Your Website Isn't a Brochure Anymore.

It's a Knowledge Graph.
‍

(Or, The attention deficit of vapid spectacle)

Logo with the text 'Formant & Thread' in white italic font on a black background, accompanied by a bright green abstract infinity-shaped design to the right.
Steve Williams
11 Jan 2022
•
5 min read
Person with dark hair and large hoop earrings leaning back and looking upward while sitting alone in a darkened theatre or auditorium with cushioned seats.

‍

For years, we treated websites like digital brochures. Write some copy. Add a few photographs. Build a navigation. Publish. If it looked professional and ranked reasonably well in Google, the job was done. That way of thinking made sense when people were your only audience.

‍

Today, they aren't.

‍

Alongside every visitor arriving on your site is another audience quietly reading everything you publish: AI search engines, language models and knowledge graphs. They aren't interested in your design. They're trying to understand your business. That's an important distinction. A person can admire a beautiful website and still leave without understanding what you actually do. An AI model has the opposite problem. It can read every word on your site in seconds, but unless your information is structured properly, it may never understand how the pieces fit together.

‍

That's where the idea of a knowledge graph becomes useful.

‍

Instead of thinking about your website as a collection of pages, think of it as a network of connected ideas.

Your services connect to industries. Your team connects to projects. Your projects connect to outcomes.

Your articles connect to expertise. Every relationship adds another piece of evidence about who you are and why you're credible.

‍

Humans experience this as a well-organised website. Machines experience it as structured knowledge. This shift is already changing how search works. Traditional SEO focused heavily on matching keywords to search queries. Modern search increasingly tries to understand entities: real organisations, people, products and concepts (and the relationships between them.)

‍

That's why structured data has become so important.

‍

Schema.org isn't really an SEO trick. It's a shared language that helps machines distinguish between a founder and an employee, a case study and a blog post, a service and a product. It turns marketing copy into information that software can interpret with confidence. The businesses that embrace this aren't simply improving rankings. They're making themselves easier for AI systems to understand, reference and recommend. The practical benefits go well beyond search. When your content is built around connected entities instead of isolated pages, it becomes easier to reuse across your site. A case study can appear on service pages, industry pages and author profiles without creating duplicate content.

Internal navigation becomes more intelligent because pages are connected by meaning rather than just menu structures.

As your business grows, your website grows with it instead of becoming increasingly difficult to manage.

In many ways, this is less about technology than it is about clarity.

‍

A brochure tells people what your business looks like.

A knowledge graph explains what your business knows.

‍

As AI becomes another gateway between businesses and customers, that distinction matters more every day.

The websites that succeed over the next few years won't necessarily be the ones with the most animation or the loudest marketing. They'll be the ones that communicate their expertise clearly enough for both people and machines to understand.

‍

That's the future of the web.

And, increasingly, it's the present.

The Invisible Failure of the Modern Webflow Site

•
Logo with the text 'Formant & Thread' in white italic font on a black background, accompanied by a bright green abstract infinity-shaped design to the right.
Steve Williams
11 Jan 2022
•
5 min read
Old, weathered metal sign reading 'WRONG WAY' leaning slightly in a grassy field under dark, ominous cloudy sky.

‍

To most clients, a newly launched Webflow site looks like a triumph. It features smooth interactions, pixel-perfect typography, and responsive micro-animations.

‍

To AI search crawlers, LLM agents, and semantic web scrapers, however, it is often completely broken.The core failure of most Webflow builds stems from a critical blind spot: treating the web as a visual canvas rather than a machine-readable entity graph.

‍

Webflow’s visual-first interface encourages designers to style elements based on appearance rather than semantic architecture, leaving search engines to blindly guess how concepts, people, and services actually relate to one another.Here is where the typical Webflow site structurally breaks down beneath the surface.

Webflow’s CMS is remarkably flexible for flat content, but real-world domain knowledge is hierarchical and relational.

‍

Out of the box, Webflow lacks native support for deep parent-child taxonomies. Designers frequently resort to flat, single-level Reference or Multi-Reference fields.

‍

This flattens complex relationships into disconnected lists, failing to express explicit sub-categories, parent concepts, or nested entities.

‍

Without explicit relational nesting, crawlers cannot map the broader context of your knowledge domain.

Structured data (Schema.org) is the primary language machines use to parse your business logic.

‍

In Webflow, it is almost always an afterthought.

‍

Because Webflow offers no native Schema validation layer or built-in entity modeling UI, designers usually resort to one of two broken approaches: Static Copy-Pasting: Injecting static JSON-LD snippets into page heads, which quickly become outdated or misaligned across dynamic template pages.

‍

Fragile CMS Injection: Binding CMS fields manually inside Custom Code blocks. A single missing quotation mark, unescaped character, or broken variable silently invalidates the entire script, leaving page code broken for crawlers without throwing a visual error on the site.

A common byproduct of manual custom-code schema injection in Webflow is a disconnect between machine code and the visible Document Object Model (DOM).

