Is your online store ready for AI commerce? What a readiness audit should check
HUB LLC · 12 September 2026 · Commerce
Online shopping used to start with a few words in a search box, a category page and a row of browser tabs. More often now, it starts with a complete question to an assistant: a waterproof jacket under €150, in XL, that can reach Riga by Friday. The answer depends on whether the software handling that question can find, interpret and trust your store’s information.
That makes an old question more pressing. Is the information a customer needs to choose and buy a product actually present, consistent and technically reachable? This article sets out what an AI eCommerce readiness audit should check, how to read its results and where an automated scan stops being useful.
AI commerce readiness is not a new SEO trick
It is tempting to treat AI discovery as a separate channel with its own secret optimizations. Google’s documentation points the other way: it states that there are no additional requirements or special optimizations for appearing in AI Overviews or AI Mode, and that no AI text files or special schema.org markup are needed for those features. Source: Google Search Central, AI features and your website.
At the same time, commerce platforms are building machine-readable interfaces specifically for agents. Shopify serves an agent discovery document at /agents.md on storefronts, with the same content at /llms.txt and /llms-full.txt by default, pointing agents to the store’s commerce endpoints and policies. Source: Shopify developer documentation. In January 2026 Google introduced the Universal Commerce Protocol, an open standard for agents working across consumer surfaces, businesses and payment providers. Source: Google’s announcement.
Both things are true. Conventional search foundations remain the base, and new commerce interfaces are being added on top. What connects them is the quality of the underlying information, so a useful audit measures a store’s data, structure and access rather than hunting for a single trick.
Two audiences, one source of truth
An online store now serves people and software acting on behalf of people. Both need the same things: accurate prices and stock, clear product attributes, fast and accessible pages, visible delivery and returns terms, and a merchant identity that can be checked. When that information is fragmented or contradictory, a human shopper may work around the gap. An automated system is more likely to skip the product or describe it incorrectly.
The eight areas a readiness audit should cover
1. Crawl and index foundation
Robots directives, XML sitemap availability, canonical tags, redirects, status codes, pagination, duplicate URLs, language versions and basic indexability. A store with indexing problems in conventional search is unlikely to do better in newer discovery environments. Review robots.txt rules for the specific crawlers you want to allow or refuse: OpenAI, for example, documents separate user agents for search, model training and user-initiated fetching, so a blanket block can have consequences a store owner did not intend. Source: OpenAI crawler overview.
2. Information software can read
Is essential product, delivery, returns, warranty and company information present in the delivered HTML or another reliably consumable form? If the price appears only after several client-side requests, or shipping terms exist only in an image or a pop-up, many systems will not see them.
3. Product data quality
Titles, descriptions, brand, identifiers such as SKU, GTIN and MPN, prices, availability, variants, category assignment and the attributes people use to choose: size, material, dimensions, compatibility, noise level, energy class. Typical problems are missing attributes, inconsistent naming, duplicated manufacturer descriptions and specifications that exist only inside PDF data sheets.
4. Structured commerce data
Product, Offer, BreadcrumbList and Organization markup where appropriate, checked for validity and, more importantly, for agreement with the visible page. Google’s guidelines require structured data to be a true representation of the page content and not to describe content readers cannot see. Source: Google structured data general guidelines. Contradictions in price, currency, availability or variant data between the page and its markup are among the most fixable findings, because one template often produces thousands of pages.
5. Performance and user experience
Core Web Vitals, mobile usability, page weight, image optimization, render-blocking scripts and obvious conversion friction. These matter to people first. A page that is slow or depends on heavy scripts to show its content is also a less dependable source for automated fetching.
6. Content that supports a decision
Does category and product copy answer the questions buyers actually ask about use cases, compatibility, comparisons, sizing and care, or does it repeat keywords? Specific FAQs, buying guides and comparison information help people decide and give AI systems concrete passages to work with. Thin or duplicated copy does neither.
7. Trust and merchant policies
Company name and address, contact options, delivery areas and times, returns and warranty terms, payment options and a consistent merchant identity across the site, feeds and business profiles. Before a person or an assistant recommends a merchant, these are the details that make the recommendation defensible.
8. Feeds, APIs and international data
Product feeds, catalog APIs, clean identifiers and reliable synchronization between storefront, PIM, ERP and inventory. Google notes that providing both structured data and a Merchant Center feed maximizes eligibility for its shopping experiences. Source: Google product structured data documentation. Merchant Center also accepts optional conversational attributes, such as product questions and answers and related products, to help AI-driven surfaces like AI Mode understand products. Source: Google Merchant Center Help. For stores selling in several countries, include hreflang, currency handling and consistency of translated product information.
Checks you can run on your own store this week
- Compare three product pages with their markup. Do price, currency, availability, brand and SKU match what a customer sees, including for a variant that is out of stock?
