Introduced at The Data Graph Convention, Cornell Tech, New York
As a information strategist who works within the Search trade, I spend my days the place information meets which means: remodeling the messy, unstructured world of product feeds and databases into structured information that each people and machines can use. At this yr’s Data Graph Convention at Cornell Tech, I launched a brand new metaphor for this transformative work: Semantic Alchemy.
Like the traditional alchemists who sought to show lead into gold, we flip information into information: helpful, actionable, AI-ready information. The method will not be magic; it’s structured, strategic, and measurable, and it’s more and more important in a digital financial system the place success hinges on discoverability and personalization.
Let me stroll you thru this alchemical course of and share how WordLift applies it to real-world digital methods.
What’s Semantic Alchemy?
Semantic Alchemy is the artwork of remodeling uncooked information into refined information graphs and ontologies – the important gas for AI techniques and clever net experiences.
It’s the strategy of taking uncooked, inconsistent product or content material information—information that usually resides in spreadsheets, legacy techniques, and product catalogs—and remodeling it into structured information belongings. These belongings energy language fashions, semantic search, product suggestions, and content material technology at scale. In doing so, we assist our purchasers automate and optimize their digital presence.
Let’s discover how this works, each conceptually and virtually.
The 5 Phases of Semantic Alchemy – that includes an actual use case implementation from Luxottica’s model Costa del Mar
1. Preparation – From Chaos to Readability
Uncooked product information is chaotic: incomplete, messy, redundant. On this first part, we cleanse, normalize, and mannequin the information. We establish product variants, group associated options, and start constructing an preliminary schema.
With Costa del Mar, we started by ingesting attribute information from their Product Data Administration (PIM) system. These attributes, resembling lens shade, body form, and match kind, had been inconsistent and siloed. We reorganized and standardized them, making ready a cleaner, extra logical information basis for additional semantic enrichment.
2. Heating – Constructing Important Semantics
As soon as cleaned, we start structuring this information into information graphs and ontologies: the true engines of semantic website positioning and AI integration.
We’ve developed a domain-specific ontology for eyewear tailor-made to Luxottica, enabling the coaching of a domain-tuned giant language mannequin able to producing high-quality, expert-level content material throughout their total catalog.
Each product variant will get its personal mini information graph. We outline the product’s attributes (e.g. lens materials, body form, UV safety stage) and their relationships. This formal semantic construction allows machines to know and cause in regards to the product in the identical method a human professional would.
Bridging the Context Hole – Why Semantics Matter Extra Than Ever
Semantic website positioning is evolving quick within the age of LLMs and autonomous brokers.
Let’s be clear:
There isn’t a scarcity of knowledge. What’s lacking is context.
We stay in a time of ample information, however scarce which means. Data graphs and ontologies present the contextual scaffolding that turns uncooked information into intelligence. They perform because the “grammar” of AI: enabling fashions not solely to foretell however to know. And this shift is essential as we transfer into an period the place LLMs are not passive instruments, however lively brokers in digital technique.
When paired with information buildings:
LLMs acquire persistent reminiscence and may cause towards objectives.
Responses shift from “believable” to verifiable and contextualized.
AI turns into explainable, traceable, and prepared for enterprise-level deployment.
3. Vaporization – Turning Data into Content material
That is the place structured information turns into wealthy, dynamic content material.
From LLMs educated on information, we are able to generate completely different content material varieties:
Extremely personalised product descriptions
Lengthy-tail website positioning content material concentrating on particular person intents
Q&A modules for content material clusters that match how customers search
These content material belongings replicate the actual person intent and are drawn from the semantic spine established in earlier phases. Each bit of textual content will be seen as a node in a bigger information ecosystem, enabling not solely higher engagement but in addition future LLM interoperability.
We’re not producing “simply any content material.” We’re creating distinctive, helpful, and context-rich content material, the type that solutions actual person questions and drives deep engagement.
4. Condensation – Formatting for People and Machines
Data has been generated and now it must be deployed.
On this part, we format content material for each machine readability and human usability:
Product feeds optimized for Google Service provider Heart (GMC)
Embedded schema.org microdata for serps
Enhanced formatting for multi-channel distribution
For Costa del Mar, we utilized beforehand unused product attributes from their PIM, attributes that had by no means been exhibited to clients or brokers, and mapped them to Google Service provider Heart “product element” discipline. By semantically structuring and publishing this information, we activated new filters in Google Procuring, enhancing product discoverability and navigation. Semantic markup was added by way of schema.org to enhance machine readability and search engine visibility.
5. Assortment – Measuring What Issues
Now we are able to accumulate the gold!
For Costa del Mar, the outcomes had been clear: enriched product listings, powered by information graphs and fueled by underutilized PIM attributes, noticed measurable uplifts in natural visitors and conversions. Each paid and free listings carried out higher, validating the end-to-end affect of semantic alchemy.
This isn’t simply idea, it’s ROI. Structured, semantic information drives higher efficiency, and in an AI-dominated panorama, information readability is enterprise foreign money.
The Future Is Information-Primarily based: website positioning within the Age of Brokers
The way forward for digital presence is more and more data-driven. As AI brokers start to interchange conventional search interfaces, the power to construction and contextualize data turns into paramount.
A strong rising normal supporting this shift is the Mannequin Context Protocol (MCP), which allows generative fashions to obtain and reuse structured, interoperable context. As an alternative of treating every immediate like a clean slate, MCP presents LLMs a semantic reminiscence house.
This modifications the website positioning paradigm: moderately than optimizing content material solely for serps, we now construct information buildings for agentic interpretation. Your digital belongings are not simply “findable”: they have to be comprehensible, interpretable, and actionable.
website positioning Past Search Engines And Towards the Age of AI Brokers
We’re transferring in direction of a future when AI brokers will turn out to be the first digital touchpoints, as a substitute of internet sites. These brokers don’t “learn” net pages; they interpret semantic buildings.
That raises an important query: Are your information ready to signify you? If conversational interfaces change search engine end result pages, visibility alone received’t win however readability and construction will. Success would require well-modeled information, interlinked ideas, logically queryable buildings. Conventional website positioning is evolving from discoverability to comprehensibility.
This issues now: the worldwide marketplace for conversational AI is projected to develop from $14.6 billion in 2025 to over $23 billion by 2027, pushed by agentic AI techniques. These brokers (from voice assistants to suggestion engines) depend on context, and content material, in flip, depends upon semantics. We have already got the information; what’s lacking is the structured which means that enables AI to behave intelligently. Data graphs and ontologies can bridge that hole. It’s about having the ability to keep on prime of this Clever Search.
Instructing AI to Assume Like a Human
Semantic alchemy is about extra than simply expertise. It’s about alignment between information and intent, between machines and folks, between content material and context.
By turning information into which means and which means into motion, we create extra revolutionary, extra related digital ecosystems.
In a world more and more formed by AI, the power to be understood is the brand new visibility.
And that, within the age of brokers, is the actual gold.
