You're a customer who comes across an intriguing peptide on the box of your dental care application and decides to look it up on Google. Google's "helpful" AI assistant suggests it's a potent anti-aging peptide for the skin based on a claim from a highly ranked, yet "fluffy" website. Convinced, you apply a dab of preparation to your face, hoping to eliminate a wrinkle. Instead, you get a painful rash. This is not a hypothetical scenario. It's a real problem caused by the fluff that now floods the Internet.
The Internet has long been filled with misleading information about cosmetic ingredients. Now, the problem becomes even worse with the arrival of text-generating AIs (LLMs - large language models of the AI type) that lack limitations concerning basic checked facts necessary for generating text content. In short, an AI can generate a "fairy tale" about any cosmetic ingredient, and all it needs is just a name or a function like "Skin Conditioning" that in fact means nothing or may be everything.
What is the root of the problem?
On one hand, there are more than 30k ingredients listed in databases like CosIng. On the other hand, the well-documented part, accessible for free on the Internet, accounts for about 10% of its total of about 3k. What about the remaining ~90%? The lack of freely accessible information about cosmetic ingredients created a massive vacuum, which was filled with vast amounts of misleading information with no factual basis.
Some of the crafty programmers obtained the CosIng DB (database) in CSV (comma-separated values) format (which can be easily imported into any modern CMS), which was accessible for download, and created websites with a vast number of pages without much effort, but also without valuable data for customers. Search engines call it "Scaled content abuse", but instead of blocking, they readily index pages and award an enormous amount of SEO credits to those generated websites, such as domain authority, over time.
The problem becomes even more severe with the advent of LLMs. Due to competition, the token price for bulk text generation dropped by just a few cents per million tokens, allowing SEO companies to generate vast amounts of fluffy text using information from the CosIng DB and a simple automated AI prompt generator. So, we've encountered a "tsunami" of low-quality AI-generated fluff that SEO companies attempt to pass off as "professional-written" and valuable content.
Many search engines, including Google and Bing, as well as AI chatbots, struggle to promptly establish a protection mechanism against AI-generated content and AI loops (where AI cites fiction from other AI) that are misleading or deceptive. Google attempted to update its algorithms with a new standard called E-E-A-T, but it also failed, as SEO companies and cunning website owners rapidly falsified the E-E-A-T requirements. Falsifications reached such proportions that they even invented a new type of science, introduced a certificate from a foreign association as a diploma from an educational institution, and presented non-existent "professionals" or biographies of existing specialists who were unaware of the existence of their website, among other tactics.
Following Google's failure as a monopolist in the industry, the situation has become increasingly out of control. Today, searching for trustworthy information about cosmetic ingredients has become a matter of serious investigation, where users must somehow separate fact from fiction.
What is the solution?
As a physician, I prefer to treat the root cause rather than just the symptom. However, the systematic solution to this problem lies in the hands of search engines and AI companies, and it may be a long time before they consistently filter out the fluff and are not hesitant to output "I don't know" or "I have no checked data concerning it." In my humble opinion, the first company to solve the "fluff" problem will win the competition, gain (or save, in Google's case) leading positions, and restore trust.
In the meantime, relying on these flawed systems puts your right to be informed, as well as your health and beauty, at risk. That's why our team developed a "simple fluff detection algorithm" to empower our readers with a quick, easy, and practical solution.
A Simple Fluff Detection Algorithm For Cosmetic Ingredients
The algorithm comprises a dozen easy checks to evaluate the quality of an article or an AI-generated search result. It can be divided into two parts, with straightforward simple checks and advanced checks, in case of you need more information, or you're going to reference that article in your creative work.
Simple checks
1. Function Check: Is the only function listed for an ingredient a general term like "Skin conditioning"? If so, it's a red flag, as 'Skin Conditioning' is the most common phrase in the CosIng DB. See What is Skin Conditioning?
2. INCI Name Focus: Does the article only define the INCI name (e.g., "Hexapeptide is a peptide with six amino acids") without explaining its specific function or action mechanism? If so, it's likely fluff generated with a program or AI. See How to generate a cosmetic ingredient website with a budget under $100?
