The three rulings the ASA published in late May were not extraordinary. Beiersdorf was pulled up for a Nivea Q10 anti-wrinkle campaign that conflated ingredient evidence with finished-product evidence. L'Oreal was held responsible for a Lancome serum ad that overstated the strength of a clinical study. 111SKIN was challenged on a "naturally improves collagen" claim that the ASA did not consider scientifically substantiated.
Individually, the rulings are the kind of thing every brand has seen before. Collectively, with the rest of the ASA's recent communications, they describe a new operating reality for UK beauty advertising.
The ASA confirmed earlier this year that its Active Ad Monitoring programme is now running at sector level. The system scans paid social ads, search ads, marketplace listings and brand websites at scale, using machine-learning models trained on the existing UK advertising code. When it identifies an ad that matches a non-compliant pattern, the ASA either opens an investigation or sends a guidance notice. The complaint inbox is no longer the front door. The software is.
For a brand at £500k-£5m, this matters operationally in three ways. The first is that exposure is now continuous, not occasional. The second is that the cost of an upheld ruling extends past the regulatory sanction into press cycles, retailer relationships and customer trust. The third is that the patterns the AI catches are exactly the kinds of claims beauty brands have built their content stacks on for years.
What the AI is trained to flag
The ASA has been clear, in its own publications and in interviews with the marketing trade press, that the monitoring system focuses on a defined set of high-risk patterns in beauty and personal care. None of them are surprising. All of them are common.
The first pattern is unsupported "clinically proven" or "dermatologically tested" claims. The ASA considers these unjustified unless paired with a methodology, a sample size, an outcome and an attribution. A claim of "clinically proven to reduce wrinkles" without a linked study, panel design or instrumental result is now treated as a candidate for investigation.
The second pattern is ingredient-versus-finished-product conflation. If a finished product contains hyaluronic acid and the ad implies that the product delivers all the benefits associated with hyaluronic acid in academic literature, the ASA reads the claim against the finished product, not the ingredient. The Beiersdorf and L'Oreal rulings both turned on this distinction.
The third pattern is mechanism overclaim. "Naturally improves collagen" was the specific phrase the ASA challenged 111SKIN on. The ASA's position is that a topical product cannot, on current evidence, "improve" endogenous collagen production in the dermis through over-the-counter cosmetic actives. The claim implies a physiological mechanism the product cannot demonstrate.
The fourth pattern is implied medical benefit. "Reduces acne", "treats rosacea" and similar phrasing trip both ASA and MHRA jurisdictions. Beauty brands often use them in influencer content, where the original claim was made by the creator and amplified by the brand. The brand is now responsible for the creator's content as if it had written the ad itself.
The fifth pattern is misleading social proof. Ads that show before-and-after images without disclosing post-production, lighting changes, makeup application or product positioning are increasingly being flagged. The ASA has signalled that AI-generated or AI-enhanced before-and-after imagery is a particular focus this year.
Why "no complaints" used to be a defence and is not now
Until recently, brands running aggressive claims could survive on the simple maths that the ASA processed only a few thousand investigations a year against tens of millions of ads. Most ads sat below the noise floor. A complaint had to arrive before an investigation began.
The AI monitoring system changes that maths. The ASA is now sampling ads continuously rather than waiting for them to be surfaced. Whether your ad gets caught is no longer a function of whether a competitor or consumer complains. It is a function of whether your copy matches the patterns the system is trained to detect.
The practical implication is that the entire portfolio of live ads is now exposed simultaneously. The risk on any single ad has not gone up. The probability that at least one ad in the portfolio is investigated has gone up significantly.
For a £500k-£5m brand, the asymmetry is brutal. One upheld ruling costs you a press cycle, an awkward conversation with at least one retailer, and a permanent search result that anyone diligent will find when researching the brand. None of that is recovered with a quiet apology.
The 14-day claim audit
A practical claim audit at this size sits between legal review and creative review. It does not need a barrister. It does need someone who reads carefully and who can hold the line on three specific calls.
Pull every active ad across paid social, Google search, Meta and TikTok into a single document. Add every email sent in the last 90 days. Add the body copy from every PDP for the top 20 SKUs. That is the surface area to assess.
Read for the five patterns above. For each match, decide which of three buckets it sits in. The first bucket is "remove": the claim is unsupported and the simplest move is to rewrite it. The second is "substantiate": the claim is defensible but the substantiation has never been written down in a form that would satisfy a regulator. The third is "restate": the claim is interesting but overcooked, and a softer version captures the marketing benefit without the exposure.
For everything in the second bucket, build a short evidence file. It does not have to be peer-reviewed. It does have to exist. A clinical study, an instrumental panel, an ingredient white paper or a consumer test, attached to the claim it supports, with a methodology summary and an outcome.
For the third bucket, rewrite the claim in language that does not imply mechanism or treatment. "Improves the appearance of" instead of "improves". "Feels firmer" instead of "is firmer". These are not weasel words. They are the legally distinct claims the ASA recognises as cosmetic rather than medical.
Reissue the rewritten ads, archive the originals with a date stamp, and document the audit. If you are ever investigated, having a documented claim review process from the last 14 days is the strongest position you can be in.
The shift in mindset
The brands that get caught in this enforcement wave will not be the ones running the most aggressive claims. They will be the ones running ordinary claims with no evidence file behind them. The category has spent five years inflating its language. The infrastructure to defend that language has not caught up.
The brands that come through the wave intact are the ones that treat claim work as part of the marketing operating model, not as a compliance afterthought. They write claims they can defend, they update claims when the evidence changes, and they have a person whose job includes saying no to a campaign that has gone too far.
The ASA's software is patient. It is also faster than any complaint inbox ever was. Run the audit this fortnight, before the audit runs you.