‍

When a designer hardcodes or auto-generates JSON-LD that declares structured data—such as product reviews, FAQs, location metadata, or author credentials—that does not visually exist or match what is rendered on the page, search crawlers flag it as deceptive markup.Rather than gaining rich snippets, the site faces algorithmic demotion or entity misclassification because its machine layer contradicts its human layer.

A well-structured knowledge graph often relies on clear, hierarchical URL paths that reflect topical taxonomy (e.g. /services/web-design/ux-audit).Webflow imposes strict dynamic URL constraints: single-segment CMS collection paths (e.g. /blog/post-title or /services/service-title).

‍

This rigid structure makes deep, programmatic taxonomy scaling difficult. It prevents the URL architecture from echoing the semantic hierarchy built into the underlying graph, forcing designers to rely entirely on workaround scripts or secondary routing layers.

Webflow remains an extraordinary tool for visual and functional deployment, but visual design is only half the web.

‍

To build web platforms that survive and rank in an AI-driven index, designers must shift from visual styling to semantic engineering. A Webflow site shouldn't just look stunning—its CMS schemas, custom code injections, and entity relationships must be engineered as cleanly as its CSS flexboxes.

Why beautiful websites no longer guarantee visibility

Logo with the text 'Formant & Thread' in white italic font on a black background, accompanied by a bright green abstract infinity-shaped design to the right.
Steve Williams
11 Jan 2022
•
5 min read
A digital artwork featuring two faces side by side on a dark background, overlaid with translucent computer code; the left face is human-like with a calm expression, the right face is more abstract and eerie with wide eyes and dark accents.

‍

Most websites are still designed as if they will be looked at. This is the first mistake. The second is more expensive: they are designed as if they will be understood by the "pixel-curious." They're rarely either.What they actually become is something stranger. A kind of performative competence in a system that no longer rewards competence-as-display (face-maxxing: just a different silicon influence). Animations get smoother. Spacing more “considered.” Someone splutters "premium" without irony. And yet nothing changes. The site shrugs like an miscreant at the Austrian border. Nice smile, no papers.

‍

Something has shifted over the past year, though few clients have been told in a way they can metabolise. They glaze over when told search is no longer a list of options. That AI systems are the new border guard in town, node-splaining the gap between browser's intention and truth, compressing results into something closer to a didactic geography teacher worried over the latest OFSTED inspection.The result is most websites are no longer "visited;" but interpreted and extracted.

‍

Design used to function as a soft proxy for authority. If it looked expensive, it was assumed to be stable. Aesthetics shorthand for legibility and humans nodded sagely over its archimedian points. That condition has since coughed politely and left the room.Now, aesthetics have been demoted. Visual language is cheap, abundant, frictionless. Entire brands can be generated in an afternoon, iterated endlessly, deployed globally, in more flavours than Ben and Jerry. When everything looks competent, “competence” stops functioning as proof of anything, and authority resides in structures less visible.

‍

This is where the new work begins. Less concerned with surface expression than coherence. Call it narrative architecture. Call it semantic discipline, if you want to sound unnecessarily evangelical about it, preceding aesthetics, rather than garnishing it afterwards. Not the end of aesthetics, then, but the end of aesthetics-as-authority. A carefully art-directed website no longer guarantees an organisation will survive extraction into an AI-generated summary, not without a json knowledge graph somewhere downstream. Design still matters. It simply no longer gets the final word.The unsexy, infrastructural work wins neither awards, nor even awwwards. It does, however, determine whether you exist in the machine-readable reality.

‍

It's existential.

‍

This leads to an awkward implication. If an organisation cannot determine how it is represented outside its own domain, then it is, in a very real sense, no longer fully authored. Its identity distributed across platforms, crawlers, models, summaries, fragments of inference. Merely a rumour with a logo.This is why the shift from aesthetics to structure is not simply a design trend.It is a change in where authority lives.

‍

‍

‍

the cost of ambiguity

Logo with the text 'Formant & Thread' in white italic font on a black background, accompanied by a bright green abstract infinity-shaped design to the right.
Steve Williams
11 Jan 2022
•
5 min read

‍

Search engines and large language models do not know your business. They infer it. They assemble a plausible outline of what you do from whatever fragments of digital debris you happen to have left scattered across the web. When those fragments conflict—outdated service pages, inconsistent directory listings, ambiguous positioning—the machinery does not ask for clarification. It simply fills the silence with plausible fiction.

To understand why this happens, it helps to borrow from Roy Bhaskar’s critical realism and its stratified view of reality:

  • The Real: The underlying structures, capabilities, intellectual property, and relational powers your company actually possesses.
  • The Actual: The tangible manifestations those powers produce—your live pages, launched products, completed client projects, and operational footprint.
  • The Empirical: The tiny, visible surface area of those events that an external observer—in this case, an indexing crawler or a vector retriever—actually captures.

Commercial ambiguity is simply the structural gap between these three layers.

