How would you feel if you want to find information about a particular thing, but the search engine or an AI chatbot returns a result about something related, not exactly what you want to find? It is frustrating, the system thinks it knows better what you need. There is a term, topic drift (also called intent drift), that perfectly describes the situation.
Topic drifts are common in the cosmetic ingredients field. Any search engine or AI system has an internal mechanism for correcting queries in case an orthographic error occurs in the keyword or question. However, the systems' evaluation leads them to overlook their own mistakes in processing queries, resulting in responses about other, even related topics. Moreover, feedback options such as "Did you mean" are inconsistently available to correct their possible mistakes.
The classic example of topic drift in the cosmetic ingredients field is "Pentapeptide-4", which is an "ancestor" of the famous mainstream lipo-peptide Matrixyl (Palmitoyl Pentapeptide-4; Palmitoyl-KTTKS). Sederma developed both peptides; however, Pentapeptide-4 was overlooked due to delivery issues at the time. More importantly, there is another lipo-peptide under the INCI designation Palmitoyl Pentapeptide-4, Palmitoyl-KTSKS (BB-Biont), with significant anti-acne and Matrixyl-like properties.
If you try to search "Pentapeptide-4" on Google, Bing, and most AI chatbots will perform a classic topic drift, returning results about Palmitoyl Pentapeptide-4's Matrixyl version and ignoring the fact that Pentapeptide-4 itself is used as a targeting peptide in modern, advanced peptide-based complexes, or there is another lipo-peptide with distinct properties.
What about a topic drift in website content?
Unfortunately, topic drifts are common on many websites that focus on cosmetic ingredients. It is a harmful practice that aims to capture as much traffic as possible, rather than providing accurate, precise, and helpful information.
If editors can't find information about a particular ingredient but know there is demand, they find a similar ingredient and rewrite the content, changing the ingredient name. Who cares about checking facts, quality, or accuracy? It is a related substance and should have similar properties.
The situation is even worse in the case of AI-generated content, which can even put a text about an unrelated ingredient in their names that have a similar base, e.g., output rewritten text about Oligopeptide-20 with the title Oligopeptide-177.
How dangerous is Topic Drift in the case of complex cosmetic ingredients?
The Illustrative Case: Synthetic Peptides Mistaken for Plant Extracts
I searched Google to find what is common between Hexapeptide-40 and Octapeptide-30, expecting a simple result based on their typical use as synthetic technical peptides (e.g., related to His-tags or specific sequence similarities).
Surprisingly, I received an AI-generated text composed of multiple false statements:
"Octapeptide-30 and Hexapeptide-40 are that they are bioengineered peptides derived from the Nicotiana benthamiana plant and are used in cosmetics for their skin and hair conditioning properties." (There are screenshots.)
This is a classic example of Google's AI drawing conclusions based on Topic Drift (or merging different ingredients) and misleading information from sources its algorithms treat as "authoritative" (high EEAT), leading to seriously misleading, false, and potentially generous statements.
How does the misinformation occur?
The error is the result of a chain of faults:
- Hexapeptide-40 (a synthetic peptide) experiences drift to the recombinant peptide, full INCI name: Nicotiana Benthamiana Hexapeptide-40 sh-Oligopeptide-1.
- Octapeptide-30 (a synthetic peptide) experiences drift to another recombinant peptide INCI name: Nicotiana Benthamiana Octapeptide-30 sh-Oligopeptide-2.
- High-EEAT websites declare the incorrect statements, making definitive, but false, claims: "This polypeptide complex, sourced from Nicotiana benthamiana," or "Nicotiana benthamiana hexapeptide-40 sh-oligopeptide-1 is a signal peptide derived from the Nicotiana benthamiana plant."
- The critical error: The AI mistakes the "expression host" (GMO, a plant or bacteria used as a bio-factory to produce the protein) for the "ingredient source" (a simple extract or botanical derivative).
- Google AI generalizes errors, concluding they represent the "truth" about the base ingredients (Hexapeptide-40 and Octapeptide-30), and makes absurd final statements:
- that both are derived from a plant,
- used in cosmetics,
- and have generic conditioning properties.
Conclusion: This case highlights the profound risks of topic drift. Moreover, Google relies on EEAT or other easily falsifiable signals and doesn't perform real content checks, such as fact-checking with an AI. A combination of those faults may pose a potential danger to users, leading to severe and sometimes irreversible consequences. There are many reports, but the case where the man replaced sodium chloride with sodium bromide with AI advice, experiencing chronic intoxication known as Bromism, is very typical.
What to do?
If you know the exact name of the ingredient (e.g., as listed on the packaging), but receive information about another one, provide negative feedback to the search engine or AI chatbot. In the case of a chatbot, you can also request that it correct its answer.
If you face topic drift in an article posted on a website, you already know that it is not a trustworthy source. Additionally, you can report the website to the search engine if you found it there, or if an AI chatbot cites it, ask it to avoid citing that website.