If Walls Could Talk — Moderation Policy
Operationalises the non-adversarial principle in
mission.md. This policy governs what a review can and cannot say; the mission explains why.
Version: 0.2 (pre-launch draft — awaiting solicitor review) Jurisdiction: England and Wales Last reviewed: 2026-04-20
Status markers used below:
[TODO: user - ...]— blocked on a fact only the founder can supply (company details, emails, solicitor appointment)[TODO: solicitor - ...]— requires qualified legal review before publicationPre-launch release requires all
[TODO: ...]markers to be resolved.
1. Purpose and scope
This policy governs how If Walls Could Talk ("the Platform") reviews, approves, rejects, edits, and removes user-submitted property reviews. It applies to every review submitted through the Platform and to all moderation staff and contractors.
The policy is designed to preserve the operator defence under Defamation Act 2013, Section 5 and the Defamation (Operators of Websites) Regulations 2013 (SI 2013/3028). To keep that defence intact, the procedures in sections 4 and 5 must be followed exactly and all decisions must be logged as described in section 7.
1.1 Legal framework
This policy sits within the following UK legal regime. See legal-context.md for the detailed analysis.
- Defamation Act 2013, s.5 + SI 2013/3028 — operator defence; notice-and-takedown procedure in §5 below.
- Online Safety Act 2023 — IWCT is a "user-to-user service". Illegal-content duties (live 17 March 2025) require proportionate measures against priority illegal content (harassment, threats, malicious communications, hate offences, fraud, CSAM). The rejection criteria in §3 operationalise these duties.
- Digital Markets, Competition and Consumers Act 2024 Schedule 20 — fake-review prevention duty (live April 2025). The fake-review rejection code (R-FAKE) and authenticity verification in §6 operationalise this. CMA208 guidance applies.
- UK GDPR / DPA 2018 + ICO Content Moderation Guidance (Feb 2024) — content moderation is high-risk data processing requiring a Data Protection Impact Assessment (DPIA), transparency, minimisation, and fairness. Logging requirements in §7 are designed to meet the data-minimisation standard.
- Protection from Harassment Act 1997 + Malicious Communications Act 1988 — underpin the harassment / threat criteria in §3.3 and the coordinated-campaign rule in §3.7.
2. What a review must be about
Every review on the Platform must relate to the physical property: the building, fixtures, fittings, communal areas, noise from the structure (e.g. pipework, poor insulation), and maintenance responsiveness as it relates to the condition of the property itself.
A review must not contain:
- Any opinion or description of a named or identifiable person (landlord, letting agent, neighbour, previous tenant, or anyone else)
- Any personal data of a third party (full name, phone number, email address, photograph, or any other identifier)
Users are reminded of this scope by a confirmation screen before submission. This policy operationalises that commitment.
3. Rejection criteria
A review submitted to the moderation queue must be rejected, in whole or in part, if it contains any of the following:
3.1 Defamatory statements
A statement of fact (not opinion) that is false, published about an identifiable person, and capable of causing serious reputational harm — as defined by Defamation Act 2013, Section 1. Pure opinion clearly framed as opinion (e.g. "in my view the agent was unhelpful") is not automatically defamatory, but any statement that could be read as a false statement of fact about a person must be removed.
3.2 Personal information of third parties
Any content that identifies, or could be used to identify, a specific individual who has not consented to appear on the Platform: full names, partial names combined with other identifiers, phone numbers, email addresses, physical descriptions used as identifiers, social media handles, or vehicle registration numbers.
3.3 Harassment, threats, and hate speech
Content that threatens, intimidates, or harasses any person. Content that constitutes a hate crime or incites hatred on the basis of a protected characteristic under the Equality Act 2010.
3.4 Landlord or neighbour references
Any substantive comment about the behaviour, character, or competence of a landlord, letting agent, or neighbour — even if framed as a property observation. Factual statements about the property management process (e.g. "repairs took 12 weeks") are permitted provided they do not single out an individual by name or clear inference.
3.5 Fake or incentivised reviews
Reviews that the moderator has reasonable grounds to believe are not based on genuine tenancy experience: for example, reviews submitted by an account with no verified tenancy evidence, reviews that appear to be coordinated, or reviews submitted shortly after account creation targeting a single property. See section 6 for the verification procedure.
3.6 Illegal content
Any content that is unlawful under English law, including but not limited to: content that infringes copyright, constitutes a privacy tort, or violates the Computer Misuse Act 1990.
3.7 Coordinated campaigns
Multiple reviews from different accounts appearing to target the same property or landlord within a short window must be flagged for investigation, even if each individual review meets all other criteria. Coordinated campaigns can constitute harassment under the Protection from Harassment Act 1997 (a priority offence under the Online Safety Act 2023) and undermine the benchmark's public-good framing.
When a pattern is detected: (a) place all reviews in the cluster into "Held pending verification"; (b) require authenticity evidence from each submitting account (see §6); (c) restore only those reviews that pass a thorough check.
4. Pre-publication moderation queue
All reviews enter a moderation queue before publication. The queue is "approve to publish" — nothing goes live without an explicit moderator approval decision.
Standard review window: 2 working days from submission.
Moderators must record one of the following decisions against each queue item:
| Decision | Meaning |
|---|---|
| Approved | Meets all criteria; publish as submitted |
| Approved with edits | Minor redactions made (e.g. a name removed); publish amended version |
| Rejected — notified | Fails criteria; author notified with reason |
| Rejected — escalated | Potential legal issue; sent to legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] before any notification |
| Held pending verification | Authenticity unclear; awaiting further checks |
If a moderator cannot reach a decision within 2 working days, the item must be escalated to the senior moderator. It must not sit in the queue unpublished beyond 5 working days total without a logged reason.
