A theatre in Barcelona tried to buy an advert for a play. The play is Una habitación propia, Clara Sanchís's decade-running stage adaptation of Virginia Woolf's 1929 essay A Room of One's Own. It has been touring without incident since 2016. This September it played Teatre Raval, a 193-seat venue in the Raval district, for a two-week run from the 16th to the 27th.
Meta rejected the ad on both Facebook and Instagram. The stated reason, quoted directly from the rejection notice: the promotion appeared to be about "social topics, elections or political issues in the European Union", which "infringes our advertising standards". When the theatre pressed Meta for an explanation, the answer, again quoted directly, was that "any mention of civil rights, feminism or social reform" are "usually classified as 'social topics' by our automatic system" and filtered out accordingly. Meta's suggested resolution was to remove the "key words" so the copy would no longer appear to encourage activism.
Woolf wrote, in her diary, that she expected to be attacked as a feminist for what she had written. She did not anticipate being flagged by a keyword classifier ninety-seven years later for the crime of having written it at all.
This is a clean Theatre Pulldown case. The institutional claim, as stated in Meta's own 2019 policy announcement introducing the social-issues authorisation framework, is that the system exists to prevent advertisers from abusing targeting options and to protect the integrity of public discourse around elections and civic life. That announcement is explicit that the category list was "left intentionally broad" as a design choice, to be refined over time rather than narrowed at launch.
The operational reality, seven years on, is a system that cannot distinguish a Super PAC media buy from a poster for a one-woman show based on a Cambridge lecture series from 1928. There is no evidence anyone at Meta looked at this ad. There is only a classifier matching against a fixed vocabulary; "feminism", "civil rights" and "social reform" all trip it, regardless of what surrounds them.
The context that makes this worse than a misfiring filter: since 6 October 2025, Meta has not merely reviewed political, electoral and social-issue ads more strictly in the EU. It has banned the entire category outright, across all 27 member states, on Facebook, Instagram and WhatsApp alike. Meta's own announcement states plainly that "political, electoral and social issue adverts are no longer able to be delivered in the EU."
Meta attributes the decision to the EU's Transparency and Targeting of Political Advertising (TTPA) regulation, which entered into force on 10 October 2025 and requires platforms to label political ads, disclose who paid for them and how much, and obtain separate, explicit consent from each targeted user before using that user's own profiling data (interests, demographics, behaviour) to decide whether they see the ad, a higher bar than the general ad-personalisation consent that covers ordinary commercial advertising. TTPA does not itself ban such advertising; it regulates how it must be done. Meta's choice, stated in its own words, was that the obligations created "an untenable level of complexity and legal uncertainty", and it withdrew from the category entirely rather than build the compliance machinery, following Google's earlier exit from the same category in November 2024.
The practical effect is that there is now no authorisation process, no disclaimer, no graduated route of any kind by which a genuinely-classified "social issue" ad can run in the EU on Meta's platforms. Set against that, Meta's advice to Teatre Raval, strip the words that describe the play's actual subject matter, is not an appeal against an overcautious but still-navigable review. It is guidance on how to avoid being correctly identified as falling into a category that is now permanently and totally closed, for an advert that is not what the category was built to catch: paid political campaigning, not a monologue drawn from a set of 1928 university lectures staged at a 193-seat venue.
Worth stating plainly, too: the ad was rejected on Facebook and Instagram, separately but identically. These are commonly framed as two platforms; for advertising purposes they run on the same backend, the same classification pipeline, the same category list, and now the same blanket EU prohibition. The theatre did not get two independent readings and a split decision. It got one classification outcome, replicated automatically across both of Meta's major surfaces, with no fallback route inside the family of products it had access to.
The underlying mechanism, an automated classifier keyed to a broad vocabulary, is not EU-specific. Its expression differs by jurisdiction, and the difference is instructive precisely because the US still runs the graduated model the EU no longer permits.
In the United States, Meta's ad-standards documentation names "Civil and Social Rights" as one of the categories under proactive monitoring, explicitly covering freedom of religion, LGBTQ rights and women's rights. Ads in this category can still run, subject to identity verification and a "paid for by" disclaimer. The enforcement pattern there tends to surface as differential treatment rather than blanket rejection: the Center for Intimacy Justice's research, which prompted formal letters to Meta from US senators in 2022 and 2023 and a complaint to the FTC, found that adverts for women's health products (menopause, pelvic pain, menstruation, fertility) were routinely rejected as "adult content" while comparable products marketed at men were approved.
So the two symptoms are not the same failure at different severities. The US pattern is a classifier applying an available authorisation route inconsistently. The EU pattern, for this theatre, is a classifier deciding whether an ad falls into a category with no available route at all.
There is an existing legal term for the EU mechanism, and it predates this story by two decades. Seth Kreimer's 2006 paper "Censorship by Proxy" describes governments enlisting private intermediaries to restrict speech they could not or would not restrict directly. Jack Balkin's 2014 Harvard Law Review piece "Old-School/New-School Speech Regulation" extends this into the platform era as collateral censorship: a government does not need to order a platform to remove anything. It only needs to impose enough cost or legal uncertainty on carrying a category of content that the platform's rational response is to overblock, since overblocking is cheaper than building the compliance machinery, and the platform bears none of the accountability a state censor would.
That is close to an exact description of the EU case. TTPA never required Meta to reject a Woolf adaptation. It required consent, labelling and disclosure machinery for political advertising that Meta judged too costly to build, so Meta's rational response was to exit the category entirely and run a deliberately broad net at the edge of it. The theatre is not collateral to a government decision about Woolf; it is collateral to a government decision about political ad transparency, mediated through a private company's cost calculation. That is precisely Balkin's model.
