Generative Engine Optimisation
AI agents cannot recommend what they cannot interpret. Your product data is now operational infrastructure.
The Discovery Revolution
Product discovery has been disrupted three times. First by search engines, which made every product findable through keywords. Then by marketplaces, which made every product comparable within a single interface. Then by social media, which made every product recommendable through human networks.
The fourth disruption is AI-native discovery. And it is fundamentally different from the first three. In every previous model, a human made the final discovery decision. They typed the query. They scrolled the results. They clicked the link. They read the reviews. AI-native discovery removes the human from the discovery loop entirely. The agent receives an intent, interprets it, searches across multiple sources, evaluates options against constraints, and presents a recommendation.
The consumer never sees your website. They never read your product description. They never compare your reviews against a competitor's. The agent does all of that, and it does it by reading structured data, not marketing copy.
The Zero-Click Reality
58 percent of searches now end without a click, according to industry research. The user gets their answer directly from the AI. For commerce, this means the traditional funnel (search, click, browse, compare, buy) is collapsing into a single prompt. If your product is not in the agent's recommendation, you do not get a second chance. There is no page two of results. There is no scroll. There is the recommendation, or there is nothing.
From SEO to GEO
Search Engine Optimisation (SEO) was designed to make content rank highly in search engine results pages. It optimises for keywords, backlinks, page speed, and user engagement signals. Generative Engine Optimisation (GEO) is designed to make content cited, referenced, or recommended by AI models. It optimises for structured data, entity recognition, answer-readiness, and machine interpretability.
What AI Agents Actually Read
When an AI agent evaluates a product, it processes:
- Structured product attributes: dimensions, materials, specifications, compatibility, certifications. Machine-readable, not embedded in prose.
- Availability signals: real-time stock status, delivery windows, geographic availability. The agent will not recommend a product it cannot confirm is available.
- Price data: current price, historical pricing, promotional offers, comparison pricing. Agents optimise for the consumer's budget constraint.
- Review aggregation: overall rating, review volume, sentiment analysis, common complaints, strengths. The agent synthesises reviews rather than reading individual ones.
- Trust signals: return policy, warranty, seller reputation, certification marks. The agent evaluates risk on behalf of the consumer.
- Schema markup: structured data that tells the agent exactly what each attribute means and how to compare it across products.
Notice what is not on this list: your brand story, your lifestyle photography, your emotional marketing copy. These matter to humans. They are invisible to agents.
The E-E-A-T Framework for Agents
Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) was designed for human search. In the agent context, the same principles apply but with a machine-readable twist.
Experience: Verified customer reviews, usage data, return rates. Not claims, evidence.
Expertise: Technical specifications, certifications, compliance documentation. Verifiable attributes.
Authoritativeness: Domain authority scores, citation by other trusted sources, industry recognition.
Trustworthiness: Consistent data across platforms, accurate availability, transparent pricing, clear return policies.
Making Products Buyable by AI
SAP's research on agentic commerce coined a useful phrase: making products "buyable by AI." This means transforming product data from marketing material into operational infrastructure.
Structured Data Requirements
At minimum, every product in your catalogue needs:
- Schema.org Product markup: name, description, SKU, brand, price, availability, review aggregate. This is the baseline that AI agents expect.
- Rich product attributes: every specification that a consumer might use to make a decision. Size, weight, colour, material, compatibility, power consumption, certifications.
- Real-time availability: not just "in stock" but specific delivery windows, geographic availability, and fulfilment options.
- Comparison-ready data: attributes formatted consistently so agents can compare your product against competitors on the same dimensions.
- Machine-readable policies: return window, warranty terms, shipping costs, promotional conditions. Structured, not buried in legal text.
The Catalogue Optimisation Process
SAP reports that their Catalogue Optimisation Agent can scale to catalogues with more than 10 million items, helping teams improve content 70 percent faster, increase data completeness by 5 percent, and reduce maintenance effort by 63 percent.
The optimisation framework follows a repeatable cycle:
Audit: Score every product on data completeness. How many of the required attributes are present, accurate, and structured?
Clean: Remove inconsistencies, fill gaps, standardise formats. A product listed as "blue" on your website and "navy" in your feed confuses agents.
Enrich: Add missing attributes using product specifications, supplier data, and customer feedback. Every gap is a missed recommendation.
Localise: Support multilingual content if you serve multiple markets. AI agents operate in the consumer's language.
Monitor: Product data is not a one-time project. Prices change, stock fluctuates, specifications update. Automated monitoring ensures your data stays agent-ready.
Owned vs. Third-Party Discovery
There are two ways AI agents find your products. You need a strategy for both.
