Narrative Architecture studio for buyers — what you should ask for
Short answer: If you are the buyer (CMO / corpcom / brand lead), a Narrative Architecture studio is measured by whether your pages and proof can be audited inside AI answers — not by article volume or a “GEO” slide. Ask for: a frozen prompt set, per-engine columns, URLs to fix, and corroboration — not a blended score or ranking promises.
Quick facts — five buyer questions at kickoff
- Which prompts are frozen? Version + date; buyer language, not internal jargon.
- Which engines are reported separately? GPT, Gemini, AIO, AI Mode, (optional Perplexity/Copilot) — not one donut.
- Which URLs get fixed this week? A page list, not a tip list.
- What external corroboration is planned? One independent trace per critical claim — not “we will go viral.”
- What gets re-measured +7–14 days after live? Per-engine columns + a caveat if citation volume thins.
The general studio definition already lives in the studio Narrative Architecture journal. This piece is the buyer seat: what to ask so you do not buy a mill labeled GEO.
One IP line (enough)
Narrative Architecture (White Wood): brand-story architecture you can audit — pages, facts, and proof arranged so humans trust it and machines have something solid to retrieve. Not a content mill.
Table: must-have in the proposal × red flag × why
| Must-have in the proposal | Red flag | Why |
|---|---|---|
| Versioned prompt set + baseline | “We use AI intuition” with no prompt list | No baseline = no audit |
| Per-engine score columns | One “AI SoV” / blended contract KPI | Hides Gemini/AIO gaps |
| List of page URLs to fix | “50 articles / month” as the main deliverable | Mill volume ≠ extractability |
| Third-party corroboration plan | “Guaranteed ChatGPT citation” | Citation promises are unethical / uncontrolled |
| +7–14 day re-measure window | Same-day before/after drama | Crawl & consensus take time |
| Market language (e.g. Bahasa for local buyers) | English-only blog for a local market | Retrievability in your market |
| Explicit scope limits (what is not promised) | SaaS feature tour as the core proposal | Studio = pages & proof, not a tool license |
What to ask at kickoff (short script)
- “Show the frozen prompts for week 1.”
- “Show a sample report with per-engine columns — not one score.”
- “Name 3–5 URLs to fix, and which claims get external backing.”
- “How does the creative brief pass both human and machine readers?” → practical answer: copy must read for humans and carry pinned facts machines can retrieve again — without turning either side into theater.
- “We are not buying a GEO listicle; we are buying page audit + corroboration.”
Proxemics (WW’s podcast) is sometimes used as an on-mic lab to test whether a narrative still reads for human ears: https://proxemicspodcast.com/ — one trail, not a contract requirement.
What a healthy buyer brief is not
- A best-of Jakarta GEO agency pitch as a substitute for audit.
- A SaaS / dashboard feature tour with no URL decisions.
- A promise to rank in “all AI.”
- A Narrative Architecture definition rewrite with no page deliverables.
FAQ
How is this different from an SEO agency?
SEO agencies are often scored on blue-link rank and traffic. A Narrative Architecture studio adds a requirement: whether the brand is named and citeable in machine answers for buyer prompts — with per-engine columns, not SERP position alone.
When is in-house enough?
When the team already has: frozen prompts, source-reading discipline, extractable answer-page writing, and a corroboration path (PR/corpcom). External studio help fits when engine gaps are large or internal capacity is full — not because the “GEO” label is trendy.
What does a brief that passes humans and machines look like?
The brief must pass two readers: humans (clear, not cringe) and machines (pinned facts, structure, consistency). If only one side wins, you win the slide or lose the extract — or the reverse.
Must we buy Jakarta GEO services?
Not by default. If a proposal is only a procurement checklist with no URLs and no engine columns, you are buying a label. Ask for craft: pages, corroboration, re-measure. Existing live jasa/checklist pages can filter vendors — this article is not a new services pitch.
How should we read SoV charts on the board?
Ask for per-engine columns + a volume/denominator note. SoV that “rises” while total citations in the harvest shrink can mislead. Do not use one blended AI score as a contract KPI.