong first recommendation. Recommendation Confidence determines quality of AI visibility, not just quantity.

A low confidence mention hedges, noting a brand might be worth considering while admitting limited information about its return policy. A high confidence mention names the brand directly as one of the most consistently recommended options in the category. Same brand, very different commercial outcome.

Recommendation Density measures how broadly a brand appears across adjacent buyer intents and related query clusters. A brand with high Recommendation Density is more resilient to changes in how buyers phrase queries and more deeply embedded in AI recommendation flows.

Recommendation Share measures presence within a defined prompt set. Recommendation Density measures how far that presence extends. High density brands are harder to displace because they appear across many intent clusters, not just one.

A running shoe brand appears in 71 percent of best running shoes prompts. But it also appears in searches for best shoes for flat feet, marathon training gear, and injury prevention footwear. That cross-intent presence is high Recommendation Density.

Recommendation Position tracks where in an AI response a brand appears and with what level of endorsement. The four positions are first recommendation, supporting mention, comparison mention, and weak citation. Position determines conversion quality, not just visibility.

Being the first confident recommendation converts very differently from being a weak supporting mention. Improving from comparison mention to first recommendation can double or triple AI-referred revenue without any change in traffic volume.

Brand A appears in 60 percent of prompts as the first recommendation. Brand B appears in 80 percent but always as a supporting mention. Despite lower share, Brand A likely drives more revenue because its position is stronger.

The Recommendation Graph is the network of relationships AI systems build between brands, products, trust signals, citations, buyer intent, and semantic relevance. It is the underlying structure AI uses to decide which brands belong together and which brand best answers a given buyer query.

Understanding your position in the Recommendation Graph is the foundation of long-term AI visibility strategy. Brands that map their position and understand their adjacencies will dominate AI recommendations over time.

In skincare, AI builds a Recommendation Graph where brands cluster around sensitive skin, clean ingredients, and sustainable packaging. A brand optimizing for these nodes enters the cluster most frequently queried by buyers.

AI Commerce Infrastructure is the umbrella layer describing the complete technical and semantic stack that enables AI systems to discover, evaluate, trust, and transact with e-commerce stores. It encompasses structured data, trust signals, semantic commerce layers, agent-readable flows, and persistent merchant identity. Atom Foundry maps AI Commerce Infrastructure across the e-commerce internet as the foundation of the AI Commerce Graph.

Just as the web required HTTP infrastructure and search required SEO infrastructure, the AI commerce era requires AI Commerce Infrastructure. Brands that build this layer now will have a compounding structural advantage as AI agents handle a growing share of purchase decisions.

A store's AI Commerce Infrastructure includes its JSON-LD schema stack, llms.txt quality, checkout agent accessibility, trust signal completeness, and semantic commerce layer. Together these determine how well that store is integrated into the AI commerce ecosystem.

The Commerce Protocol Layer is the standardized stack of machine-readable files, APIs, and structured formats that allow AI agents to interact with e-commerce stores at the protocol level rather than the HTML level. It includes llms.txt, llms-full.txt, agents.md, Schema.org markup, product feed APIs, and checkout agent APIs. The Commerce Protocol Layer is to AI commerce what HTTP was to the web. It defines the rules by which AI agents and merchants communicate.

As commerce moves from human-navigated to agent-navigated, the Commerce Protocol Layer becomes the primary interface between merchants and the AI systems that route buyers to them. Stores without a complete protocol layer are invisible to agent-native commerce flows.

A store's Commerce Protocol Layer includes a high-quality llms.txt that accurately describes its product catalog, a valid agents.md that enables checkout flows, complete Product schema on all product pages, and a Google Shopping feed that stays in sync with live pricing. All four layers working together make the store protocol-native for AI commerce.

The Recommendation Gap is the quantified difference between a brand's current AI Recommendation Share and the share held by the top competitor in its category. It represents the revenue opportunity available through AI Commerce optimization. A large Recommendation Gap means significant revenue is flowing to competitors through AI recommendation channels.

The Recommendation Gap translates AI visibility data into commercial stakes. A brand with 12 percent Recommendation Share competing against a category leader at 67 percent has a 55 point gap. Every point of gap is estimated revenue flowing to a competitor through AI channels.

