Marks and Spencer Lily AI Product Discovery 2026
M&S shows why AI visibility now starts with product data, not just ranking tactics.
Marks & Spencer's Lily AI partnership shows a simple shift: retail AI visibility is moving from page copy to product data. The brands that win in Google Shopping, AI Overviews, AI Mode, Gemini, and agentic commerce will make every SKU legible before a shopper ever types a query.
The Marks & Spencer AI product discovery signal is about structured product data
Marks & Spencer is treating product discovery as a data architecture problem, not a campaign problem. On July 9, 2026, Retail Gazette reported that M&S partnered with Lily AI to improve how products are found across Google, organic search, and emerging AI-powered shopping channels.
The operational detail matters more than the vendor name. The report says M&S is using Lily AI's Product Intelligence Platform to automate structured product data creation at scale, improve product attribution, and make new products discovery-ready at launch. That is the retail version of AI visibility: the brand is not only writing better descriptions; it is giving discovery systems cleaner inputs.
Google has been moving in the same direction. In May 2026, Google announced AI Performance Insights in Merchant Center to show how products are discovered on AI Mode in Search, the Gemini app, and AI Overviews. For retailers, that makes the product feed a visibility surface.
M&S's timing is not random. In its 2026 annual report hub, the retailer named Digital, Data and Technology as a key investment priority and said each managing director now has a digital and technology plan embedded into the business. Product discovery is becoming part of operating infrastructure, not a marketing side project.
AI shopping visibility depends on attributes machines can resolve
AI shopping systems cannot recommend a product they cannot understand. Lily AI's own product intelligence materials frame the problem bluntly: AI surfaces need richer product attributes, schema, feed data, and language that maps to how shoppers ask for things.
The practical failure is easy to see. A human merchandiser may understand that a dress is "office-ready," "travel-friendly," "modest," or "wedding guest appropriate." A product feed may only say "green midi dress." Search engines, AI shopping agents, and recommendation systems then have to infer the missing buying context. Sometimes they will. Often they will not.
Lily AI's June 2025 product content announcement argued that product pages need technically robust, consumer-informed content for both traditional and AI-driven search. It cited its consumer research finding that roughly 80% of consumers had abandoned a search because they could not find what they wanted, while 40% reported using an AI-powered search engine for online shopping. The exact numbers are less important than the pattern: shoppers are asking in natural language, and product catalogs are still too often built in retailer language.
That creates a new operating question for retail teams:
| Visibility layer | Old retail habit | AI-era requirement |
|---|---|---|
| Product title | Short merchant label | Searchable, specific product identity |
| Attributes | Internal taxonomy fields | Consumer-language traits machines can parse |
| Feed data | Channel distribution input | Source material for AI shopping surfaces |
| Product page copy | Human browsing support | Extractable evidence for search, agents, and recommendations |
| Measurement | Clicks and revenue by channel | Visibility across AI Mode, AI Overviews, Gemini, and shopping agents |
The M&S case is useful because it does not treat AI visibility as a content hack. It treats it as product truth becoming machine-readable.
The retail lesson is bigger than SEO or GEO
Marks & Spencer is showing why retail AI visibility now sits below SEO, not above it. SEO can help a page rank. GEO can help a page get cited. Neither solves a product catalog that lacks the attributes, schema, and semantic specificity machines need to match a shopper's intent.
Google's 2026 Search update language reinforces the point. Google said Search is becoming more conversational, multimodal, and agentic, with follow-up questions flowing from AI Overviews into AI Mode. That means product discovery is no longer just "rank for blue linen shirt." It is "find me breathable office shirts that do not wrinkle on a commute," followed by comparison, refinement, availability, and purchase intent.
In that environment, the product itself has to carry more evidence. The PDP, feed, structured data, image attributes, and availability signals all become part of the answer system. If those inputs are thin, the brand can be present in inventory and still invisible in interpretation.
