Editorial Policy

AgentDesk exists to help operators — engineers, founders, researchers, marketers, and support leaders — make better decisions about AI agents. To earn that trust we hold ourselves to the editorial standards below. We update this policy whenever our process changes.

Our Coverage Areas

We cover six beats and nothing else:

• Autonomous agents — general-purpose, browser, and OS-level agents.
• Coding agents — Devin, Cursor, Copilot, Replit, Codex-class systems.
• Customer-support agents — voice and chat resolution platforms.
• Research agents — deep-research, literature-review, and analyst tools.
• Marketing & sales agents — SDR, content, and outbound automation.
• Productivity agents — knowledge, inbox, calendar, and meeting agents.

How We Pick Topics

Each weekday we publish one deep article. Topic selection is driven by three signals: (1) launches and benchmarks from the past 7 days, (2) reader questions submitted via our contact form, and (3) trending search demand on long-tail AI-agent keywords. We avoid duplicating coverage of the same product within a 30-day window unless something materially new has shipped.

Sourcing Standards

• Primary sources first: vendor docs, model cards, official benchmarks, arXiv preprints, GitHub repos.
• At least one independent secondary source (peer-reviewed publication or major tech outlet) for any quantitative claim.
• Direct quotes are attributed to a named person, role, and publication date.
• We link to sources inline, not just in a footer.

Fact-Checking

Before publication every article passes a four-point check: (a) every numeric claim has a citation, (b) every product capability is verified against current vendor documentation, (c) competitor comparisons are based on the same evaluation criteria across vendors, (d) pricing claims include the date the price was observed.

AI Assistance and Human Review

AgentDesk uses an AI-assisted editorial pipeline: large language models help draft, structure, and stress-test articles against the standards above. Every article is then scoped by a human editor against a topic brief, fact-anchored to known products and benchmarks, reviewed for tone and accuracy, and signed off before going live. We label AI assistance transparently and never present AI-generated text as original interviews, quotes, or first-person experience.

Conflicts of Interest

We don’t take paid placements, sponsored posts, or vendor-written articles. When a writer has a financial relationship with a covered company — equity, advisory, consulting — we disclose it inside the article. Affiliate links, when present, are disclosed per our Disclaimer.

Corrections Policy

We correct material errors quickly and openly. Email editorial@agentsdesk.app with the article URL and the issue. Confirmed corrections are typically applied within 48 hours and noted with a dated correction line at the top of the affected article.

Right of Reply

Companies and individuals named in an unfavorable light may submit a response of up to 200 words to editorial@agentsdesk.app. We publish substantive on-record replies as an inline editor’s note.

Reader Safety

When we cover agents that can execute code, browse the web, or spend money on a user’s behalf, we flag risks (prompt injection, runaway costs, data exfiltration) and link to mitigations. We don’t publish jailbreak how-tos or content that primarily enables fraud, surveillance, or harm.

Contact

Tips, corrections, and on-record feedback: editorial@agentsdesk.app. General inquiries via our contact form.