A protein supplement brand appears in 14 percent of relevant buyer prompts. The category leader appears in 61 percent. The Recommendation Gap is 47 points. Closing it is the commercial case for AI Commerce optimization investment.

AI Recommendation Rate is the percentage of tested purchase-intent prompts in which a merchant appears as a recommendation, in any position and at any confidence level. It is the broadest measure of AI visibility. A merchant with a high AI Recommendation Rate appears across many queries. A merchant with a low rate is systematically excluded from most recommendation flows in its category.

AI Recommendation Rate is the most direct measurement available of whether AI commerce optimization is working. It converts every technical improvement into a measurable business outcome: did our brand appear in more buyer conversations this month.

A home goods brand tests 100 purchase-intent prompts relevant to its niche. It appears in 23. Its AI Recommendation Rate is 23 percent. After fixing schema markup and niche positioning it retests and scores 41 percent. The rate improvement is directly attributable to the optimizations made.

Buyer Intent Coverage is the percentage of the distinct buyer intents in a category for which a brand appears in AI recommendations at least once. A brand can hold strong Recommendation Share on a few prompts yet have low Coverage if it never surfaces for the rest. Coverage measures breadth across the Buyer Intent Graph, not depth on any single intent.

Recommendation Share can be inflated by dominating one or two prompts. Buyer Intent Coverage reveals how much of a category's actual demand a brand is even eligible to capture. Low coverage is hidden risk: whole clusters of buyers never see the brand.

A supplement brand appears for searches like best creatine and best pre workout but never for the other 18 intents in its category. Its Buyer Intent Coverage is 10 percent, even if its share on those two prompts is high.

Intent Authority is the strength with which AI associates a brand with a particular buyer intent. A brand with high Intent Authority is not just present for an intent but treated as a default answer to it, appearing early and described with confidence. It is intent specific: a brand can hold high authority for one intent and none for an adjacent one.

Being the authority for best magnesium for sleep is worth more than appearing weakly across many intents. Intent Authority concentrates Recommendation Position and Confidence on the exact queries that drive a brand's economics, rather than spreading thin.

For searches like best fresh dog food delivery, AI consistently names one brand first and describes it in the most detail. That brand holds high Intent Authority for that intent, even if larger competitors appear lower in the same answer.

Recommendation Demand is the volume of buyer intent directed at a category or specific intent cluster through AI systems. It is the AI-era equivalent of search volume: how much buyer attention actually flows through a given set of prompts. High Recommendation Demand intents are where recommendation visibility is most commercially valuable.

Not all intents are worth the same. Winning a high-demand intent like best protein powder matters more than dominating a niche one. Recommendation Demand tells a brand where to concentrate by sizing the buyer attention behind each part of the Buyer Intent Graph.

Best running shoes carries far higher Recommendation Demand than best shoes for plantar fasciitis in wide sizes. A brand prioritizes coverage of high-demand intents first, then expands into the long tail.

Agent-Readable Commerce is the next layer beyond Machine-Readable Commerce. Where Machine-Readable Commerce ensures AI can parse and understand a store, Agent-Readable Commerce ensures autonomous AI agents can navigate the entire commerce flow end to end, from product discovery through price verification through checkout completion. It is the infrastructure requirement of the agentic commerce era.

Shopify launched Agentic Storefronts in May 2026. AI agents now complete purchases autonomously. Stores that are not agent-readable will be bypassed in the growing share of AI-completed transactions. This is not a future trend. It is already happening.

An agent-readable store has llms.txt, llms-full.txt, and agents.md. Cart and checkout are accessible without JavaScript popups. Prices and availability are in HTML. Returns and shipping are in crawlable text. An agent can complete a purchase without any human input.

Agentic Commerce is the emerging model of e-commerce in which AI agents act as autonomous buyers. A human delegates a purchase intent to an AI agent, and the agent independently discovers products, compares stores, evaluates trust and pricing, and completes the transaction. Shopify launched its Agentic Dashboard in May 2026, generating llms.txt, llms-full.txt, and agents.md for every store to enable this model.

Agentic Commerce is not a future prediction. Shopify confirmed it and built infrastructure for it in May 2026. Stores that are not configured for agentic flows will be systematically bypassed as AI agents handle a growing share of transactions through 2026 and 2027.

A buyer tells Apple Intelligence to find the best organic coffee subscription under $40 and set it up. The AI agent searches, evaluates stores, selects the highest-scoring option, and completes checkout. The buyer never opens a browser. Agentic Commerce in action.

