Creative Touch Permanent Makeup and Aesthetics

In April 2026, Nature published an analysis confirming what many in the research world had suspected: 2.6% of computer-science papers accepted to major academic conferences in 2025 contained at least one potentially hallucinated citation. That is a near-tenfold increase from the year before. One journal editor reported rejecting a quarter of January 2026 submissions on the basis of fabricated references alone. ^[1]

This crisis has now reached health information on the wider web. AI-generated articles on skincare, treatments, and clinical evidence routinely appear in search results, many carrying citations that either do not exist, describe a different paper entirely, or have since been retracted.

We don’t think this is acceptable for health content, especially in a category like medical aesthetics, where readers may base real decisions about their body on what they read.

So we built a system that makes hallucinated citations structurally impossible for any reference used to support a clinical claim.

This page explains exactly how we do that, what sources we use, how we use AI, how we handle conflicts of interest, and what to do if you ever spot an error. We’re publishing it because readers and search engines deserve to know how our health information was created, not just what it says.

We do not trust citations. We verify them.

In plain English: how we verify every citation

Every academic reference that appears in our content goes through a verification pipeline before it can support any clinical claim. Here’s what that actually means:

1. We only accept real, traceable sources.
   Any citation must resolve to a genuine PubMed ID, DOI, PMC ID, NCBI Books accession, or authoratitive journal or website. If it doesn’t exist, it cannot be used to support a clinical claim.

2. We pull the official record directly from PubMed.
   Our system automatically fetches the full bibliographic details, abstract, official medical subject headings (MeSH terms), and crucially, whether the paper has ever been retracted, corrected, or updated.

3. Where possible, we examine the actual methods section.
   For open-access papers we download the full text and check how the study was really conducted, not just the abstract summary. Where a paper is not available under open access, classification relies on PubMed’s publication-type metadata and abstract.

4. We keep checking every week.
   Every reference in our system is automatically re-checked for retractions. If a paper we have cited is later retracted, our system flags it and we immediately review every page that uses it. References carrying no retraction flag are re-audited every 90 days as a further safeguard.

This is not standard practice across most health websites or even many academic journals. We built it because retracted or fabricated citations are now a documented public-health problem.

How we classify evidence

Not all research is equal, and we don’t treat it as if it were. References are classified against the Oxford Centre for Evidence-Based Medicine (OCEBM) hierarchy, the same framework used by NICE and other clinical guideline bodies.

In simple terms:

• Tier 1: Systematic reviews and meta-analyses (the highest level of evidence).
• Tier 2: Randomised controlled trials (RCTs).
• Tier 3: Cohort and case-control studies.
• Tier 4: Case series and case reports.
• Tier 5: Expert opinion and narrative reviews (the weakest form for clinical claims).

We also separately tag mechanistic studies (laboratory, in-vitro, or computational research). These are cited only to explain how something might work at a biological level, never as proof of clinical effectiveness.

Our language matches the strength of the evidence:

• Strong evidence (Tier 1–2) → “demonstrates”, “shows”, “confirms”.
• Moderate evidence (Tier 3) → “suggests”, “indicates”, “is associated with”.
• Preliminary or mechanistic findings → “preliminary evidence points to”, “has been observed in”.

Where evidence is uncertain or conflicting, we say so and present both sides. Where a tier classification cannot be automatically determined, the reference is treated conservatively and the associated claim uses hedged language until classification is complete.

Our editorial standards and independence

Who we are. Creative Touch is a family-run aesthetics clinic and beauty salon in Rotherham, South Yorkshire. Our content team is led by Zoe Bayliss (owner and aesthetician), with clinical input from our nurse Michelle, aesthetician Claire, skincare therapist Jessica, and SMP specialist Grace.

None of us hold academic research credentials and we’re not going to pretend otherwise.

What we do have is something most academic papers and large health websites lack: a fully automated, documented verification system that makes hallucinated or retracted citations structurally impossible. Combined with more than thirty years of hands-on clinical experience, this is the foundation for trusting our content.

