Content standards for pages that need to earn trust.

Gadex runs production with AI. Each article is governed by planning, source checks against named references, claim verification, and CMS preparation, and your team can approve every draft before it goes live.

Claims and statistics need traceable sources.

Source quality is verified against named references.

Every article is prepared for your review, approval, or automatic publishing.

01

Source quality

Primary sources are preferred where available. These include official documentation, research papers, standards bodies, government sources, product documentation, and first-party data.

When secondary sources are used, they need to be credible, relevant, and current enough for the claim being made. A source should prove the point, not merely sit near it.

02

Fact checking

Claims, statistics, citations, and definitions are checked before an article is delivered. Risky or unsupported claims should be removed or rewritten.

The review focuses on the substance first: whether the article answers the intended question, whether important claims can be defended, and whether the page adds useful detail.

Visual summary

The claim-to-source decision path

Claim-to-source fact-checking workflow A draft claim is checked against a source for recency, authority, relevance, and exact support, then retained with a citation, rewritten with a boundary, or removed. EXAMPLE EDITORIAL WORKFLOW · NOT AN AUTOMATED TRUTH SCORE 01 Draft claim “AI-assisted content is not penalized merely for using AI.” 02 Primary source Google Search Central Current public guidance Linked beside the claim 03 Support check ✓ Current enough for the claim ✓ Authoritative for Google policy ✓ Supports the exact nearby wording RETAIN + CITE The source supports the claim. Keep the boundary and link visible. REWRITE Narrow what the sentence says. Match the source’s actual scope. REMOVE No support No claim.
Example editorial workflow. Source quality depends on the claim; the diagram is not an automated truth score.
03

AI-run production

Gadex uses AI to run research, outlining, drafting, editing, and fact-checking against named sources. The pipeline owns the article job, source quality, claim accuracy, structure, tone, and CMS readiness.

Your team decides how content goes live. In approve-first mode you review and approve every draft in your CMS before publishing. In autopilot mode, approved work publishes to your CMS automatically.

04

Editorial specificity

Generic content is edited down or rebuilt. Useful pages need a clear reader, concrete examples, constraints, comparisons, and next steps tied to the site.

If an article could appear on any competitor site with only the brand name changed, it is not specific enough.

05

Publishing readiness

A finished article should include the page title, meta description, internal links, heading structure, CMS formatting notes, source notes, and approval status.

The handoff should make it clear what is ready to publish and what still needs client confirmation.

06

Approval boundaries

Gadex fact-checks claims against named sources inside the pipeline. Your team holds final approval of company-specific, legal, medical, financial, regulatory, or product-sensitive claims, and can approve every draft before it publishes.

Those boundaries are agreed during the discovery call so the approval path is clear before production starts.

How this page is sourced

This page describes how Gadex works, written by Gadex. It is not a study and carries no dataset: statements about our own process are what we do, and statements about search or AI-answer behaviour follow the platforms’ current public documentation. Where something could not be supported, it is written as our view rather than as fact.

Search and answer engines change without notice, so check anything you are about to act on against the source. Our methodology sets out how pages are researched and reviewed, and our editorial standards set out what we will and will not claim — including how to tell us if something here is wrong.