Agent Transaction Readiness is the degree to which an autonomous AI agent can complete a full purchase on a store without human help: discover the product, verify price and availability, add to cart, and check out. It sits one layer beyond Agent-Readable Commerce, which ensures an agent can read the store. Transaction Readiness ensures an agent can act on it.

As agentic checkout scales, being recommended is not enough if the agent cannot complete the purchase. A store an agent can read but not transact on loses the sale at the last step. Agent Transaction Readiness is the difference between being considered and being bought from.

An agent selects a store as the best match, then abandons because checkout requires a JavaScript popup and account creation. The store had recommendation visibility but failed Agent Transaction Readiness. The order goes to the next option.

AI Actionability is the degree to which a store's data is structured precisely enough for an AI agent to take action on it: compare a price, confirm stock, apply a filter, or trigger a checkout. Readable content can still be unactionable if prices, variants, or availability are ambiguous or unstructured. Actionability is the bridge between extraction and transaction.

An agent that reads a vague price range cannot reliably compare or buy. Actionable data is exact, structured, and current: a numeric price in schema, a boolean availability, a defined variant. Without it, an agent either guesses or skips, and skipping is the safer choice for the agent.

A product page shows a price described only as starting around $29 in text, but the actual variant prices live in a JavaScript widget. A human understands it. An agent cannot act on it. Putting exact prices and availability in Product schema makes the page AI Actionable.

Recommendation Velocity measures how quickly a brand's AI Commerce Score and Recommendation Share are changing over time. Positive velocity means the brand is becoming more visible. Negative velocity means competitors are gaining ground.

A static score tells you where you are. Velocity tells you which direction you are heading. A brand at 55 with positive velocity of plus 8 per month will overtake a brand at 68 with negative velocity.

In January a fashion brand scores 48. In March it scores 61. Recommendation Velocity is plus 6.5 per month. It will reach Moderate Confidence within two months at this pace.

The Recommendation Economy is the structural shift from a search-and-click model of commerce to an AI-mediated model where intelligent systems recommend, compare, and transact on behalf of buyers. In the Recommendation Economy, the entity that controls recommendation determines who gets the sale. Google controlled the Link Economy. AI agents control the Recommendation Economy.

The Recommendation Economy is not a future prediction. Shopify confirmed rapid year-over-year growth in AI-referred orders in early 2026. The question is not whether the Recommendation Economy exists. The question is whether your brand is optimized for it.

In the Link Economy, a buyer searched for running shoes and picked from 10 Google results. In the Recommendation Economy, a buyer tells Apple Intelligence to find the best trail running shoes for wide feet under $120 and receives two or three confident recommendations. Winner takes most.

ACO, or Agentic Commerce Optimization, is the discipline of preparing a store for AI shopping agents rather than search engines. It covers product data structure, trust signals, and commerce infrastructure together, since an agent has to read, trust, and act on a store in a single pass.

An AI shopping agent does not browse a store the way a human does. It reads structured data, checks trust signals, and decides whether to recommend or transact in seconds. A store with strong SEO but weak ACO is often invisible to the agent completely.

A DTC skincare brand ranks on page one of Google but scores 41 out of 100 on ACO. AI shopping agents cannot verify its return policy or shipping cost from structured data, so they recommend a competitor instead.

GEO is the practice of structuring content so generative AI systems such as ChatGPT, Perplexity, and Google AI Mode can extract, cite, and recommend it inside a generated answer, the same way SEO made pages legible to search crawlers.

AI systems do not rank pages, they synthesize answers. Content that is not structured for extraction and citation never makes it into the answer, no matter how well it would have ranked in search.

A supplement brand rewrites its ingredient page into direct question and answer format. Citations in AI answers rise from 2 per month to 34 per month within six weeks, with no change in search ranking.

SEO and GEO share some tactics but different goals. SEO wins a ranking position a human clicks through. GEO wins a citation inside an answer the human never has to click past.

A page can rank first in Google and still never appear in an AI answer, because AI extraction and search ranking are evaluated by different mechanisms entirely.

A mattress brand ranks second in Google for best mattress for back pain but appears in zero of twenty tested ChatGPT answers for the same query, because its content is comparison table only with no extractable prose.