                               

Most digital strategies fall victim to what Bhaskar called the epistemic fallacy: the mistake of confusing what can be observed with what actually exists. Search engines and AI systems commit this fallacy by design. They operate strictly at the empirical level, treating the fragment they observe as the totality of your business. If your structural signals are unmanaged or contradictory, you allow statistical chance to determine how your capabilities are indexed, synthesized, and ultimately sold—or overlooked.

This is where retroduction becomes necessary. Rather than attempting to guess what an algorithm wants to hear, retroduction works backward from observed empirical events to identify the underlying mechanisms that must have produced them. In practice, this is precisely what a semantic audit performs: tracing contradictory indexation signals and broken entity graphs back to the structural disconnects in your company’s underlying digital footprint.

It is also worth separating the transitive from the intransitive dimension. The transitive dimension encompasses our changing descriptions, schemas, and models of the world; the intransitive dimension consists of the actual entities those models describe. Semantic architecture does not fabricate an identity, manufacture relevance, or invent expertise out of thin air. Its job is purely epistemic: to ensure the intransitive reality of your business—what you actually are and what you can actually deliver—is clearly, unambiguously knowable to machine readers.

What to Tell Your Dev Team Today

If you want to close the gap between what your business actually does and what machines infer, give your team these three immediate directives:

  1. Audit SameAs Array Integrity: Have engineering dump every @id and sameAs URI across your primary JSON-LD blocks. Ensure they resolve strictly to definitive, canonical authority sources (e.g., Wikidata, official registry entries, primary company profiles) rather than generic third-party directories or defunct domain variants.
  2. Eliminate Schema Collisions across Templates: Check whether legacy CMS templates are injecting conflicting Organization or Service schema types on the same page. If Page A asserts your business is a "ProfessionalService" while Page B calls it a "TechArticle" or generic "WebPage," vector retrievers will treat the entity as low-confidence noise.
  3. Map Canonical Parent-Child Relationships: Ensure product or service variants explicitly use isPartOf or subOrganization nodes back to your primary @id root rather than floating as isolated content blocks.

‍

What a Semantic Report Actually Tells You

Logo with the text 'Formant & Thread' in white italic font on a black background, accompanied by a bright green abstract infinity-shaped design to the right.
Steve Williams
11 Jan 2022
•
5 min read

‍

Most brands inhabit what psychoanalyst Jacques Lacan termed the Imaginary: a polished self-image reflected back to them in internal slide decks, brand guidelines, and executive summaries. It is comfortable, cohesive, and largely fictional.

The problem arises when this imaginary identity meets the search engine or large language model—which acts as a modern Big Other. The Big Other does not care about your aspirational vision statements. It enforces a cold Symbolic Order, indexing only the hard, structured facts it can parse.

                 

When a company's internal self-perception fails to match machine indexation, you get misrecognition. The brand believes it is communicating leadership in enterprise infrastructure; the algorithm categorises it as a mid-tier local agency.

This is where a Semantic Report intervenes. It performs three critical functions:

  • Establishing the Quilting Point (Point de Capiton): The anchoring signifier that stops meaning from sliding into infinite, ambiguous regress. Words like "innovative" or "scalable" mean nothing in isolation. An entity schema provides the digital quilting point—a unique URI and defined relationships—that pins down precisely what your company is, what it owns, and what it delivers.
  • Mapping the Relational Web: An entity has no standalone meaning; it exists purely through its network of relationships—its client projects, board members, proprietary methodologies, and geographic roots. A semantic audit maps this graph to ensure search engines see your business as a node in a real ecosystem.
  • Exposing the Broken Hammer: Friedrich Schelling wrote of the "ground"—the dark, unseen foundation that supports existence without drawing attention to itself. Heidegger famously noted that we only notice a hammer when it breaks. A Semantic Report operates like that broken hammer in the workshop, exposing where your schema breaks, where entity references conflict, and converting invisible digital plumbing into visible, actionable strategy.

What to Tell Your Dev Team Today

To stop treating AI retrieval as a copy-editing problem and start addressing it as an architectural one, hand your engineering team these three tasks:

  1. Establish a Single Global Root @id: Verify that every page on your domain references a single, centralized entity URI (e.g., [https://yourdomain.com/#organization](https://yourdomain.com/#organization)) in its JSON-LD headers, rather than redefining local, unlinked Organization blocks on every page render.
  2. Run an Entity-Extraction Test on Core Landing Pages: Pass your 5 key service URLs through an open-source NLP or LLM extraction prompt (or Google's Natural Language API). If the machine extracts competitors, generic buzzwords, or incorrect parent entities instead of your core capabilities, your DOM markup lacks semantic anchor nodes.
  3. Deprecate Unstructured Microdata in Favor of Pure JSON-LD: Inspect legacy codebases for inline Microdata or RDFa mixed into HTML tags. Consolidate all entity definitions into a clean, dynamic JSON-LD injection block in the document <head> to prevent parser drop-offs.

‍