From SEO to AIO: Exploring the Next Frontier of Ecommerce Optimization

Today, I stumbled across a LinkedIn post by Nick Weisser, the organizer of the WordPress Zurich meetup, about the relevance of SEO. The post made me think about how AI tools like ChatGPT are changing product discovery. As LLMs become a major player in how people engage with the internet, they will also change the way we find products. Shouldn’t we optimize our stores specifically for AI understanding?

AI Transforms Product Discovery and Purchase Decisions

Through my research, I explored the field of AIO (Artificial Intelligence Optimization). AIO is not just another marketing buzzword. It acknowledges that the purchase decision process and how products are discovered are fundamentally changing.

Davenport et al. (2020) examine how AI transforms consumer interactions with products, emphasizing changes in four key areas: predicting customer behavior, enhancing personalization, improving customer service, and facilitating product discovery. Their research reveals a standout insight: AI-driven discovery tools significantly reduce search friction, aligning products more closely with consumer preferences and making the shopping experience more efficient.

One of the underlying reasons shoppers are excited about this help from AI tools is that, according to Google research published in March 2024, 62% of consumers find decision-making increasingly challenging. Two-thirds postpone or forgo purchases due to an overload of choices and information.

How AI Reads Ecommerce Product Data

One insightful aspect of my deep dive into Artificial Intelligence Optimization (AIO) was learning about the Transformer architecture, which underpins advanced models like GPT-4, the underlying engine for tools such as ChatGPT. This architecture uses a mechanism called “self-attention,” introduced by Vaswani et al. (2017), that allows the model to determine the relative importance of each token (word or symbol) in a sentence. Self-attention enables Transformers to weigh connections between distant words and concepts, building a contextually rich understanding of the data. Unlike traditional models, which process text sequentially, Transformers analyze all tokens in parallel, applying self-attention layers to capture subtle and complex interrelations within the text. This capability is crucial in applications like product catalogs, where an AI model can recognize connections not just between products and their attributes but also across categories, subcategories, and even customer preferences, constructing a sophisticated network of relationships.

To optimize AI models for handling commerce product data, structured data formats can play an essential role. For instance, Ning et al. (2023) demonstrate how knowledge graphs can make product information more “AI-readable.” By implementing structured product information within a three-layer knowledge graph (with layers for entities, schema, and data relationships), they achieved notable results: a 76.3% precision rate in data integration and a 43.1% improvement in search accuracy. Structured data like this can help AI systems integrate and interpret the relationships among diverse product attributes better.

Ning et al. (2023) propose a three-layer knowledge graph structure tailored for ecommerce:

• An entity layer (basic attributes) captures fundamental product information, including details like name, brand, and specifications.
• The schema layer (extended attributes) defines relationships and hierarchies among different product categories and attributes, establishing a structured framework for organizing data.
• The data layer (relational attributes) integrates actual product data from various sources, ensuring that information remains comprehensive and up-to-date.

Here’s an example of how such a three-layer knowledge graph of product data might look:

{
"product" : {
  "basic_attributes" : {
    "id" : "GH-X1",
    "name" : "Gaming Headset X1",
    "price" : 99.99,
    "currency" : "USD",
    "sku" : "GH-X1-BLK",
    "availability" : "In Stock",
    "release_date" : "2023-08-15"
  },
  "extended_attributes" : {
        "specifications" : {
          "connectivity" : "Wireless",
          "battery_life" : "30 hours",
          "sound" : "7.1 Surround",
          "microphone" : "Detachable",
          "weight" : "250g",
          "color" : "Black"
        },
        "features" :
            [ "Noise cancellation", "Memory foam cushions", "RGB lighting" ],
        "warranty" : {"period" : "2 years", "type" : "Manufacturer"}
      },
   "relational_attributes" : {
    "category": ["Gaming", "Audio", "Accessories"],
	      "brand": {
	        "name": "Example brand",
	        "website": "https://example.com"
	      },
	      "related_products": [
	        {
	          "id": "GH-X2",
	          "relationship": "upgrade"
	        },
	        {
	          "id": "WM-1",
	          "relationship": "compatible"
	        }
	      ],
	      "reviews": [
	        {
	          "review_id": "R1",
	          "rating": 4.5,
	          "content": "Great sound quality and battery life.",
	          "date": "2023-10-01",
	          "reviewer": "User123"
	        },
	        {
	          "review_id": "R2",
	          "rating": 4.0,
	          "content": "Very comfortable for long gaming sessions.",
	          "date": "2023-09-15",
	          "reviewer": "Gamer42"
	        }
	      ]
  }
}
}

