Editorial Policy

How aggregation works

Every 30 minutes our fetcher (HentaidudeBot) checks the public RSS and Atom feeds of our sources. It identifies itself honestly, obeys each site’s robots.txt including crawl-delay rules, and uses conditional requests so unchanged feeds cost the publisher almost nothing. For each new item we:

1. run the safety filter described below - items that fail are discarded before anything is stored;
2. remove tracking parameters and boilerplate, and detect duplicates of stories we already carry;
3. assign a section (Hentai, Anime, Manga, Games or Industry) and tags using written rules;
4. keep a short excerpt, download the story image and resize it for fast loading, and publish the page.

When several outlets report the same news, we publish it once and list the other reports under “Also reported by”, so you can compare coverage.

Choosing sources

We only aggregate established English-language news outlets, official publisher and studio channels, and long-running community news desks that publish a feed, plus a few official Japanese studio and publisher channels whose news we translate (see below). We prefer the outlet that broke a story over sites that rewrite other coverage. The full list, with what each source covers, is on our sources page. Publishers can ask to be added or removed at any time.

Excerpts, not articles

We never store or show a full article. A story page contains the headline, an excerpt of two to three sentences clearly marked as a quote from the source (or an original summary, below; for a Japanese source, a short translation clearly marked as ours), a credit, and a prominent link to the full report. Our goal is to send readers to the publishers, not to replace them - the “Read the full story” link is the most prominent element on every story page.

AI summaries and quality checks

For some stories we publish an original summary written with the help of an AI language model instead of the quoted excerpt. Every such page says so directly under the summary. The summary is a short news brief, not a replacement for the report: it leads with what is new, is kept to about a third of the source article, and is written in its own structure rather than retelling the article paragraph by paragraph. When we have them, a closing paragraph adds context from our own archive - earlier stories on the same studio, publisher or series, and which other outlets covered the news - stating only what those stories’ headlines and dates say, without guessing at links between them; when it would add nothing, it is left out. The model sees the source’s report and those headlines, nothing else, and its output must pass automated quality checks before it is shown:

• Length: at most 250 words, and the part based on the source never more than 40% of the source article’s length. A summary needs at least 120 words for a long report, or 80 words for a short one.
• Originality: no run of 12 or more consecutive words copied from the source, fewer than 20% of its five-word sequences shared with the source text, no direct quotes, and no sentence-by-sentence retelling in the source’s order.
• Faithfulness: every number, date and proper name in the summary must appear in the source report - or, in the context paragraph, in the archive stories it refers to. A context paragraph that mentions anything else is dropped.

An AI summary that fails any check is not published - the page shows the quoted excerpt instead. Our editors can also write or correct a summary by hand: text copied from the source is refused (and a hand-written summary later found to copy the source is replaced by the excerpt), the length and faithfulness checks decide whether the page is offered to search engines, and only a summary rewritten in the editor’s own words counts as the editor’s - a lightly edited AI summary keeps its AI label. Publishers that ask not to have their work processed by AI get excerpts only. Summaries can still contain mistakes: the source article is always the authoritative version, and we fix errors as soon as we learn about them.

Stories translated from Japanese

Some news - new adult anime releases in particular - is published only in Japanese, by the studios and publishers themselves. For a small number of official Japanese sources we write the English headline and summary with the help of an AI language model, and instead of a quoted excerpt the page shows a short translation of the opening of the original report, labelled “Translated from Japanese”. It is our translation, not a quote, and the original Japanese headline is shown next to it. A translated story is never published without its translation: when the translation is not available it waits, and when it keeps failing the story is dropped. The same safety filter checks both the Japanese original and our English text, and every number and date in the translation, headline and summary must appear in the Japanese report. Translations can contain mistakes: the Japanese original is always the authoritative version.

What we ask search engines to index

Pages that only quote an excerpt exist for readers browsing the site, and we ask search engines not to index them. We only ask to index story pages that add something of their own: a summary that passed the checks above, plus either coverage of the same news by at least one other outlet or a summary written by one of our editors. Background stories from our archive and a box of key facts taken from the same report are useful to readers but are not enough on their own, and we cap how many new pages we submit each day. We never claim another publisher’s story as ours.

Corrections and updates

If a source corrects or updates its report and the change appears in its feed, our story is refreshed on the next fetch and shows an “Updated” time. If you spot an error in a summary, a wrong section or a broken link, email contact@hentaidude.tv with the page address. We correct or remove the story, and significant corrections are reflected in the page’s updated time.

Zero tolerance: content involving minors

We never publish anything that sexualises minors or involves minors in adult content - no exceptions. This applies to real people and to fictional characters, whatever the art style or claimed age.

• Every incoming item is checked against a strict block list before it is stored: terms such as loli, shota, lolicon, shotacon, underage and school-coded terms in an adult context cause the item to be rejected automatically.
• When a story is summarized, the AI model is also asked whether it involves minors - or school-age or child-like characters - in a sexual or suggestive context, or is an adult work featuring them. If the model says yes, gives no clear answer, refuses, or its reply is blocked by a content filter, the story is removed. The model’s output is run through the same block list before anything is published.
• Rejected items are not published and their content is not kept. We record only a one-way hash of the link and the reason, so the same item is never fetched again.
• These checks apply to every single item from every source: whichever outlet sends such material, it is never published.
• If anything slips through, report it to contact@hentaidude.tv with “URGENT” in the subject. Such reports are handled before anything else and the story is taken down immediately while we review it.

If you come across child sexual abuse material anywhere online, report it to the NCMEC CyberTipline or the Internet Watch Foundation.

Reporting a problem

Corrections, broken links and general questions: contact@hentaidude.tv. Copyright and takedown requests follow the process on our content removal page. Removed stories are replaced by a notice and disappear from our feeds, sitemaps and search.