Seekde AI Search and Discovery Intelligence
Seekde
⌕ Try Seekde →
Seekde

How Seekde Works

Learn how the Seekde interactive preview demonstrates query entry, intent breakdown, sample answers, and illustrative source attribution.

Understanding the Seekde Discovery Model

Seekde is designed to explore how modern search interfaces present structured sub-intents, synthesize multi-topic answers, and display transparent source references.

Today, Seekde is organized around two complementary initiatives:

  1. An Interactive Explorer Preview: A client-side demonstration on the homepage that illustrates intent classification, modular topic breakdown, sample answer presentation, and illustrative source attribution cards.
  2. An Independent Research Publication: A specialized editorial desk focused on researching AI search platforms, crawling architectures, and content optimization methods.

This guide explains how the current preview operates step-by-step, details its exact frontend behavior based on accepted code, and outlines the planned architecture for future multi-engine discovery tools.

The Current Interactive Preview

The interface on the Seekde homepage is deployed as an Interactive Preview. It demonstrates a structured alternative to traditional search results pages using curated client-side demonstration datasets rather than live commercial AI model APIs.

PROCESS WORKFLOW
01

[ User Query Input ]
→
02

[ 1. Intent Breakdown ]

(Curated Intent Label & Topic Scope)

→
03

[ 2. Sample Answer UI ]

(Direct Answer Summary + Topic Cards)

→
04

[ 3. Attribution View ]

(Illustrative Source Cards + Follow-ups)

The preview workflow demonstrates three discovery stages:

  • Query Entry: Visitors enter a search question or select a preset demonstration topic.
  • Intent Breakdown: The interface displays an illustrative intent classification and presents structured topic branches.
  • Sample Answer & Attribution: The preview presents a modular summary paired with illustrative source cards and contextual follow-up chips.

Step-by-Step Exploration Walkthrough

Step 1 — Ask a Question

Visitors begin at the single-prompt explorer on the homepage. The interface includes:

  • A text input field (#productSearchInput) pre-populated with an example question and supporting arbitrary text entry.
  • A primary action button (Explore →) to trigger the demonstration workflow.
  • Preset example chips providing one-click access to curated demonstration queries:
    • AI search citations: "How do AI answer engines choose sources?"
    • Generative search: "How does generative search work?"
    • SEO vs GEO: "SEO vs GEO search optimization"
  • An animated status indicator displaying simulated progress messages ("Understanding your search intent…", "Finding the most relevant topics…", "Organizing the answer and sources…") during an 1,150ms interface transition.

Step 2 — Explore the Illustrated Intent

When a query is submitted, the frontend selects an illustrative dataset using deterministic client-side keyword matching:

  • Intent Classification Label: The preview displays a curated intent badge (such as Informational & Technical exploration, Conceptual & Architectural inquiry, or Comparative analysis).
  • Goal Context: A sidebar card displays explanatory notes illustrating how an answer engine can adapt response depth to match the user’s primary goal.
  • Topic Scope Segmentation: The demonstration presents four modular subtopic cards illustrating how complex questions can be broken into prerequisite concepts.

Step 3 — Review the Sample Answer

The preview presents an illustrative answer container structured for clear information hierarchy:

  • Quick Answer: A concise 2–3 sentence overview addressing the core question directly.
  • Modular Subtopic Cards: Four interactive buttons displaying topic titles and descriptive subheadings. In the current interface, clicking any topic button populates that topic into the search bar and triggers its corresponding preview dataset.
  • Status Pills: Contextual tags indicating illustrative parameters (such as Source-aware and Exploration-ready).

Step 4 — Inspect Illustrative Sources

Below the answer summary, the interface displays curated attribution cards:

  • Rendered Fields: Each source card renders a numerical index (1–3), a source organization name button, and a descriptive summary explaining its topical relevance.
  • Interactive Feedback: Clicking any source link triggers an illustrative notification toast ("Sample source reference — illustrative preview."), clearly reminding users that citations in the current preview are demonstration examples.
  • Attribution Design: Demonstrates how answer engines can keep underlying references visible alongside synthesized summaries rather than burying links.

Step 5 — Explore Related Topics and Follow-Up Questions

In the right sidebar, the interface displays four contextual follow-up suggestions (#exploreNext):

  • Presents logical next research questions connected to the primary query.
  • Allows visitors to click any suggestion to immediately load that topic into the explorer.
  • Demonstrates how conversational search interfaces facilitate continuous research journeys.