3. CosIng Data Check: Does the content repeat or programmatically modify data from official sources like the CosIng database without adding new analysis or context? If so, it's likely fluff. See Is the CosIng helpful database for customers?
4. Rating System Check: Does the website give the ingredient a single, simple score or rating (e.g., "Good," "Bad," or a number) that oversimplifies its complex properties? If so, it's likely fluff. See Unified rating system for personal care ingredients: is it possible or is it a marketing bluff?
5. Topic Drift Check: Does the website or AI change the ingredient name if you input (search) the exact INCI name with phrases like "refers to", "same as", etc, and describes another ingredient? If so, it's likely fluff. See What is topic drift in cosmetic ingredients?
6. Non-existing Ingredient Check: Search on their website for "Hexapeptide-40" or "Octapeptide-30" and if there is a listing as a "skin‑conditioning agent", claim anti‑inflammatory and soothing properties, or that it is often used in skincare formulations. None of this is true; it is generated content, and it's a red flag.
Advanced checks
7. The "Jargon-Slang Mismatch" Check: Look for a jarring contrast between complex scientific terms and unprofessional, colloquial language. This is a sign that the author is not a true expert: the content is rewritten to trick Google or is AI-generated. If so, it's likely fluff.
8. Verifiable Author Check: Can you find a specific, named author with verifiable professional credentials and a clear professional history? Author information may be on the website's "About Us" page. If not, it's a red flag.
9. Language Check: Is the tone overly promotional, using vague and superlative claims instead of neutral, scientific language? You can also check if the website sells cosmetic products, but that ingredient is not included in their products. If so, it's likely fluff.
10. Factual Accuracy Check: Are there any claims that you know for sure are false? This includes misinformation, incorrect scientific classifications, or fabricated benefits. If so, it's definitive fluff.
11. Voice and Confidence Check: Does the author's voice sound overly confident, presenting every statement as an absolute fact without references or any nuanced or cautious language like "according to manufacturer claims" or "suggests"? If so, it's likely fluff.
12. Cross-Verification Check: Can a key claim made in the article be easily disproven with a quick search of a reputable, third-party source like an official government database? If so, it's definitive fluff.
What Can You Do?
You can also train a large language model (LLM) to act as your personal fluff filter. While no AI can be fully trusted to verify facts, it can be trained to recognize patterns of low‑integrity content.
Step 1: Save the Rules
Copy the complete 11‑point Fluff Detection Algorithm from this article into a chat with an AI assistant.
Step 2: Give Instructions
Tell the AI:
“I want you to act as a Cosmetic Ingredient Fluff Detector. Analyze any article, text, or search result I provide and rate it against the 11‑point algorithm. Give me a score and explain which points the content fails on.”
Step 3: Test and Refine
Feed it examples you already know are good or bad. Adjust your instructions to improve accuracy over time.
Your ability to identify fluff is not just a personal skill; it's a critical step in making the internet a more reliable place. When you use this algorithm to spot misleading information, you can take a decisive next step - STOP the fluff:
1. Report the Fluff: Many platforms and search engines have reporting or feedback mechanisms for low-quality or misleading content. Use them. By flagging these pages, you help train algorithms and human reviewers to identify fluff in the future.
2. Ignore the Fluff: It's the easiest way, by just ignoring (not clicking or spending your time on their website), you will make their website useless in the eyes of search engines.
2. Share a Reliable Source: If you find a misleading article, share a trustworthy resource article in the comments or on your social media. Sharing of truth helps redirect other users to accurate information and counters the spread of misinformation.
3. Spread the Word: Share this algorithm with friends and family members who are interested in cosmetic ingredients. Or better, make your own algorithm in a field where you're an expert. The more people who are equipped to spot fluff, the less effective these deceptive tactics will be. By joining our efforts, you are helping to create a healthier and more trustworthy online environment for everyone.
This way, you’re not just spotting fluff — you’re starving it, reporting it, replacing it, and teaching others to do the same. That’s how you turn detection into deterrence.