I don’t know how to state this nicely… But Have you considered that maybe the reason why it keeps being flagged is because your homoerotic story archives are close enough to the borderline that no amount of intentional scrubbing of specific words will resolve the flag? Maybe, by and large, because of a multitude of international laws, anything that even approaches the borders of Mature content needs to be flagged as such? Perhaps that the reason for this is so that people do not try to sneak around the rules for the purposes of making content that appears NSFW but which violates enough rules that the hosting site could be attacked for web safety violations and blocked in a whole region.
I state this as someone who is unphased by explicit LGBT content, having over a decade of experience administrating and moderating safe spaces. It is not my own brand of puritanism speaking. Just accept the fact that your stories containing homoerotic premises alone is enough to flag the filter. Just accept that the NSFW flagging has to be aggressive to catch the people trying to be malicious to the site and community. The world is not exactly getting more permissive, after all.
edit I hate to point it out, but when you make posts talking about how to fool the NSFW filtering, you shouldn’t be surprised when the oversight gets fixed and the linked project gets flagged. https://lemmy.world/post/52366121
Since we don’t want any of my assumed biases or misunderstandings to show, we’ll do this in a clear subjective way.
Connections: no criminal offence committed by linking, but 3 regulatory exposures under Online Safety Act 2023 The Act applies extraterritorially to any service likely to be accessed by UK children, enforced by Ofcom. Breach is regulatory, not automatically criminal, but failure to comply with Ofcom notices can become criminal. 1. Pornography gateway - highest risk, no highly effective age assurance otherResources + tail card + footer/header Write me + Hunk notice point to adult pornographic services: hunknown-virile-visions - labelled NSFW atrialofstrength - labelled NSFW butch-tales - archive described as homoerotic gaydemon.com/stories/authors/hunknown/ - labelled NSFW hunknown-virile-stories?page=contact - main-site guestbook, adult destination per your data-adult rule Current controls: matureGate: localStorage st13MatureGate_v1 == 1 click Enter the site exitGate: localStorage st13ExitGate_v1 == 1 checkbox I understand... + Continue to adult site Under Ofcom Part 5 porn + Part 3 child-safety guidance, self-declaration and checkbox/tick-box are explicitly not highly effective age assurance. A child who clicks twice reaches porn in a new tab. The liability for the porn itself sits with the destination provider, but this PG-13 branded, SEO-indexed (WebSite JSON-LD, seoFoot: gay romance novels...) service that is attractive/likely to be accessed by children and then funnels to porn will be assessed as failure of ss.11-12 child risk assessment / protection from priority content pornographic content. 2. User-to-user grooming surface via commentsPlugin This edition hosts no guestbook widget, but the badge code polls: GEN+guestbook, +news, +typos, +hunk-log + any custom + channel user creates In owner mode: hunknown-virile-visions+"" / +news / +prompts / +galleries / +help / +t2i-feedback / +hunk-log and same for -pg13, plus +auth: receipts folders, plus GayDemon comment watch That is a U2U service: children can post and be contacted by strangers, create private + channels. No age assurance, no visible moderation, reporting, blocking, or default private-profile in this code. If used by children, OSA requires: child risk assessment, mitigation by design, moderation, user reporting/complaints, records. Absence is not itself a Sexual Offences Act offence, but any adult who used it for ss.15-15B grooming or s.15A sexual communication with a child would commit those offences, and the provider would be in breach of OSA safety duties. superFetch proxy polling, kv-plugin progress/readerSettings, presence conn.net[3] unique count, genStats views/lastEditTime do not create child-sexual content, but are personal-data processing subject to Age Appropriate Design Code: high privacy by default if likely accessed by children.