This is where the Machine Relations frame is useful as a neutral diagnostic lens. Machine Relations, coined by Jaxon Parrott, describes the broader discipline of making brands legible, retrievable, and credible inside machine-mediated discovery. In retail, that discipline begins at SKU resolution: can the machine identify what this product is, who it is for, why it fits the query, and whether it can trust the data?
Product data is becoming earned visibility infrastructure
The strongest retail brands will connect product intelligence to external authority, not isolate it inside the catalog. Better attributes improve machine interpretation, but attributes alone do not create trust. AI systems still need corroboration: customer reviews, third-party coverage, product comparisons, retailer authority, and consistent entity signals across the web.
That is why earned authority still matters. A well-structured M&S product feed helps machines parse the product. External retail coverage, brand reputation, and trustworthy category references help machines decide whether the brand deserves to be surfaced.
AuthorityTech's publication intelligence tracks the publications AI engines retrieve and cite across categories, which matters because retail visibility increasingly depends on both owned product data and third-party corroboration. The product feed tells machines what the item is. External authority helps machines decide whether to trust and recommend the brand.
For operators, the sharper question is not "Do we need AI search optimization?" It is:
- Can a machine understand every product without guessing?
- Can a search or shopping agent map those products to natural-language buying intent?
- Can external sources corroborate the brand and category claims?
- Can the team measure visibility across AI shopping surfaces, not only classic search?
That is the citation architecture problem underneath the M&S signal. Retailers need machine-readable products and machine-trustworthy proof around those products.
What CMOs should copy from the M&S case study
The copyable move is to audit product legibility before chasing AI visibility dashboards. Visibility tools can show whether a brand appears. They cannot fix weak source material by themselves.
A practical retail audit should start with the top commercial product clusters, not the whole catalog. Pick the products that matter most to margin, seasonality, or inventory exposure. For each cluster, review whether product data answers the questions shoppers actually ask: fit, occasion, material, compatibility, durability, use case, constraints, comparisons, and availability.
Then match those fields against the surfaces that now mediate discovery:
| Audit question | Why it matters for AI visibility |
|---|---|
| Do product titles contain the searchable identity a shopper would use? | AI systems need clear entity labels before they can match intent. |
| Do attributes describe use case, occasion, material, fit, and constraints? | Natural-language shopping queries are more specific than legacy category filters. |
| Does schema expose the same product truth as the page and feed? | Inconsistency makes machines less confident. |
| Do reviews and third-party mentions reinforce the same claims? | Corroboration improves trust beyond owned content. |
| Can the team measure AI surface visibility by product group? | Classic SEO dashboards miss where AI shopping discovery is moving. |
The M&S story is not that every retailer needs the same vendor. The story is that the catalog is no longer back-office plumbing. It is the source layer for AI-mediated shopping.
FAQ
Why is the Marks & Spencer Lily AI partnership important for AI visibility?
The partnership is important because M&S is improving product discovery by enriching structured product data, not just writing more content. That matches Google's move toward AI performance insights in Merchant Center, where product data affects visibility across AI Mode, AI Overviews, and Gemini.
Is product data now part of GEO?
Yes, for retail. GEO is often discussed as content formatting for generative engines, but retail discovery depends on product feeds, attributes, schema, and PDP content. A product cannot be cited, compared, or recommended reliably if the machine cannot resolve what it is.
What should a CMO audit first after seeing the M&S case study?
Start with the product clusters that carry the most commercial value. Audit whether titles, attributes, schema, page copy, reviews, and third-party mentions all describe the same product reality in machine-readable language. If the inputs are inconsistent, AI visibility measurement will only diagnose the gap.
Where does Machine Relations fit in retail AI discovery?
Machine Relations is the broader framework for making brands legible, retrievable, and credible inside AI-driven discovery. In retail, the first layer is product legibility; the next layer is corroboration from trusted sources. Teams can benchmark that gap with an independent AI visibility audit.