Editorial Guidelines
Every review on Oversite follows the same process. We test the product, measure the results, and publish transparent findings. This page explains exactly how we do that.
Testing Methodology
We do not review products based on press releases, marketing materials, or vendor demos. Every product reviewed on this site has been used directly by our editorial team under real-world conditions.
For AI Models
- - We run standardized prompt sets across writing, coding, analysis, and creative tasks
- - We record and compare outputs across models for the same prompts
- - We reference published benchmark scores (MMLU, HumanEval, GPQA, Arena Elo) and cite sources
- - We verify pricing, context window sizes, and rate limits against official documentation
For AI Tools
- - We sign up for a paid plan (not just the free tier) whenever possible
- - We use the tool for a minimum of two weeks before publishing
- - We evaluate output quality, ease of use, pricing value, support responsiveness, and integration options
- - We compare directly against competing tools in the same category
Scoring Criteria
We rate tools on a 1 to 5 scale. Scores are not assigned by algorithm. They reflect our editorial team's assessment after hands-on testing.
| Score | Meaning | Description |
|---|---|---|
| 5.0 | Exceptional | Best in class. Sets the standard for its category. |
| 4.0 - 4.9 | Excellent | Highly recommended. Minor shortcomings that don't affect core use. |
| 3.0 - 3.9 | Good | Solid choice for the right use case. Notable limitations. |
| 2.0 - 2.9 | Fair | Below average. Better alternatives exist in most scenarios. |
| 1.0 - 1.9 | Poor | Not recommended. Significant issues with quality, pricing, or reliability. |
Update Schedule
AI moves fast. A review written six months ago may already be outdated. We follow a strict update schedule to keep our content accurate.
Monthly
All tool reviews and model pages are checked for pricing changes, feature updates, and accuracy.
Weekly
Benchmark leaderboards and API pricing tables are refreshed with the latest data.
Immediately
Major model releases, significant pricing changes, and breaking news are covered within 24 hours.
Every page on Oversite displays a "Last Updated" date. If you see a date older than 60 days on a review, let us know.
Independence Policy
Advertisers and affiliate partners do not influence our editorial content, scores, or rankings. Ever.
Some links on Oversite are affiliate links. When you click one and make a purchase, we may earn a commission at no additional cost to you. This is how we fund our testing and keep the site running without paywalls.
Our editorial process is separate from our business relationships. The team that writes and scores reviews does not have access to affiliate revenue data for individual products. A tool with a generous affiliate program will receive the same scrutiny as one with no program at all.
For full details, read our Affiliate Disclosure.
Corrections
We make mistakes. When we do, we correct them promptly and transparently. If you spot an error in any review — wrong pricing, outdated feature information, incorrect benchmark data — contact us at editors@oversite.org. Corrections are noted inline with the date of the fix.