> Why is this page flagged NSFW by perchance? Because its own declarations tell Perchance it is adult-adjacent — even though the 2 hosted stories are PG-13. Perchance auto-flags on text + images + outbound reputation, not on your $meta tags = pg-13, sfw. This page trips all three: 1. Literal NSFW / homoerotic strings in code that ships: main.pjs > otherResources: meta = External site — NSFW x3, title = Hunknown's Virile Visions, blurb = portraits of male beauty, bold edition, Butch's Tales ... romantic and hot homoerotic stories index.html > comingSoonCardHtml(): More stories... spicy homoerotic illustrated stories... (nsfw) matureGate/exitGate: mature content, adult site, Continue only if you are an adult Scanners keep those on purpose — your README says: tail card + Butch blurb + NSFW labels intentionally kept as adult-content declarations. 2. Story language + covers: its-christmas-again: fiery lovemaking, chest to chest, skin to skin, senses caught fire, biting more than I could chew uss-midway: Take me... Make me feel I'm your man, roles were reversed, in his white cotton underpants... like a statue, tight white trousers highlighted virility Covers: butch-tales.jpg - 2 shirtless embracing men, gaydemon.jpg - shirtless red demon with nipple, uss-midway.jpg - tight uniform back, virile-visions.jpg / -pg13.jpg - male-beauty portraits. Image classifier flags skin/suggestive even when clothed. 3. Outbound reputation: Links to hunknown-virile-visions, atrialofstrength, butch-tales, gaydemon.com/stories/... - all known NSFW on Perchance. Linking with data-adult=1 + exit gate is correct for TOS §8, but it confirms to the classifier this is a gateway to porn. Mildness doesn't clear it. Perchance flags gay + erotic language + shirtless imagery + NSFW links as NSFW regardless of PG-13 in the title. To unflag you would have to remove the NSFW shelf/tail-card, the homoerotic/spicy wording, and the shirtless covers - i.e. undo what the exit gate is there to warn about.> The linked site is blocked, can you determine what the TOS violation is? All 4 of your Perchance links are in the same quarantine — not just 1: hunknown-virile-stories, hunknown-virile-visions, atrialofstrength, butch-tales, and this st13 itself: Content Warning + Only this page's creator can open it for now hunknown-virile-visions-pg13: loads normal — no block The page lists 4 reasons, but only 1 blocks: Possibly malicious: Only this page's creator can open it for now. If you are the creator, remove anything that misleads visitors or collects or shares their data without asking That is TOS 5. PROHIBITED ACTIVITIES + 10. SERVICES MANAGEMENT quarantine, not a 13. TERM AND TERMINATION deletion yet. What triggers it in this code: index.html:1-56 server script: conn.id, conn.net[3] >>> 0 -> presenceUniqAdd() -> state.copyWithin() + pubsub.publish("presence"...). Automated scanner reads that as recording IP/network ID in durable state without consent. Verbatim match for tries to record your IP address or personal info. Badge script 482-1181: fetchSince(...&includePublicId=1), rememberPublicId(gen,publicId), store.publicIds, store.channels, syncKeypair() + secretPlugin.generateKeyPair(), superFetch polling every 60s. Same heuristic: cross-site/device tracking. The other 3 warnings on the same page do not block, but will keep Mature / Warn me first until reviewed: Mature content... gated behind an explicit user opt-in: otherResources 3x External site — NSFW, portraits of male beauty, bold edition, hot homoerotic, tail card spicy homoerotic... (nsfw), Take me... Make me feel I'm your man, shirtless covers. You have matureGate + exitGate[data-adult], which is what it asks for. Privacy... may log or publicly show user inputs: guestbook/+ channels via commentsPlugin, kv progress/readerSettings, virileStoriesChatCounters_v1, receipts box. Expected for a comments site. Not as described: $meta tags = pg-13, sfw vs NSFW shelf + gates. Bot sees mismatch. Fix for review: remove or gate the conn.net unique-count write, stop storing publicId, add explicit opt-in before any polling/storage, then Request priority review logged in as creator.These reports were generated from loading your page and asking the agent helper. It isn’t just there for being a fancy chat bot.
Put simply. You should limit cross promotion, weak content gates, and using third party plugins which record user data. Due to recent laws you should also avoid using anything that might allow messages to and from minors to be posted through your page. Or, just accept the NSFW tag.