What we cite. Our primary sources are peer-reviewed literature indexed on PubMed, official clinical guidelines from NICE and the MHRA (Medicines and Healthcare products Regulatory Agency), and reference works from the National Library of Medicine. We do not cite secondary health journalism, competitor content, social media, or press releases as evidence for clinical claims. When we reference manufacturer data for a specific product, we name the source and note that it is manufacturer-funded.

Commercial independence. We publish no sponsored content. We maintain one disclosed affiliate relationship: we sell Zinzino omega-3 products (BalanceOil, omega-3:6 ratio testing kits, and related supplements) and earn commission through our Zinzino affiliate account. We have no other affiliate relationships with ingredient or device manufacturers, and no undisclosed commercial relationships of any kind that could influence our scientific characterisations.

When we write about Zinzino products, or any other treatment or product we sell, we disclose the relationship clearly in the content. Our recommendations are based on clinical evidence and our hands-on experience with clients; we would recommend these products even if we earned no commission.

How we use AI. We use AI tools to assist with research briefing, evidence-gap identification, and citation discovery. However, no clinical claim is ever published without human verification against the original source. Our citation pipeline is deliberately designed so that a hallucinated citation cannot pass through it; it requires a live, resolvable identifier.

Evidence language. We match the strength of our language to the strength of the evidence (see above). Where no good peer-reviewed evidence supports a popular claim, we say so, even when that claim would benefit us commercially.

Our content process, from research to publication

1. A research brief is prepared using PubMed searches, our internal knowledge base, and relevant clinical guidelines. Content is anchored to a structured knowledge base of science entities — ingredients, biological processes, conditions, and treatments — each with their own citation records. This means related articles share a consistent, cross-referenced evidence foundation rather than being researched in isolation.
2. A first draft is written with every clinical claim explicitly mapped to its verified source citation.
3. The draft undergoes editorial review for accuracy, evidence-language compliance, counter-evidence, and brand voice.
4. Clinical framing is checked by a member of our clinical team.
5. On publication, the article’s citations are automatically added to our reference management system, where the weekly retraction-audit cycle begins.

We display the publication date and “last reviewed” date on every Lab article. Content is reviewed when significant new evidence appears or at regular intervals for fast-moving topics.

Corrections, retractions, and feedback

We get things wrong sometimes. When we do, we correct them promptly and openly.

• When significant new research emerges or we improve the content, we update the article and reflect the change by updating the visible “last reviewed” date on the page.
• If a claim is factually inaccurate or a cited paper is retracted, we update the text and add a dated correction note explaining what changed.
• We never make substantive changes without updating the “last reviewed” date.

If you spot an error (a factual inaccuracy, a mischaracterised study, a broken or retracted reference) please tell us. We want to know.

Contact us at editor@creativetouchrotherham.co.uk or via the website contact form. Please include the page URL, the specific claim or citation in question, and any evidence you have. We will investigate and respond.

A note on what this policy is not claiming

This policy describes our process. It does not claim that our content is equivalent to peer-reviewed research. It does not claim that our clinical observations carry the same weight as randomised controlled trials. It does not claim that our team’s practical expertise is interchangeable with academic or medical research credentials.

What it does claim is this: our content is built on a documented, automated, verifiable system for handling scientific evidence. The references we cite exist. Their retraction status is checked weekly. The studies are classified against a recognised evidence hierarchy. Our claims match the strength of that evidence. And if we get something wrong, we correct it openly.

In an information environment where hallucinated citations are now a documented public-health issue, we believe this is a reasonable standard to hold ourselves to.

Creative Touch Aesthetics Ltd is a private aesthetics clinic and beauty salon. Our Lab content is for educational purposes and does not constitute medical advice. For clinical concerns, consult a qualified healthcare professional.

Technical note: This policy corresponds to the following Schema.org trust properties, which are declared in the site-wide structured data: publishingPrinciples, correctionsPolicy, verificationFactCheckingPolicy, and ethicsPolicy.