AI Shopping describes the buyer side of the shift. Instead of typing a query into a search box and browsing links, buyers ask an AI assistant a question and act on whatever it recommends.

This shift moves the entire purchase decision inside the AI system. A store's visibility to that system now matters more than its position on a results page.

One in four buyers under thirty five now start a purchase research journey inside ChatGPT or a similar assistant rather than a search engine.

Semantic Commerce means a store's product data speaks buyer language, such as waterproof trail shoes for wide feet, instead of internal naming, such as Model TX-4 Wide.

AI matches buyer prompts to product meaning, not product names. A mismatch between how a store names things and how buyers ask for things is invisible to a human shopper but fatal to an AI match.

A shoe brand renames Model TX-4 Wide to Wide-Fit Waterproof Trail Runner across its product feed. Matches against buyer queries containing the word wide rise from 9 percent to 61 percent.

AI Authority Signals are the concrete, checkable proof points, such as press coverage, review volume, expert citations, and third party mentions, that an AI system scans before trusting a brand enough to recommend it.

AI does not take a brand's word for its own quality. It looks for independent confirmation elsewhere on the internet. Without visible authority signals, even accurate product claims get discounted.

Two supplement brands make the same claim. One has 40 independent reviews and three press mentions. The other has neither. AI recommends the first brand in 8 of 10 test prompts and the second in 1 of 10.

An AI Commerce Audit is a structured scan across a store's product data, trust signals, and technical infrastructure that produces a scored, prioritized list of what is blocking AI visibility and recommendation.

Most stores do not know why they are missing from AI recommendations. An audit turns a vague problem into a specific, ranked list of fixes with the largest impact first.

An audit of a home goods store finds 18 of 40 checks failing, with missing FAQ schema and inconsistent pricing data responsible for over half the AI Commerce Score gap.

AI Commerce Intelligence is the analysis layer that sits on top of raw scoring and graph data. It turns a number and a network map into a specific, prioritized set of actions a merchant can take.

A score alone tells a merchant where they stand. Intelligence tells them what to do about it, in what order, and what impact to expect from each fix.

A brand's AI Commerce Score is 52. AI Commerce Intelligence identifies that fixing FAQ schema alone would be worth an estimated plus 9 points, more than any other single change available.

AI Interpretability is the gap between reading and understanding. A page can be fully readable by AI and still be misinterpreted if the meaning is ambiguous, inconsistent, or buried in jargon.

A store that AI can read but cannot correctly interpret gets matched to the wrong buyer intent, or no intent at all. Readability without interpretability still fails to convert into a recommendation.

An AI system correctly extracts a product description that calls a jacket versatile outerwear but cannot determine whether that means rain, wind, or cold protection, so it excludes the product from all three searches.

AI Readiness is a snapshot measure of whether a store's data, structure, and trust signals currently meet the minimum bar for AI systems to read, evaluate, and consider recommending it.

Readiness is the gate before scoring even matters. A store that is not AI ready may never be evaluated at all, regardless of how good its actual products or trust signals are.

A new DTC brand launches with a JavaScript-only product catalog. It scores 0 on AI Readiness, not because its products are weak, but because AI systems cannot parse the catalog to evaluate them.

AI Structured Signals are the schema markup, structured data feeds, and machine-readable markers a store exposes so AI systems can parse products, pricing, and policies precisely rather than inferring them from prose.

AI systems trust structured data far more than they trust inferred prose, because structured data is unambiguous. A price in a schema field is certain. A price mentioned in a paragraph is a guess.

A store adds Product and Offer schema to its top 200 pages. AI extraction accuracy for price and availability rises from 61 percent to 98 percent on those pages.

AI Trust Confidence measures how certain an AI system is that a store's trust signals, such as return policy, reviews, and business details, are accurate and current, not simply present.

A trust signal that exists but looks stale, inconsistent, or unverifiable lowers confidence even if the underlying information is technically correct. AI treats uncertain trust as no trust.

A store's return policy page has not been updated in three years and conflicts with a newer policy mentioned on its FAQ page. AI Trust Confidence for that store drops even though a return policy is present.

ChatGPT Shopping is OpenAI's built-in product search and recommendation feature, which surfaces, compares, and links to merchant products directly inside a conversational answer rather than a separate results page.