In theory, structured data provides a foundation for AI crawlers to identify and organize information more efficiently. AI crawlers rely heavily on consistent, structured data to “understand” the content they’re indexing. Standards like Schema.org, a collaborative vocabulary developed by major search engines, have become central to this process. Schema.org provides a standardized markup vocabulary to embed structured metadata within HTML. For example, the Product schema in Schema.org includes attributes for product name, price, category, and more, which can help search engines and AI models recognize essential product details.

While structured data like Schema.org improves search engine comprehension, AI crawlers’ current capabilities to fully understand complex relational data (as seen in the three-layered knowledge graph above) are still evolving. Traditional crawlers primarily recognize Schema.org’s straightforward, attribute-based structure but have limited ability to interpret deeper relationships like “compatible with” or “similar to.” Advanced AI models integrated into newer crawlers are beginning to bridge this gap, enabling a more nuanced understanding of structured data.

Product Descriptions That Make Sense

As AI models interpret data in layers of abstraction, the role of product descriptions has become more crucial than ever, especially given the limitations of current structured data for Artificial Intelligence Optimization (AIO). AI systems process information differently from traditional search engines. Therefore, we need a more nuanced approach to content creation.

What this means is that a superficial product description is not effective when optimizing for AI:

❌ “Premium gaming headset with awesome features!”

Instead, a more comprehensive and detailed description can help AI understand the full context and relevance of a product:

✅ “Wireless gaming headset featuring 7.1 surround sound, 30-hour battery life, and memory foam ear cups. Compatible with PC, PS5, and Xbox Series X.”

This need for detail isn’t new; superficial descriptions have long been ineffective in SEO and in building relevance for shoppers. What has changed is the increasing necessity to make product information rich and structured enough for both shoppers and AI models to interpret fully.

The challenge lies in striking the perfect balance: creating descriptions that communicate essential product details to shoppers, persuade them to take action, and provide AI with the context for deep understanding. Based on the research, here’s a framework for creating effective product descriptions:

What This Means For Your Store

AI is changing how people shop online. From my research into Artificial Intelligence Optimization (AIO), I’ve found that success depends on two key factors: making your product information clear for AI to process and ensuring it’s helpful for shoppers to navigate.

The numbers tell an interesting story. Google’s research shows that 62% of shoppers feel overwhelmed by options when shopping online. Effective AIO implementation shines in this area, making your products more discoverable and easily understood by both AI tools and shoppers.

Here’s what I’ve found works to future-proof your e-commerce store:

• Write detailed product descriptions  that provide AI systems with clear context. Tools like ChatGPT can help you create consistent, detailed product content that serves both AI and user needs.
• Structure your product data using knowledge graphs, starting with Schema.org markup on product pages, to ensure key product information is accessible and machine-readable.

Quick Start: Take one low-performing product, ideally with high traffic but a low conversion rate, and upgrade its description using the prompt template I shared. After 30 days, check the results in your analytics. This simple test could reveal significant improvements that make it worthwhile to apply the approach across your store.

This isn’t about “gaming” the system; it’s about enhancing your products’ discoverability and clarity. Whether customers find your product through ChatGPT, Google, or any other platform, good product representation follows the same core principles.

---

References

• Davenport, T., Guha, A., Grewal, D., & Bressgott, T. (2020). How artificial intelligence will change the future of marketing. Journal of the Academy of Marketing Science, 48(1), 24-42. https://link.springer.com/article/10.1007/s11747-019-00696-0
• Ashley, S., Bradley, S., & Park, Y. (2024, May). Build confident consumers to improve metrics. Think with Google. https://www.thinkwithgoogle.com/intl/en-emea/consumer-insights/consumer-journey/build-confident-consumers-to-improve-metrics/
• Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., Polosukhin, I. (2017). Attention is All You Need. In Advances in Neural Information Processing Systems (pp. 5998-6008).
• Ning, H., Yang, J., Chen, W., Sun, H., Liu, Z. (2023). Web Data Integration and Knowledge Graph Construction for E-commerce Product Information Management. IEEE Access, 11, 74100-74115. https://ieeexplore.ieee.org/document/10167798