Third-Party Agent Discovery
This is when an AI agent on a platform you do not control (ChatGPT, Google Shopping, Perplexity, a competitor's agent) discovers and recommends your product. You control none of the experience. You control only the data the agent receives.
Optimising for third-party discovery means:
- Structured product feeds submitted to every relevant platform and aggregator.
- Schema markup on every product page so crawling agents can extract attributes.
- Consistent data across all channels. Price, availability, and attributes must match everywhere.
- Review velocity and quality. More reviews, higher ratings, and recent reviews all increase the likelihood of agent recommendation.
Owned Conversational Commerce
This is when you deploy your own AI agents on your properties. The agent operates on your behalf, using your data, serving your customers. This is where durable competitive advantage is built.
When a consumer interacts with your agent, you capture:
- Intent data: what the consumer actually wants, in their own words, not filtered through a search algorithm.
- Decision context: why they chose product A over product B, what constraints mattered, what trade-offs they made.
- Preference signals: price sensitivity, brand loyalty, feature priorities, delivery requirements.
- Conversion patterns: what information the agent provided that led to purchase, what objections were raised.
This intelligence feeds merchandising, pricing, inventory, content strategy, and personalisation. It creates a feedback loop that gets stronger with every interaction. Third-party agents do not share this data with you.
The Dual Strategy: Optimise for third-party discovery to capture demand you would otherwise miss entirely. This is table stakes. Build owned agents to capture the intelligence that creates long-term advantage. This is the strategic play. Neither alone is sufficient. Together, they create a flywheel: third-party discovery drives awareness, owned agents drive intelligence, intelligence improves both discovery and conversion.
AI Engine Optimisation (AEO)
AEO is the technical discipline that sits alongside GEO. While GEO is the strategic framework, AEO is the execution: the specific technical practices that make your content citable by AI systems.
Answer-Ready Content
AI agents look for content that directly answers questions, not content that hints at answers. Structure your product information as explicit answers to the questions agents ask.
Not: "Our premium leather wallet comes in a variety of colours." Instead: "Available colours: black, brown, tan, burgundy. Material: full-grain Italian leather. Dimensions: 11cm x 9cm x 1.5cm."
Not: "Fast shipping available." Instead: "Standard delivery: 3-5 business days. Express delivery: next business day. Free shipping on orders over $50."
Technical Implementation
The technical checklist for AEO readiness:
- Schema.org structured data on every product page (Product, Offer, AggregateRating, Review).
- robots.txt configured to allow AI crawlers access (check for blocks on common AI user agents).
- XML sitemaps updated and submitted to Google Search Console and Bing Webmaster Tools.
- Page load speed optimised. AI crawlers have timeout limits. Slow pages get skipped.
- Content formatted for extraction. Clear headings, structured lists, explicit specifications, FAQ sections.
Practical GEO Audit
The following framework walks you through auditing a product catalogue for AI agent readiness.
Step 1: Select Your Sample
Choose 10 products from your catalogue that represent your highest-revenue or highest-priority items.
Step 2: Structured Data Audit
For each product, score its data completeness across six categories: structured attributes, availability signals, price data, review aggregation, trust signals, and schema markup. Use a 1-5 scale for each. Calculate the average.
Step 3: AI Agent Test
Ask ChatGPT, Perplexity, and Google's AI Overview about your products. Search for your product category and see if your products appear in recommendations. Search for your brand name. Note what information the AI surfaces and what it gets wrong.
Step 4: Competitor Comparison
Run the same AI agent test for your top three competitors. Compare their data completeness, AI visibility, and recommendation frequency against your own. Identify where competitors are ahead.
Step 5: Roadmap
Produce a prioritised list of improvements ranked by impact and effort. Quick wins (schema markup, consistent naming) first, then structural improvements (real-time availability feeds, review aggregation), then strategic investments (owned agent deployment).
Which gaps in your product data would make your offerings invisible to AI agents, and what would fixing them unlock?
Key Takeaways
- Discovery is changing: 58 percent of searches are zero-click. AI agents find, evaluate, and recommend products without the consumer ever visiting your site.
- GEO replaces SEO: Generative Engine Optimisation makes products citable and recommendable by AI. Structured data, entity authority, and answer-ready content are the new ranking factors.
- Product data is infrastructure: Structured attributes, real-time availability, schema markup, and machine-readable policies are not nice-to-haves. They are operational requirements.
- Dual strategy wins: Optimise for third-party agent platforms (awareness) and build owned conversational agents (intelligence). Neither alone is sufficient.
- AEO is execution: Answer-ready content, entity authority, schema markup, and technical accessibility are the specific practices that make your content visible to AI.
- Audit first: Start with a structured data audit. Quick wins in schema markup and data consistency can improve AI visibility before any strategic investment.