Everything that governs how Agent Oracle operates — editorial standards, AI methodology, privacy, safety, and legal — in one place.
Agent Oracle publishes advice that people act on with money and headcount, so the standards behind it should be legible before you take any of it seriously. This page is the index to those standards — how articles are researched and approved, where AI is used and where a human has to sign, what we collect about you, and what you are entitled to ask us to delete.
Nothing on this page is aspirational. Each linked document describes what we actually do today, and each is dated so you can see when it last changed.
Editorial Policy
How content is created, reviewed, and corrected.
AI Content Methodology
How we use AI — and where humans stay in the loop.
Responsible AI Statement
Safety, privacy, and fairness principles.
Privacy Policy
What we collect, why, and how to control it.
Terms of Service
The agreement between us and you.
Cookie Policy
Cookies, analytics, and advertising trackers.
Disclaimer
Educational content, not professional advice.
Refund Policy
Subscription refunds and cancellations.
Contact
Reach editorial, support, or security.
Editorial independence
Agent Oracle is independently owned. No article, diagnosis path or recommendation is sold, and no advertiser, sponsor or partner is given sight of copy before publication or the ability to change it after. Where advertising appears, it is bought programmatically against the page and has no relationship to what the page says.
The publisher approves every article before it goes live and is the named point of accountability when something published here turns out to be wrong. Corrections are made on the page itself with a note of what changed, rather than by quietly editing history.
Where AI is used, and where it is not
AI drafts research summaries and first passes from public sources. It does not decide what is true, what gets published, or what a reader should do. Every published piece is read and approved by a human, and factual claims that could change a business decision are checked against a primary source before the piece goes live. Our AI methodology page sets out the full workflow, including what the models are allowed to touch.
Your data and your rights
Sessions belong to the account that created them. We do not sell prompts, transcripts or reader data, and we do not hand conversation content to third parties for model training. You can export or delete your data at any time from your account, and formal privacy requests are confirmed in writing.
Cookies fall into three buckets — the ones needed to keep you signed in, the ones that measure aggregate traffic, and the advertising cookies set when ads are served. The cookie policy explains each bucket and how consent is recorded and withdrawn.
Limits worth knowing
Agent Oracle is a publication and a thinking tool, not a licensed advisory firm. Nothing here is legal, tax, accounting or investment advice, and no article accounts for the specifics of your contracts, jurisdiction or cap table. Read the disclaimer for the full scope, and take material decisions to a professional who is accountable for them.
Why you should — and should not — trust what you read here
The standard we hold ourselves to
Every published piece states its assumptions in the body rather than burying them. Where a conclusion depends on a company's size, margin structure or sales cycle, the article says which range it is reasoning about, because advice for a firm with a six-week cycle actively harms one with a nine-month cycle.
Claims about the outside world link to their source. Claims about our own product — turn limits, retention windows, refund handling — are stated as numbers, not adjectives, so they can be checked against what the app actually does.
Where AI is involved and where it is not
Research and first drafts are AI-assisted against public sources. Structure, argument, fact-checking and the decision to publish are human. No article is published unread, and pieces that survive the draft stage unchanged are the exception rather than the rule.
Articles that update over time carry a version indicator, so you can see that a piece has been revisited rather than assuming a 2026 date means a 2026 rewrite.
Independence
There are no sponsored posts, paid placements or affiliate incentives shaping what gets covered. Advertising, where it appears, is served by a third-party network and has no influence on editorial selection or conclusions.
Subscriptions fund the publication. That creates its own bias — towards making the tool look useful — so the honest counterweight is publishing the cases where structured analysis does not help, and we do.
The limits you should hold in mind
This is a reasoning aid, not a consultant with access to your accounts. It cannot see your data, verify your numbers, or know what your last three hires were like. Everything it says is conditional on what you told it in the session.
Treat the output as a structured second opinion to argue with. The decisions remain yours, and the pieces here are written to make you better at defending them rather than to hand you a verdict.