- Reload a product page with JavaScript disabled. Are the product name, price, key specifications and delivery information still present?
- Read robots.txt line by line. Confirm every rule is intentional, including crawler rules added by a plugin, a CDN setting or a previous agency.
- Choose one bestseller and list the five facts a buyer needs to choose it. Check whether each exists as an attribute, not only inside a description or PDF.
- Start on a product page and find the delivery and returns terms. Then check that they match the terms in your feeds and marketplace listings.
- Look for agent discovery files. On Shopify, request
/agents.mdand review what it says. Elsewhere, note whether such a file exists and what it claims, without treating it as a substitute for correct pages and feeds.
How to read a readiness score
A score is useful only if it can be explained. A single number with no evidence behind it invites either complacency or panic. A credible report breaks the result into categories and ties every finding to a page, a template and a reproducible observation.
The table below illustrates such a breakdown. It is an example, not the result for any real store.
| Category | Example score | What it reflects |
|---|---|---|
| Overall readiness | 63 / 100 | Weighted summary of the categories below |
| Technical SEO | 82 | Indexability, canonicals, sitemap, redirects |
| Performance | 71 | Page weight, images, scripts, mobile behavior |
| Structured commerce data | 68 | Markup present, valid and consistent with the page |
| Product data | 61 | Attribute completeness and consistency |
| Content quality | 59 | Specific, decision-supporting, non-duplicated copy |
| AI discoverability | 54 | Whether key commercial information is reachable and readable |
| Agentic commerce readiness | 38 | Feeds, APIs, identifiers and policy data |
This example describes a store where SEO work has addressed crawlability but the catalog was never modeled for comparison: a solid technical base held back by thin attributes and weak integration data.
Scoring should be deterministic wherever possible. HTTP status, markup presence, canonical values and sitemap availability are facts to validate directly, not to be judged by a language model. AI is useful for the interpretive part, such as classifying page types, spotting generic copy and explaining a finding in plain language, and those conclusions should be labeled as interpretation.
What priority findings look like
Each finding should state the problem, the evidence, why it matters and what fixing it involves. Three examples of the kind of finding a readiness audit produces:
High priority: product attributes are incomplete
Brand, identifiers, material, size or compatibility appear visually on some pages but are not consistently represented in structured product data. A comparison or recommendation system cannot use information it cannot reliably extract. At catalog scale, the fix is an enrichment workflow that extracts, normalizes, validates and publishes missing attributes, not manual page edits.
High priority: markup contradicts the commercial page
A template outputs the parent product’s price in the markup while the customer sees a variant price, or availability is hard-coded as in stock. The report should name the responsible templates, because fixing one template can correct every page it renders.
Medium priority: copy is written for keywords, not decisions
Category text repeats generic phrases while omitting use cases, compatibility and purchasing criteria. Restructuring copy around those questions, while preserving brand voice, helps conventional search, onsite conversion and AI-assisted discovery at the same time.
What an automated scan cannot tell you
- Back-office causes. A public scan sees symptoms. Why an attribute is missing, whether an ERP mapping, a supplier file or an import script, needs admin, feed or code access.
- Catalog-wide truth. A sample of pages reveals patterns. Measuring completeness across a large catalog needs feed, API or database-level checks.
- Business impact. Prioritizing by revenue requires analytics and knowledge of which categories matter to the business.
- Guaranteed outcomes. No legitimate provider can guarantee that an independent AI system will recommend or cite a store. An audit improves the foundations that make a store easier to understand, evaluate and integrate.
Platform notes
Magento and Adobe Commerce
Findings should be traced to the responsible part of the stack: attribute sets, configurable product relationships, store views, custom modules, third-party extensions, feed generators, cron jobs and indexers. A markup contradiction can come from a theme or an extension rather than from catalog data, and the audit should say which. For store-level search guidance, see our Magento AI optimization tips.
Shopify
Review theme output, app-injected markup and scripts, variant and collection structure, international content and the platform-managed agent discovery file. Shopify states that for most stores its managed agents.md file is all that is needed, so customizing it is rarely the first priority. Catalog quality remains the merchant’s responsibility.
Custom platforms and marketplaces
For custom PHP, headless or marketplace systems, the public storefront is only part of the picture. The data model, catalog APIs, search architecture and seller data validation often determine readiness more than any template does.
From findings to implementation
An audit is worth commissioning only if its findings become work: templates to fix, attributes to enrich, feeds to repair and integrations to stabilize. We describe how we structure that implementation in AI Commerce Ready: preparing Magento and Shopify stores for AI-driven shopping, and how to measure whether the changes affect visibility in AI visibility monitoring.
If you would like an engineer to look at your own store, start with our free eCommerce store audit or send us your store URL and the platform it runs on. We will tell you which findings are worth acting on first and what implementing them involves. Our AI Commerce Ready service covers the implementation, and the AI for eCommerce page describes our wider commerce work.