Fallback & Keyword Routing Behavior

The current preview relies on local client-side datasets defined in product-home.js:

  • Mapped Keyword Routes: The frontend evaluates input text for specific keyword triggers (such as geo, vs, generative, source, citation, or ai) to load one of three specialized demonstration datasets.
  • Fallback Dataset: When an arbitrary question is entered that does not match these predefined triggers, the interface loads a general demonstration dataset with the intent label Exploratory research (Interactive Preview).
  • Preview Boundary: The preview is not connected to a live web-scraping engine or a live LLM backend; it demonstrates interface architecture and response structure.

What the Preview Does Not Do Yet

To maintain complete product truthfulness, Seekde explicitly outlines the operational boundaries of the current preview:

  • No Live Semantic Parsing: The interface does not run natural language processing models on arbitrary user input; intent labels are assigned from curated local datasets.
  • No Real-Time AI API Querying: The explorer does not execute external API calls to OpenAI, Google, Anthropic, or Perplexity.
  • No Live Web Retrieval: The system does not crawl or scrape live web pages upon query entry.
  • No User Query History Storage: The explorer operates entirely client-side and requires no user account, profile creation, or database storage of entered queries.
  • No Automated GEO Scoring: Seekde does not currently generate automated domain visibility scores or crawlability audits via API.

Current Preview vs. Planned Future Product

Capability Dimension Current Interactive Preview Planned Future Production System
Query Processing Curated demonstration datasets with keyword routing Real-time parallel multi-engine querying
Engine Surfaces Illustrative answer engine interface Comparative evaluation across supported AI search platforms
Source Citations Curated demonstration source cards with preview toast Dynamic citation extraction and persistence tracking
Brand Monitoring Educational research guides Automated AI Share of Voice measurement and brand alerts
Crawlability Audits Editorial checklists and teardowns Automated robots.txt and structured data extraction API
Data Storage Client-side only (no user account or query logging) Optional team workspaces and custom prompt monitoring sets

Why Seekde Separates Intent, Answers, and Sources

Traditional search engines collapse information retrieval into a list of ranked blue links, requiring users to manually visit multiple websites and synthesize disparate paragraphs. Conversely, some generative chat interfaces present synthesized answers without transparent attribution, obscuring source origins.

Seekde’s product philosophy focuses on three principles:

  1. Explicit Intent Identification: Illustrating how an interface categorizes user goals helps researchers refine their queries and evaluate search scope.
  2. Modular Answer Synthesis: Segmenting complex answers into clear subtopic cards improves comprehension compared to long walls of generated text.
  3. Transparent Source Attribution: Keeping original publishers and citations visible ensures ground truth remains verifiable and creators receive proper credit.

How the Editorial Publication Fits In

The software explorer represents one component of Seekde. The project is equally dedicated to independent research and technical analysis through the Seekde Blog, anchored by our flagship monograph on the Architecture of Generative Answer Engines.

Seekde’s editorial program is being built around:

Seekde’s editorial standards and claim verification protocols are defined in our Research Methodology.

Real Current Demo Case Walkthrough

To see how the preview operates in practice, consider the default preset query available on the homepage:

> Demonstration Query: "How do AI answer engines choose sources?"

When this query is executed in the interface:

  1. Curated Intent: The preview assigns the intent label Informational & Technical exploration and displays a badge indicating structured intent.
  2. Curated Answer: The interface renders a summary explaining multi-stage retrieval, vector semantic search, BM25 keyword matching, and neural reranking.
  3. Topic Cards: The preview renders four interactive subtopic cards: Retrieval-Augmented Generation, Citation scoring, Information gain, and Structured data & schema.
  4. Source Cards: Three illustrative reference cards are displayed: ACM Digital Library, arXiv Information Retrieval, and Search Engine Land.
  5. Follow-Up Suggestions: Four exploration chips appear in the sidebar: Generative Engine Optimization, Vector embeddings vs keyword search, Direct citation tracking, and LLM crawler behavior.

Future Direction (Planned Roadmap)

The long-term vision for Seekde involves evolving from curated demonstration interfaces toward live search intelligence tooling:

  • Multi-Platform Search Comparison: Simultaneously dispatching queries across supported AI search platforms to evaluate differences in synthesized responses and source selection.
  • Citation Recurrence Tracking: Building an empirical index tracking which publisher domains are consistently cited across competitive industry queries.
  • Enterprise Brand Monitoring: Enabling marketing teams to monitor brand recommendations and factual accuracy across conversational search surfaces.

Where to Go Next

Action completed.