AgentStack exists to help leaders make better decisions about AI agents, copilots, AI platforms and workflow automation. That only works if readers can trust what we publish. This page explains how we choose topics, research, test, disclose paid relationships, correct mistakes and use AI in our own workflow.
We update this page when our process changes. Last updated: [DATE].
How we choose what to cover
We pick topics based on what our readers are researching, building or trying to decide on.
- Reader questions: We track the questions readers send us and the gaps we see in search results, where vendor blogs and forums often provide the only answers.
- Market changes: We cover launches, pricing and licensing changes, acquisitions, new protocols and regulation that change how teams build or buy AI.
- Direct experience: We prioritize areas where we can add a perspective that goes beyond a feature list.
- Partner proposals: Partners can propose topics for sponsored work, and those pieces are labeled. Partner proposals do not set our independent editorial calendar.
We skip trending topics when we have nothing useful to add, and we come back when we do.
How we research
Every article goes through a research phase before drafting begins.
- We read official documentation, pricing pages, changelogs, API references and model cards.
- We check primary sources such as protocol specifications, regulatory texts, public filings and research papers.
- We review independent benchmarks and published evaluations, and we note their methodology.
- We read practitioner discussions on GitHub issues, Hacker News, Reddit and community Slack or Discord groups to see how products behave in real use.
- We look at review platforms such as G2 and Gartner Peer Insights for patterns in feedback.
- We speak to practitioners and vendors when a question cannot be answered from public sources.
If a price, feature, benchmark or claim cannot be verified through a credible source, we leave it out or flag the uncertainty in the article. Figures supplied by a vendor are attributed to that vendor.
How we test products
For reviews, comparisons and benchmarks, we test products ourselves whenever it is practical.
- We sign up for a trial, a sandbox or a paid plan.
- We walk through setup, admin controls and integrations.
- We build a representative workflow or agent for the use case we are writing about.
- We record what we can measure, such as task success, latency, failure modes and cost per run.
- We note where the product matches its documentation and marketing and where it falls short.
When we publish benchmarks or teardowns, we publish the methodology alongside the results. When hands-on testing is not possible, for example with enterprise-only platforms that require a long sales process, we say so clearly and explain which documentation, interviews and third-party evidence we relied on. We never claim to have used something we have not used.
Ratings and recommendations
When we score or rank products, we publish the criteria behind the result. Scores and verdicts come from those criteria alone. If a paid placement affects which products appear in a list or the order they appear in, we disclose that at the top of the article.
How we disclose paid relationships
AgentStack is funded in part by companies that pay to be featured in our content. We disclose every one of these relationships because reader trust is what makes the publication worth reading. Our disclosures follow the US Federal Trade Commission’s guidance on endorsements and native advertising, so they are clear, close to the content they describe and written in plain language.
- Paid placements: When a company pays to be mentioned or featured in an article, we label the mention as a paid placement where it appears. We also add a disclosure at the top of the article, before the first paid mention.
- Sponsored articles and reviews: Sponsored content is labeled “Sponsored” or “Partner content” at the top of the article, in the byline area and in any social or newsletter promotion. The sponsor can suggest the topic and review the draft for factual accuracy. The sponsor cannot dictate our opinion, rewrite our analysis or remove limitations we found.
- Affiliate links: Some links earn us a commission if a reader signs up or buys, at no extra cost to the reader. We disclose affiliate links in the article, and commissions never change our assessment of a product.
- Free access and travel: We disclose when a company has given us free accounts, extended trials, demo environments or event travel. Access never buys a positive review.
- Other relationships: If an author or [COMPANY LEGAL NAME] has an investment, advisory role or consulting relationship with a company we cover, we disclose it in the article. [CONFIRM CONFLICT OF INTEREST POLICY]
- Links: Paid and affiliate links carry rel=”sponsored” attributes.
Our standard disclosure at the top of an article reads: “Disclosure: [Company] paid to be featured in this article. Our assessment of every product, including [Company], is our own. See our Editorial Standards for details.” Inline labels read “Paid placement” or “Sponsored.” Our Disclosure page summarizes how we label paid placements, sponsored content and affiliate links.
If you ever see something on AgentStack that looks like an undisclosed paid relationship, email [CORRECTIONS EMAIL] and we will investigate and fix it.
How we handle corrections
We aim for accuracy, but AI products change weekly and mistakes happen. When we get something wrong, we fix it.
- Minor edits: Typos, formatting and broken links are fixed without a note.
- Factual corrections: Errors in pricing, features or claims are corrected in the article, with a short note at the end explaining what changed and when.
- Significant changes: Revised recommendations or retracted claims get a clear correction notice at the top of the article.
If you spot something that looks wrong, email [CORRECTIONS EMAIL] with the article link and, if possible, a source.
How we use AI in our workflow
We cover AI for a living, and we use AI tools in our own work. We think readers deserve to know exactly how.
Where AI helps us:
- AI tools help us summarize and organize research from many sources.
- AI tools help us draft outlines and structure long articles.
- AI tools help us edit for clarity, grammar and consistency.
- AI tools sometimes produce early drafts, which our writers then revise, fact-check and rewrite in their own voice.
What we never do with AI:
- We never publish AI-generated content without human review, editing and fact-checking.
- We never use AI to invent quotes, statistics, case studies, benchmark results or testing experiences.
- We never let AI decide what we recommend or how we evaluate a product.
- We never treat an AI output as a source, so every claim is checked against a primary source.
Every article on AgentStack is reviewed and edited by a named human before it goes live. The judgment, testing and accountability behind each piece belong to our editors and writers.
Questions or feedback
If you have questions about how we work, or you want to flag something that does not meet these standards, please get in touch.
Email: [EDITORIAL EMAIL]
Address: [ADDRESS]