The US women's health pattern is the same shape of harm, but the sourcing does not support claiming it by the same precise mechanism. Nothing in the Center for Intimacy Justice's findings or the Senate correspondence identifies a specific external liability trigger comparable to TTPA; Meta has never offered "avoiding a named regulation" as its stated reason for rejecting menopause or pelvic-pain adverts, the way it explicitly named TTPA for the EU exit. What can be said honestly is narrower: both patterns are automated systems built to manage a platform's own risk exposure, calibrated broadly enough that women's health and women's rights content absorbs a disproportionate share of the overblocking, in one case because a government made the category expensive and Meta chose to exit it, and in the other for reasons Meta has not disclosed and the record does not establish. Same demographic of speech paying the cost. Not provably the same mechanism.
The EU's Digital Services Act gives users of Very Large Online Platforms real, statutory redress rights over automated decisions. Article 17 requires a "clear and specific statement of reasons" for restrictions on content, including explicit disclosure of "the use made of automated means in taking the decision." Article 20 requires an internal complaint-handling system whose decisions are taken "under the supervision of appropriately qualified staff, and not solely on the basis of automated means." Article 21 escalates unresolved complaints to a certified, independent out-of-court dispute settlement body, at the platform's cost.
On its face, this looks like precisely the mechanism this case needs: Meta's own language frames the rejection as the output of "our automatic system", which is exactly the kind of automated restriction Article 17 was written to force an explanation of. Whether it actually applies here is a genuinely open question rather than a settled one. Article 17's text is anchored to restrictions on content already "provided by the recipient of the service", framed around removal, demotion or restriction of visibility, which most naturally describes content already live and then acted upon. An ad rejected before it ever runs sits in less-tested territory, and advertising specifically is also addressed by a separate provision, Article 26, which governs transparency rather than appeal rights. Whether a pre-publication ad rejection carries the same statutory statement-of-reasons and appeal obligations as a post-publication takedown does not appear to have been tested. It is worth an enquiry rather than an assumption either way.
It is tempting to read this as a Meta-specific problem with a Meta-specific fix: use a different platform. The reach numbers alone do not rule that out. Meta's own ad-planning tools put Facebook's estimated audience in Spain at 37.7 million and Instagram's at 29.5 million as of July 2026, per NapoleonCat's tracking; TikTok is the only other platform in a comparable order of magnitude, at roughly 30 million per GfK DAM's June 2026 figures. It's worth flagging that Meta's own tools carry an explicit caveat on these numbers: "Estimates aren't designed to match population, census estimates or other sources," and audience-inflation complaints against Meta go back to at least 2017, when the Video Advertising Bureau found Facebook's reported reach exceeded the actual population in every US state. Treat the reach figures as an order-of-magnitude comparison, not a headcount.
TikTok's own advertising policy, however, does not run Meta's broad "social issues" framework at all. Its ban is narrower and specific: candidates, parties, elected or appointed officials, ballot measures and legislative advocacy. The same policy explicitly permits eligible advertisers to promote "tourism, cultural, or entertainment initiatives." On the text of the policy alone, an advert for a stage adaptation of a 1929 essay is not obviously the kind of content TikTok's political-advertising ban was written to catch, in the way it plainly was for Meta's classifier.
That is a real correction to make, not a minor hedge: it would be wrong to claim every comparable platform is equally hostile to this specific ad. What can be said honestly is narrower. Nobody has actually tested whether this ad would have cleared TikTok's review; the claim above is what the policy text says, not what TikTok's own moderation would have done in practice, and moderation systems have a well-documented habit of diverging from their own stated policies. Google Ads runs a separate verification and restriction regime for political and issue content that would need its own scrutiny before any comparison is drawn. X has materially smaller reach in Spain than either Meta property. None of that adds up to a platform that reliably solves this problem; it adds up to an absence of evidence either way, which is a different and weaker claim than "every platform fails the same way," and the honest version is the one worth publishing.
What can be said with confidence is narrower and more specific to Meta: it is currently the only platform of comparable reach in Spain running both a total EU-wide prohibition on an entire ad category and a keyword-level classifier with no demonstrated contextual judgement deciding who falls into it. That combination, not "all platforms are the same," is the actual finding here.
The only genuine escape from this failure mode is the one this publication already runs on: owned channels and non-algorithmic federated reach, neither of which sell advertising space nor run a social-issues classifier against the publisher at all. The cost is real. A mailing list or a Mastodon following does not hand a 193-seat venue anything close to Facebook's reported 35 million potential eyes in Spain.
That is the trade-off this theatre is now actually facing, not an abstract sovereignty argument but a live operational one: reach, purchased at the cost of being subject to an opaque classification system with an EU-wide total prohibition behind it and a redress mechanism of uncertain reach in front of it, against control, purchased at the cost of nearly all of that reach. Digital sovereignty is usually discussed as an institutional or infrastructural question. This is what it looks like at the scale of a single theatre trying to sell tickets to a play about a woman needing a room, and the money, of her own.
Editor's note: this piece applies the analytical framework set out in "The Theatre Pulldown", examining the gap between institutional governance claims and operational reality.
The Sovereign Auditor covers digital sovereignty, cybersecurity governance, and data protection policy, with particular focus on Isle of Man jurisdiction and Crown Dependency issues.
Payments via PayPal. Credentials delivered by email. No Substack. No Stripe. No middlemen.
