View as
Method & case

77 days: how an overseas website became an AI-citable source

A method-focused case: how brand knowledge, real user questions, website content, quality checks, and performance monitoring connect into a continuously optimized loop.

This case describes a reusable operating method: sort out facts and knowledge first, construct user questions next, then plan, produce, and check website content, and finally let monitoring results feed the next iteration. The article does not disclose the client's identity or performance data without public authorization.

In brief

The focus of the 77 days was not publishing a batch of articles at once, but building a GEO working system that keeps running: brand knowledge, user questions, growth planning, content production, quality checks, website publishing, performance monitoring, and next-round optimization connect to each other.

This article focuses on the method and the working assets retained; it does not disclose the client's identity or performance figures without public authorization.

Starting point: a website like a brochure

When the project began, the website already held a fair amount of information, yet it did not operate as a growth system built around user questions. Key facts were scattered across materials, user questions were not the basis of the content structure, external communication was not fully aligned with the website, and performance monitoring could not reliably guide the next round.

The work therefore had to fix fact consistency and process first, rather than immediately adding articles.

The working loop that keeps running

The team organized the work into eight connected steps:

  1. Organize brand and product knowledge.
  2. Construct real user questions.
  3. Set the growth plan and content priorities.
  4. Produce content by question and evidence.
  5. Check facts, sources, question match, and answerability.
  6. Publish to the brand website.
  7. Monitor AI visibility, site behavior, and business signals.
  8. Plan the next round of optimization from the results.

The first round builds the system; later rounds keep calibrating against feedback.

Why start from the knowledge base

The knowledge base is GEO's infrastructure. If brand facts are scattered, vague, or contradictory, content production and citation cannot stay reliable.

First, turn marketing language into checkable facts — material names, certifications, test results, usage scenarios, craft details. Second, identify conflicting accounts across pages and materials, and produce a unified version later pages can follow. Third, let content production, scoring, editing, and review all build on the same factual base, so automation does not amplify inconsistency.

What the client keeps is a knowledge asset that keeps working, not just published articles.

How query construction changes the website

Traditional websites often start from what the brand wants to show; a content system aimed at GEO starts from which questions users and AI need answered: what problem the user is solving, what they are comparing, which evidence the answer needs, what the website is responsible for stating, and which page AI should cite when a related question appears.

A broad topic can split into several pages, each targeting a concrete question. Each page's duty is clearer: answer the question directly, state the scope of applicability, show the basis, and link back to the website's authoritative source. This structure also improves human browsing and AI retrieval at the same time.

External sources still matter

The website should be the first-party source, but it should not be the only signal AI can find. External sources corroborate the website when their accounts agree — they should not replace the website's official statement of brand facts.

Third-party content can be edited, deleted, or read differently; the website is more stable and easier to attribute to the brand. The working principle: establish the website's official source first, then supplement with external material.

The assets the client keeps

The project left a set of working assets the client can keep using:

  • A unified brand and product knowledge base
  • A validated query baseline
  • A content planning structure tied to user intent
  • Rules for checking source clarity and answerability
  • Published first-party website pages
  • Records of external sources used for corroboration
  • Monitoring configuration and a review cadence
  • Rules and Agents that can keep producing and checking content

Content services deliver concrete outputs; a GEO system also leaves behind inputs, rules, evidence, and process capability.

What this case does not claim

What transfers is the working method, not unproven promises of results. This version of the article publishes no performance figures, because public authorization and metric definitions must be confirmed first; nor does it claim different categories will get the same results. High-decision-cost products may produce more comparison and checking questions, while low-decision-cost products may need different content, ads, and conversion paths.

The reusable working order is: build first-party knowledge, construct real questions, publish answerable source pages, add external corroboration, monitor AI and user reactions, and let data arrange the next round.

How to run the first diagnosis

If an overseas website still reads like a brochure, start by answering: How does AI currently describe the brand? Which sources does it cite? Which user questions is the website missing? Which brand facts exist only in decks, PDFs, or sales conversations? Which claims should become first-party source pages on the website?

The answers help judge whether the main gap lies in content, structure, measurement, or evidence.

FAQ

What does it mean for a website to become an AI-citable source?

The website holds clear, crawlable, checkable passages that AI can use as evidence when composing answers. The website is then both a brand showcase and a source of facts.

Why not write articles first?

If fact consistency is not in order, writing at volume first can widen the inconsistencies. A knowledge base and a query graph make later content more accurate, better structured, and better suited to answer questions.

Is buying external media placement required?

External sources can supply corroborating support, but they should not replace the website. Key claims should appear on the brand's own pages first, supported by other sources as needed.

Can a method case be done without publishing client numbers?

Yes. A case can describe the process and the working method. Public figures should wait until authorization and measurement definitions are confirmed.

What should the first diagnostic step be?

Ask an AI assistant how it describes the brand, note the sources it cites, compare the answer against the website's official facts, and identify the content gaps to fix first.