Kimi AI Guide publishes independent English-language research about Kimi products, models, pricing and developer tools. This policy explains how we select topics, verify claims, distinguish evidence from opinion and correct errors.
Kimi AI Guide is not affiliated with Moonshot AI or Kimi. Official Kimi pages remain the authority for accounts, billing, legal terms and service availability.
Our editorial priorities
We prioritize:
- Accuracy over speed. A page is not complete merely because text has been published.
- Primary evidence. Official documentation and reproducible observations come before summaries from other sites.
- Independent search intent. Each page should answer a distinct reader question and contribute useful evidence, not exist only to increase the article count.
- Clear product boundaries. We separate Kimi Membership, Kimi Code, Open Platform API and Kimi Business.
- Visible uncertainty. We label unverified, changing or conflicting information.
- Practical value. When appropriate, pages include original screenshots, calculations, checklists, code or repeatable tests.
Topic selection
We publish a page when it can do at least one of the following:
- answer a distinct Kimi-related question more clearly than an existing page;
- verify a fast-changing fact such as a price, model ID or usage rule;
- document a workflow readers can reproduce;
- compare products using a stated method and common criteria;
- expose a limitation or source conflict that affects a decision; or
- provide a practical reference, calculator, checklist or troubleshooting path.
We do not publish near-duplicate pages solely for small keyword variations. When two queries have the same intent, one stronger page should cover both.
Source hierarchy
Our preferred source order is:
- Kimi or Moonshot AI API documentation, product documentation and changelogs;
- official Kimi product pages, help-center articles and account interfaces;
- official model reports, repositories, announcements and terms;
- direct, reproducible observations made under our testing methodology;
- high-quality independent research for context, clearly identified as secondary; and
- community reports only when primary evidence is unavailable and the limitation is explicit.
Search snippets, copied price tables and unsourced social posts are discovery aids, not final evidence.
Verifying changing facts
Prices, quotas, model availability, interface labels, policies and API behavior can change without notice. For these claims we aim to record:
- the exact source URL;
- the date checked;
- the relevant product and account type;
- the displayed currency or billing unit;
- any regional or login dependency; and
- conflicts with another official page.
The most specific current source normally controls. For example, a dedicated billing page is usually stronger than a general marketing paragraph. If equally authoritative official sources disagree, we publish the discrepancy instead of choosing the more attractive claim.
Tests and first-hand claims
We only call a result “tested,” “observed” or “measured” when the stated test was actually completed.
Every meaningful live test should identify:
- the date and access level;
- model, product or interface version where available;
- task, prompt, input file or dataset;
- relevant settings and tools;
- number of runs;
- output or measurable result;
- failure cases; and
- limitations that prevent broader conclusions.
If required login, payment, software or data is unavailable, we stop. We may still publish a documentation-based guide, but it must not imply that we ran the product.
See Testing Methodology for the full protocol.
Screenshots, images and evidence
Original screenshots should show the exact interface or result being discussed and include a caption with the date and context. Private details, API keys, billing identifiers and unrelated personal content must be removed or obscured.
We do not use decorative evidence. A screenshot is included only when it helps a reader verify a step, understand a limitation or see a material interface state. Images should have descriptive alternative text and must not link to an unrelated page.
Third-party images are used only with an appropriate license or another valid basis. Product screenshots and logos are not presented as our property.
Reviews and comparisons
A review or comparison should state:
- the decision it is designed to support;
- the products and versions compared;
- common criteria applied to each option;
- what was tested directly and what came from documentation;
- pricing date and billing assumptions;
- material advantages and disadvantages; and
- any relationship that could influence the coverage.
We avoid declaring an absolute “best” product from a single prompt, vendor benchmark or subjective impression.
Technical content and code
Technical guides are checked against official documentation. When possible, code is run in a controlled environment before we label it tested. Samples use placeholders for keys and secrets, include required versions and avoid unsafe defaults.
A working demonstration is evidence only for the documented environment and date. It does not guarantee compatibility with later SDKs, models or account configurations.
Authorship and responsibility
Content maintained by this project uses the byline KI AI Team. The byline indicates responsibility under this policy; it does not imply official Kimi status or undocumented professional credentials.
The current verification date and relevant sources should be visible on high-change pages. Substantive editorial changes should update the page’s modified date without rewriting the original publication date as though the page were new.
Corrections
Readers can report possible errors to [email protected]. A useful report includes the page URL, the disputed sentence, supporting source and the date the source was viewed.
When a material error is confirmed, we correct the page and add a note that explains what changed. Typographical or formatting fixes that do not alter meaning may be made without a correction note.
Read Sources & Corrections for the submission process.
Advertising, affiliates and sponsored material
Advertising may fund the site. Ads do not buy favorable conclusions or access to unpublished test results.
If affiliate links are used, they will be disclosed. The possibility of earning a commission does not change the price paid by a reader unless the offer itself states otherwise. Sponsored material, if ever accepted, must be clearly labeled and separated from independent editorial coverage.
We do not encourage readers to click advertisements to support us. See Advertising Disclosure.
Independence and conflicts of interest
We disclose a free product, paid access supplied by a vendor, sponsorship, affiliate relationship or other material consideration when it relates to coverage. We do not allow a vendor to approve an independent conclusion before publication.
Product companies may be asked to clarify facts. Their response can be included, but it does not replace our evidence or give them editorial control.
Policy changes
We may revise this policy as the site, its testing capabilities or applicable requirements change. Material revisions should update the “Last reviewed” date.
Questions about this policy can be sent to [email protected].