Editorial standards
How we review
MurasakiHub publishes reviews, comparisons, rankings, and guides to make it easier to decide what to read next. This page explains the framework behind those pieces and the limits of what a label or score means.
Start with what is recorded
The archive was built over time, so not every published article has the same level or type of documented evidence. We do not assume that a review was hands-on tested just because it has a rating, and we do not retrospectively add proof, dates, or labels that were not recorded when the article was prepared.
Review framework
The questions a useful review should answer
What is it for?
Describe the platform or topic plainly, including the intended use and relevant category.
Who may like it?
Surface a practical strength, a limitation, and the reader profile the article can support.
What can be checked?
Keep facts such as a free tier, dates, pricing, or sources tied to what the article records.
What are the alternatives?
Use internal links and comparisons to help readers continue researching instead of forcing a decision.
Evidence labels
Labels are evidence, not decoration
These labels appear only when the relevant evidence is recorded for the specific article. Their absence does not imply the opposite; it means the archive does not make that claim for that page.
Hands-on tested
Used only when a direct test is recorded for that article.
Research review
Used only when the article records the research behind its assessment.
Price/source checked
Used only when the relevant price or source check and its date are recorded.
Scores and facts
How a score is made
Every scored review uses five criterion scores from 1 to 10, with at most one decimal each. We multiply each score by its weight, add the contributions, then round once to one decimal. Exactly halfway values round up: 7.850 becomes 7.9. The same calculated result appears in the review, comparison, listings and structured data.
There are no undisclosed bonuses, penalties or partner overrides. The criterion scores are editorial judgements supported by the stated evidence; the final arithmetic is reproducible. An affiliate payment does not change either.
Quality, consistency and relevance of the content or output covered by the review: catalogue and translation for stores, play experience for games, responses or generated media for AI tools.
What can actually be tried without paying, including quotas, time limits and whether the free route demonstrates the core experience. A verified absence of free access scores 1; unknown access stays unscored.
Useful access relative to its verified cost, including billing interval, credits, add-ons and restrictions. For a free service, assess the documented limits and trade-offs rather than automatically awarding 10.
Coverage and usefulness of the core capabilities in the stated review scope. Distinguish directly tested functions from documented but untested ones.
Friction in the reviewed tasks: navigation, clarity of controls and limits, readability, responsiveness and errors on the devices actually checked.
S = 0.30Q + 0.25F + 0.20V + 0.15C + 0.10U
Published score = floor(10S + 0.5) / 10
Example: 0.30 × 7.0 + 0.25 × 7.5 + 0.20 × 9.0 + 0.15 × 8.5 + 0.10 × 8.0 = 7.850, published as 7.9 / 10. Individual contributions are not rounded before adding.
How we assign criterion scores
For each criterion, record the observed strengths, limitations, source and test scope, then select the matching band below. Explain the chosen value within that band. Decimal precision does not make an editorial judgement an objective measurement.
- 1-2
- Absent or unusable for the stated purpose; major failures dominate.
- 3-4
- Limited usefulness; substantial restrictions or recurring problems.
- 5-6
- Usable for the stated purpose, with material compromises.
- 7-8
- Strong within the reviewed scope, with clear remaining limitations.
- 9-10
- Excellent within the reviewed scope, with only minor limitations supported by evidence.
Free access has a 25% weight because trying a service before paying is part of this review framework. Paid-only products can therefore score lower even when their content is strong. Compare products within the same category and read the criterion breakdown.
Missing evidence is not zero and is not an assumed average. If any criterion cannot be supported, publish the review without an overall score until the gap is resolved. Do not redistribute its weight. A numeric score does not certify licensing, privacy or safety; material concerns remain part of the written verdict regardless of the result.
Score corrections: September 23, 2026
We recalculated the archive from its recorded criterion scores. This corrected arithmetic and rounding without changing the underlying judgements or claiming a fresh test. Each affected review shows its earlier result.
- HentaiPros: 5.0 → 5.1
- Nutaku: 7.0 → 7.1
- E-Hentai: 6.9 → 7.7
- nhentai: 6.6 → 7.8
- Promptchan AI: 8.5 → 8.6
- Candy AI: 7.8 → 7.9
- FAKKU: 6.2 withdrawn because the limited Free-access test lacks a complete criterion breakdown.
- Published
- The article's original publication date.
- Updated
- The page was revised; this alone does not mean the product was retested.
- Last checked
- The latest recorded verification of the article's product claims.
- Price checked
- When the stated plan or price was checked. Availability and regional prices may change.
Product screenshots
Interface evidence, not promotional artwork
When a review includes a product screenshot, it is a non-explicit interface capture made by MurasakiHub for a specific editorial claim. Each figure identifies its source, capture date, and access context; we crop it to the relevant feature and redact personal or sensitive information when needed.
These images illustrate what was checked, rather than advertise a platform or replace the review's written evidence. Product screenshots are kept out of article heroes, social preview images, and other promotional metadata.
- Editorial art
- Original illustration that sets the article's mood. It is not product evidence.
- Product UI
- Dated, self-captured evidence tied to one named claim and a recorded access level.
- Diagram
- An explanation of a choice or process, labelled as explanatory rather than tested.
What a review may not cover
A hands-on label does not by itself include paid checkout, cancellation, private account settings, customer support, long-term use, or every region. Check the article's test scope before assuming one of those was verified.
Links and recommendations
One clear next step
Current external links point to platforms covered in the relevant article and are editorial. They are kept separate from the review itself so a reader can continue with the article, compare alternatives, or leave for an official site without repeated prompts.
Paid placements and affiliate links require a visible disclosure beside the call to action and a sponsored link attribute. A commercial relationship does not change the review score. Partnership rules and contact details are kept on the partnerships page.
Updates and corrections
Keep the record useful
When material information needs correction or clarification, the aim is to update the relevant page and keep its recorded dates and evidence scope clear. We do not rewrite the archive to make older work appear more thoroughly verified than it was.