Primary authority
Statutes, regulations, standards bodies, government guidance, vendor advisories, repositories, release notes, filings, specifications, and official technical documentation are preferred when they directly answer the claim.
Evidence first
Human accountability
How Zeph Tech decides what is publishable.
A useful research library needs more than volume. These standards explain how Zeph Tech approaches sources, factual review, uncertainty, corrections, AI-assisted work, commercial conflicts, and the maintenance of older material.
Last materially reviewed: August 20, 2026. These standards describe the intended editorial process; they do not turn every historical article into a current statement of fact.
Evidence hierarchy
Use the closest credible source to the underlying fact.
Source quality depends on the question. A regulation is best supported by the regulator or legal text; a product capability by current technical documentation; a vulnerability by the advisory, maintainer, or primary disclosure; and a market claim by evidence that actually measures the market.
Independent verification
High-quality reporting, research papers, testing organizations, and specialist analysis can add context, challenge a primary source, or cover facts that the original actor cannot independently validate.
Vendor claims
Vendor-provided metrics, security statements, benchmark results, roadmaps, and performance claims are identified as vendor claims when independent verification is absent or materially different.
Secondary summaries
Aggregators, summaries, newsletters, and social posts can be discovery signals, but they should not substitute for an available primary source when the distinction affects accuracy.
Date fit
The publication date of a source is not enough. The source must describe the version, rule, event, product state, or requirement relevant to the article's claim.
Specificity
A homepage or generic vendor page rarely supports a detailed factual statement. Citations should reach the particular document, advisory, standard section, release note, or dataset used in the analysis.
Corroboration
When a claim is consequential, surprising, disputed, or difficult to verify, a second independent source is preferred unless the primary record itself is definitive.
Evidence gaps
If the available evidence cannot support a confident conclusion, the article should narrow the claim, explain the gap, or avoid publishing the unsupported conclusion.
Editorial review
Review the claim, not only the prose.
Editing for grammar can make an unsupported statement sound more authoritative. The review target is therefore the relationship between the claim, the source, the date, and the confidence expressed in the language.
- 1
- Scope. Confirm the article answers a defined question instead of combining unrelated developments into a generic technology summary.
- 2
- Claims. Identify the statements that materially affect a reader's decision and verify that the cited evidence supports what the sentence actually says.
- 3
- Dates. Check effective dates, publication dates, version numbers, transition periods, end-of-life dates, and whether a proposal has become a final rule.
- 4
- Language. Remove certainty that the evidence does not justify. Distinguish observed facts from forecasts, recommendations, assumptions, and vendor assertions.
- 5
- Action value. Explain who is affected, what should be checked next, and which decision should not be made from the article alone.
AI-assisted work
AI may assist the workflow. It is not the authority.
Zeph Tech may use AI-assisted tools for bounded tasks such as organizing notes, comparing documents, identifying questions to verify, drafting structure, formatting data, or improving clarity. Published responsibility remains human.
- No synthetic sourcing. An AI output is not accepted as evidence for a factual claim merely because it sounds specific or includes a citation.
- Verify generated details. Dates, quotations, identifiers, product names, legal references, technical parameters, and numerical claims require inspection against the underlying source.
- Protect sensitive material. Research and client workflows should not place protected records, credentials, personal data, confidential client information, or restricted source material into tools without an approved handling basis.
- Keep responsibility visible. Automation can accelerate a task, but it does not transfer accountability for what Zeph Tech publishes.
Corrections and challenges
Make it easy to challenge a claim and possible to see meaningful changes.
A correction is not a failure of editorial independence. Refusing to correct a verified error is. Zeph Tech distinguishes material corrections from routine copy edits so readers can understand whether the substance changed.
- 01
Receive the challenge.
Readers, vendors, researchers, and affected organizations can send a correction or source question through the contact page. The most useful reports identify the URL, disputed statement, and supporting evidence.
- 02
Re-check the record.
The disputed claim is reviewed against the cited material and any stronger evidence supplied after publication. A vendor disagreement is considered, but it does not automatically override independent evidence.
- 03
Classify the change.
Spelling, formatting, and non-substantive clarity edits may be corrected silently. Material factual changes, reversals, or date corrections should be disclosed in the article or revision history when practical.
- 04
Fix dependent surfaces.
When a material error affects summaries, search metadata, related guides, downloadable assets, or other generated pages, the correction should extend beyond the first article when those surfaces repeat the error.
- 05
Learn from the defect.
Recurring error patterns should become validation rules, source requirements, template changes, or editorial checks so the correction improves the publishing system rather than only one page.
Commercial independence
Advertising and product interests do not get a veto over evidence.
Zeph Tech operates a research library, publishes advertising-supported editorial pages, and develops ZephCMS. Those activities can coexist only when readers can distinguish editorial judgment from commercial intent.
- A
- Advertising. Advertising relationships do not purchase favorable research conclusions. Ads should remain visually distinct from publisher content and are intentionally excluded from high-intent decision and trust pages.
- B
- ZephCMS. Research may link to an internal ZephCMS overview when the product is contextually relevant, but a product link does not convert a research conclusion into a product recommendation.
- C
- Sponsored work. If Zeph Tech publishes paid or sponsored material, the commercial relationship should be disclosed clearly enough that a reasonable reader can identify it before relying on the content as independent analysis.
- D
- Conflicts. A material relationship with a company, product, or subject should be disclosed when it could reasonably affect how a reader interprets the work.
- E
- Vendor responses. Vendors may challenge factual claims and provide evidence. They do not receive automatic editorial control over conclusions, wording, or whether a substantiated criticism remains published.
Maintenance and historical content
An archive should preserve history without pretending history is current.
Technology, law, products, and security conditions change. Older research can remain useful when its date and scope are visible, but stale material should not be promoted as current guidance merely because its URL still works.
- Preserve publication dates. Readers should be able to understand when an article described the world and avoid treating an old state as the current state.
- Quarantine suspect work. Material that cannot be confidently verified may be removed from public discovery while the underlying source is retained for deliberate repair.
- Prefer durable updates. Evergreen guides and decision resources should be updated when requirements, links, products, or practices change enough to affect the recommendation.
- Avoid artificial freshness. Changing a modified date without materially reviewing the content does not make the underlying guidance current.
Inspect the work
Use the standards against the research itself.
Read a current briefing, open the cited sources, and challenge anything that does not meet the standard described here. The goal is not to ask readers for blind trust; it is to make responsible verification practical.