5. Notice-and-takedown procedure (Defamation Act 2013, Section 5)
This section implements the procedure required by the Defamation (Operators of Websites) Regulations 2013 (SI 2013/3028) to preserve the Section 5 operator defence. Follow it precisely — any deviation can forfeit the defence.
5.1 Receiving a complaint
A takedown complaint must be submitted in writing (email is sufficient) to complaints@ifwallscouldtalk.uk. To be a valid "notice of complaint" under the Regulations, the complaint must include:
- The complainant's name and email address
- The URL of the specific review complained about
- The specific words or statements said to be defamatory
- The complainant's explanation of why the statement is false and defamatory
- Confirmation of whether the complainant consents to their identity being passed to the reviewer
A complaint that does not include all of these elements is not a valid notice under the Regulations and does not trigger the formal procedure. Acknowledge it, request the missing information, and restart the clock only once a complete notice is received.
5.2 The 5 working day response window
Once a complete and valid notice is received, the Platform has 5 working days to take one of the following actions:
Option A — Remove the review Remove the review within 5 working days. Log the removal (see section 7). Notify both the complainant and the reviewer (if contactable) that the review has been removed. The Platform retains its Section 5 defence.
Option B — Refer to the reviewer If the reviewer has consented to their identity being disclosed (or is identifiable to the complainant), pass the complainant's details to the reviewer within 5 working days. Log the referral. The complainant must then pursue their complaint directly with the reviewer. The Platform retains its Section 5 defence provided no further action is needed.
Option C — Identify the reviewer to the complainant Where the reviewer has not consented to disclosure, the Platform may still pass identifying information if: (a) a court order requires it, or (b) the reviewer has explicitly agreed. Do not disclose reviewer identity without one of these conditions.
Failure to act within 5 working days forfeits the Section 5 defence. If the 5-day deadline is at risk, escalate immediately to legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] — do not let it lapse silently.
5.3 Anonymous reviewer — cannot be identified
Where a review was submitted using the anonymous display option and the Platform cannot identify the reviewer from internal records (e.g. the account has been deleted and logs purged):
- Remove the review immediately upon receiving a valid notice.
- Log the removal, the reason ("reviewer unidentifiable"), and the complaint reference.
- Notify the complainant that the review has been removed.
- Do not attempt to reconstruct reviewer identity from incomplete data.
The Platform retains its Section 5 defence by removing the content. There is no obligation to identify an anonymous reviewer in order to keep the defence.
5.4 Counter-notification by reviewer
If the reviewer disputes the complaint after being notified, the Platform must:
- Pass the reviewer's response to the complainant.
- Make no further editorial decision on the disputed content for 14 days, to give parties time to seek legal advice.
- Seek legal advice before making any further decision if the dispute is unresolved after 14 days.
6. Authenticity verification
Where a moderator has grounds to suspect a review is fake or incentivised:
- Place the review in "Held pending verification" status.
- Send the submitting account a standard verification request asking for the tenancy year range and postcode (already collected at submission), plus any other reasonable corroborating detail the moderator specifies.
- If no satisfactory response is received within 5 working days, reject the review and log the reason.
- Do not publish the review while verification is pending.
The Platform does not require tenants to provide tenancy agreements or other documents at submission — but may request them during the verification step. Refusal to provide supporting detail when asked is grounds for rejection.
7. Logging requirements
Every moderation decision must be logged in the moderation record system with the following fields:
| Field | Required |
|---|---|
| Review ID | Yes |
| Property postcode (partial: e.g. M1 3) | Yes |
| Submission timestamp | Yes |
| Moderator ID | Yes |
| Decision (from section 4 table) | Yes |
| Decision timestamp | Yes |
| Reason code (see Appendix A) | Yes |
| Free-text notes | Optional but recommended |
| Complaint reference (if applicable) | Yes, if a Section 5 notice was received |
| Escalation flag | Yes, if escalated |
Logs must be retained for a minimum of 3 years from the decision date. They are legally privileged materials and must not be disclosed without legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] sign-off.
Moderators must never log reviewer personal data (name, email, IP) in the free-text notes field — reviewer identity is stored separately in the account database and linked only by Review ID.
8. The 30-day edit window
Reviewers may edit their published reviews within 30 days of original publication.
An edited review re-enters the moderation queue and must be re-approved before the edited version goes live. The original approved version remains visible until the edit is approved or rejected.
After 30 days, the review is locked. A reviewer who wishes to amend a locked review must contact support@ifwallscouldtalk.uk. The Platform may, at its discretion, unlock a review for editing if there is a legitimate reason (e.g. factual error identified by the reviewer).
Any edit that introduces content that would have been rejected under section 3 must be rejected. The original approved version is then reinstated.
9. Escalation contacts
| Situation | Escalate to |
|---|---|
| Potential defamation claim | legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] |
| Section 5 deadline at risk | legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] |
| Content involving a minor | legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] immediately |
| Law enforcement request | legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] and the founder (solo-founder stage; until a governance body is appointed) |
| Hate speech / criminal content | legal@ifwallscouldtalk.uk [TODO: user - appoint solicitor; alias currently forwards to support@ and is monitored] and the founder (solo-founder stage; until a governance body is appointed) |
Appendix A — Rejection reason codes
| Code | Meaning |
|---|---|
| R-DEF | Contains or implies defamatory statement |
| R-PII | Contains personal data of a third party |
| R-HARASS | Harassment, threat, or hate speech |
| R-SCOPE | Outside property scope (landlord/neighbour reference) |
| R-FAKE | Suspected fake or incentivised review |
| R-ILLEGAL | Other unlawful content |
| R-NORESPONSE | Verification request unanswered within 5 working days |
| R-LEGAL | Removed following valid Section 5 notice |