ChatGPT Shopping pulls from structured product data and trust signals, not general web ranking. A store invisible to this feature loses a growing share of buyers who never leave the chat interface.

A buyer asks ChatGPT for the best noise canceling headphones under 200 dollars. ChatGPT Shopping returns three products with prices and links, none from stores lacking structured product data.

ChatGPT Visibility measures whether and how often a specific brand appears when buyers ask ChatGPT for recommendations in its product category, tracked across a consistent set of test prompts.

ChatGPT is the largest single AI shopping surface by user volume. Low ChatGPT Visibility in a category usually means low visibility across most other AI assistants as well, since they draw on similar signals.

A pet food brand tests 40 category prompts in ChatGPT. It appears in 5, for a ChatGPT Visibility rate of 12.5 percent, compared to 70 percent for the category leader.

Human View vs AI View describes the gap between a page as a human sees it in a browser, fully rendered with images and styling, and a page as an AI crawler reads it, often stripped down to raw HTML with no JavaScript execution.

A page can look complete and persuasive to a human while being nearly empty to an AI system, if key content loads only through JavaScript or sits inside a visual element with no underlying text.

A product page shows a rendered comparison chart to human visitors. The AI View of the same page shows an empty div, because the chart is drawn entirely in JavaScript with no text fallback.

llms.txt is a plain-text file, placed at a store's root domain, that gives AI systems a structured summary of what the site sells, how it is organized, and how it should be interpreted, similar in spirit to a robots.txt file but written for language models rather than crawlers.

AI systems increasingly check for llms.txt as a fast, low-ambiguity summary before crawling a full site. A missing or empty file is a small but real signal that a store has not prepared for AI access.

Shopify's Agentic Dashboard auto-generates a llms.txt file for every store. Two stores both have the file, but one filled it with real product and policy summaries while the other left the auto-generated placeholder. AI systems score them very differently.

Prompt Visibility Testing is the practice of running a consistent, representative set of real buyer prompts through AI systems on a recurring basis to measure whether, how often, and in what position a brand appears.

Prompt Visibility changes over time as AI models update and competitors improve. A single test is a snapshot. Testing on a schedule turns that snapshot into a trend a merchant can act on.

A brand runs the same 50 prompts through ChatGPT and Google AI Mode every month. Over one quarter, its appearance rate climbs from 12 percent to 27 percent after fixing schema and FAQ content.

Retrieval Intelligence describes how effectively an AI system searches its available knowledge, including a store's own content, to find and surface the most relevant answer to a specific buyer query.

Even a store with excellent content loses if the AI system's retrieval step never surfaces that content for the right query. Strong retrieval intelligence connects the right buyer question to the right existing answer.

A store has an accurate, detailed sizing guide, but it is retrieved for only 4 of 30 sizing related buyer prompts because the page title and headings do not match common buyer phrasing.

Shopify AI Readiness measures how prepared a Shopify store is for AI shopping agents, accounting for platform specific defaults such as theme rendering, app generated content, and Shopify's own llms.txt and Agentic Dashboard behavior.

Shopify stores share a common technical foundation, so common failure patterns repeat across thousands of stores, such as app injected content that AI cannot parse or default themes that hide key data behind JavaScript.

Two Shopify stores use the same theme. One customized its product template to expose full specs in plain HTML. The other left specs inside a JavaScript accordion. Shopify AI Readiness differs by 34 points between them.

Shopify AI Visibility measures whether a Shopify store's products, pages, and trust signals are visible enough for AI shopping systems to discover the store at all, before evaluation or recommendation ever happens.

Discovery comes before evaluation. A Shopify store can be perfectly ready and trustworthy once found, but if AI systems never discover it in the first place, none of that readiness gets the chance to matter.

A Shopify store blocks AI crawlers in its robots.txt by default, a common but often unnoticed theme setting. Correcting it increases the store's presence in AI search results within two weeks.

Unverified Visibility is a brand appearing in AI answers or mentions without the underlying trust signals to support that appearance, such as verifiable reviews, business details, or accurate policies.

Visibility built without trust signals is fragile. AI systems periodically re-evaluate trust, and a brand relying on Unverified Visibility can disappear from recommendations as quickly as it appeared.

A new brand gets mentioned by AI for a trending product category based on early buzz alone. Three months later, once AI systems re-check trust signals and find no verifiable reviews or return policy, mentions drop by 80 percent.