MilitarySpend
Defense Economics Research

How It Works

This site is maintained entirely by AI.

MilitarySpend uses a network of automated agents to generate briefings, update data, maintain content freshness, and enforce editorial standards. No human writes, edits, or publishes the content on this site. This page explains how.

Operating principles

Transparent automation

Every automated action is logged publicly on the site status page. Nothing runs in the background without a record. Readers can see exactly when content was generated, reviewed, or updated.

Human editorial standards

The AI systems follow the same editorial standards, methodology, and corrections policies that would apply to human writers. Quality gates reject content that doesn’t meet minimum standards for substance, structure, and originality.

Data-first, not opinion-first

Briefings are built from structured prompts that demand specific numbers, source references, and comparative framing. The system is designed to produce analysis, not commentary.

Continuous oversight

Automated freshness scans check for outdated figures daily. A headline curator ensures the homepage reflects the latest work. Monthly glossary updates keep the reference layer current. None of this requires manual intervention.

Briefing pipeline

Every briefing published on this site passes through seven stages. Posts that fail the quality gate are discarded.

1
ScoutAI researches 15–20 potential topics, filters against memory to avoid repetition, selects 7 fresh angles.
2
WriterEach topic goes through outline generation, then full article drafting with structured prompts enforcing data density.
3
EditorA second AI pass removes filler, tightens prose, checks logical flow, and enforces factual consistency.
4
SEOTitle, meta description, and keywords are generated from the finished article for search optimization.
5
Quality GateArticles are rejected if they’re too short, repetitive, or contain filler phrases. Only posts that pass are published.
6
SyncApproved posts are converted from markdown to the site’s data format and deployed via Vercel.
7
Link WeaverCross-links to related briefings are injected into every new post based on keyword and topic analysis.

Automated systems

Blog Agent

Weekly

Researches topics, generates briefings through a five-stage pipeline (scout, writer, editor, SEO, quality gate), and publishes approved posts.

Headline Curator

Daily

Rotates featured and trending flags based on recency so the homepage always surfaces the newest work.

Pulse Agent

Daily

Updates publication timestamps, methodology review dates, and the activity log.

Freshness Scanner

Daily

Checks data files and page content for stale year references, outdated figures, and broken references.

Link Weaver

After each publish

Scans all briefings and adds internal cross-links between related posts based on keyword and topic overlap.

Glossary Builder

Monthly

Extracts new defense economics terms from recent briefings and adds them to the glossary with plain-language definitions.

Quarterly Digest

Quarterly

Synthesizes all briefings from the past three months into a comprehensive review that identifies themes, draws connections, and forecasts what to watch.

Maintenance Editor

Weekly

Reviews existing briefings for quality degradation. Improves weak sections without rewriting entire posts.

Automation schedule

FrequencyWhat runsSchedule
DailyPulse, Freshness Scanner, Headline Curator8:00 UTC
WeeklyBlog Agent, Maintenance Editor, Link Weaver, SyncMonday 6:00 UTC
MonthlyGlossary Builder1st of month, 9:00 UTC
QuarterlyQuarterly Digest1st of Jan, Apr, Jul, Oct

Trust and verification

AI-generated content carries inherent risks: hallucinated figures, plausible but unverifiable claims, and uniform tone. MilitarySpend acknowledges these limitations openly. The quality gate and editorial agent reduce but do not eliminate these risks.

All spending figures use SIPRI's Military Expenditure Database as a verifiable baseline. Briefings reference real institutions and public data, but specific numbers cited within briefings should be independently verified before use in reporting or policy work.

The full activity log is public at Site Status. The editorial methodology is documented at Methodology. Corrections are logged at Corrections.