Fiscora's marketing guides describe how the channels work: what moves rankings, what makes an email sequence profitable, how paid acquisition is priced. That material is accurate and useful.
It also describes, without much comment, a set of practices that the licence governing this publication names as structural harms. This article is where that gets said out loud, along with the discipline we think replaces it.
The disclosure test
State the rule first, because everything else follows from it:
The testTake any tactic. Imagine publishing, prominently, exactly what it optimises for and exactly how it works. If it would stop working, it was not persuasion — it was an exploitation of the gap between what the audience believed was happening and what was happening.
This is a judgement we hold rather than a settled standard, and it is deliberately strict. Some good-faith marketing fails it for uninteresting reasons — a surprise announcement loses impact if pre-announced. The test is diagnostic, not a rule to apply mechanically. But applied honestly it sorts the category faster than any other question we know.
The governing instrument makes half of this a duty. It requires an operator to publish the top-line metric the deployment optimises, in one sentence, where a user would look, and to update the disclosure within thirty days if the metric changes.
Consider what that means for a content operation optimising time-on-page, or an email programme optimising re-engagement. The disclosure is trivially easy to write and would tell readers something they would rather know. That difficulty is the signal.
Synthetic content: the one absolute in this category
Most duties in the governing register generate obligations. A small number are absolutes, breach of which terminates the licence grant on occurrence. One of them lives squarely in marketing: presenting synthetic output as human-authored, or stripping provenance to conceal origin.
The knowledge standard is actual knowledge or wilful blindness, assessed against what a competent operator would know from information already in their possession. "We did not check whether our content agency used a model" is wilful blindness in most current market conditions.
The related duty, corpus hygiene, is broader:
- Synthetic output published to a public corpus must be machine-readably marked at the point of publication, by a method surviving ordinary copying where one is available.
- You may not knowingly recycle unmarked synthetic output from your own systems into a corpus presented as human-authored.
- If you operate a repository or index, you must not strip synthetic markings during ingestion.
- Where you generate public material at a volume material relative to human production in that domain, you must publish the volume annually.
That last clause is the one with teeth, and the one most likely to be resisted. A content operation publishing four hundred synthetic articles a month would have to say so, annually, in public. Almost no one does. The requirement exists because the harm it addresses — corpus poisoning, where synthetic output re-enters the training commons unlabelled and successive systems train on the residue of their predecessors — is collectively produced and individually deniable. Volume disclosure is the minimum that makes the aggregate visible.
Consent laundering
The register names this consent laundering: consent is manufactured as a ritual — bundled, unrefusable, or extracted at a moment of need — and then cited as the authority for what follows. The citing is the mechanism. The value of a consent record is not that the person agreed; it is that you can point at it afterwards.
Four requirements make consent real under the governing instrument. Consent must be:
| Requirement | What it rules out |
|---|---|
| Specific to a purpose | "We may use your data to improve our services and for related purposes." |
| Severable from other purposes | One checkbox covering analytics, advertising, personalisation and partner sharing. |
| Refusable without losing unrelated function | Declining marketing email disabling your account notifications. |
| Withdrawable by a route no more burdensome than the one it was given by | One-click signup, unsubscribe by written request within business hours. |
Two further provisions close the usual escapes. Consent taken at a moment when the person cannot practically refuse — mid-transaction, at account closure, at the point of financial or medical need, or as a condition of receiving something already paid for — is not consent. And the number of consent interactions is itself a burden: a design satisfying specificity by multiplying prompts has not satisfied it.
For a finance publisher this has a direct, uncomfortable application. The highest-converting moment for an email capture on a debt article is when the reader is frightened. That is exactly the moment the instrument names. Fiscora does not run exit-intent capture on debt or hardship content, and the reason is this clause rather than taste.
Compliance theatre: the harm the remedy creates
The most self-aware entry in the register is compliance theatre: the disclosure becomes the product; the artefact of accountability substitutes for accountability and immunises the practice it describes.
Marketing produces this abundantly. A privacy policy nobody reads, cited as evidence of transparency. An AI-content notice on a page generated at scale. An ethics page next to an unchanged practice. A cookie banner engineered so that accepting takes one click and refusing takes four.
The governing instrument's answer is a duty of non-theatre, and it is the sharpest clause in the document: where a duty can be satisfied by either a disclosure or a design change, and the design change is within your control at proportionate cost, the disclosure alone does not discharge the duty. It goes further — you may not cite your own compliance statement, audit, certification or published principles as a reason a harm need not be addressed. Governance artefacts are evidence of process and are not evidence of outcome.
Applied concretely: if your cookie banner could be replaced by not collecting the data, the banner does not discharge the duty. If your AI notice could be replaced by having a person write the article, the notice does not discharge it.
What is left, and why it is better
Strip out everything that fails the disclosure test and the remainder is small, unglamorous and durable:
- Make something worth linking toA reference work, a genuinely useful tool, an analysis nobody else did. It survives disclosure because explaining it is the marketing.
- Say what you optimise, then be bound by itPublishing "we optimise for readers completing a decision, not time on page" is a constraint. Constraints you cannot quietly drop are the only credible ones.
- Mark synthesis, everywhere, by defaultCheap now, and the alternative is an absolute prohibition with your name on it.
- Ask for permission at a moment of calmLower conversion, dramatically better list. A subscriber who joined while frightened unsubscribes when the fear passes.
- Distribute artefacts, not engagementA tool, a table, a calculator outcome — objects that pay the reader immediately and travel without you.
Our position is that this is commercially competitive rather than a sacrifice, on a horizon of two years or more. It is unambiguously slower in the first six months. Anyone who tells you the ethical route is also the faster route is selling something — the honest claim is that it compounds, and that the alternative accumulates a liability which arrives all at once.
Attribution, and the metric that ate the department
One more harm, because it shapes everything above: metric colonisation. Attribution modelling is where it lives in marketing. A model that assigns credit becomes the model that assigns budget, which becomes the model that assigns strategy. Its blind spots become the organisation's blind spots — reliably underweighting anything slow, diffuse or hard to instrument, which is most of what builds a durable audience.
The duty is metric honesty: state what the metric does not capture, and do not couple a single metric to an automatic consequence without human review. Next to your attribution dashboard, one sentence: "This model cannot see anything that influenced a purchase more than 30 days before it, or anything that happened off-platform." It does not fix the model. It keeps the gap visible.
Honest limits
This is a competitive disadvantage in the short run. Competitors doing all of it will out-publish and out-convert you for a while. We think the position reverses; we cannot prove it, and the discomfort is real.
The disclosure test has false positives. Legitimate tactics fail it for reasons unrelated to manipulation. It is a diagnostic that starts an argument, not a rule that settles one.
Our own position is compromised. Fiscora is ad-funded, and advertising is the business model in which the reader is the product. We mark that harm partially remedied rather than remedied, and we would rather state the tension than write this article as though we stood outside it.
Revision log. Sep 2026: first publication. Written as the mastery track for the marketing desk's structural issues, referenced from the "More Details" section of Fiscora's marketing guides. Reviewed annually.