AI search engines do not always repeat a perfume's official note list faithfully. A new Scento study found that the engines tested sometimes added notes that do not appear in the official listing at all, and the rate at which they did so varied sharply from one platform to another.
The study tested ChatGPT, Google Gemini, and Perplexity on 100 perfumes from 45 fragrance houses on October 20, 2026. Each answer was compared against the published note pyramids in Scento's Perfume API database, so the test measured the engines against the official listings rather than against reviewers, retailers, or marketing copy. That design matters because a note pyramid is the closest thing a fragrance has to a canonical ingredient story, and shoppers treat it as the authoritative description of what a scent smells like.
For SEOs, the setup is a useful reminder of how AI answers should be judged. An answer can read smoothly and still drift from the source material, and the only way to catch that drift is to check it against a primary reference. In this case the reference was unusually clean, which makes the gaps the study found harder to dismiss.
Gemini added at least one note absent from the official listing in 49% of its answers, compared with 24% for ChatGPT and 25% for Perplexity. In other words, roughly half of Gemini's answers introduced at least one extra note, while ChatGPT and Perplexity did so in roughly a quarter of theirs. That is a meaningful gap between platforms answering the same questions about the same perfumes, and it suggests that shoppers asking different engines about the same fragrance may walk away with different impressions of what it smells like.
Completeness told a related but different story. ChatGPT returned every officially listed note in 82% of answers, followed by Perplexity at 78% and Gemini at 69%. So ChatGPT was the most likely to give shoppers the full official list, while Gemini was the most likely both to leave something out and to add something extra.
The study also applied a stricter test: reproducing the official note list exactly, with nothing added and nothing missing. ChatGPT and Perplexity each did so in 70% of answers, while Gemini did so in 43%. Across the 99 perfumes answered by all three platforms, all three matched the official note list exactly for just 39 fragrances. That last figure is the one worth sitting with, because it shows how rarely the engines agreed with the official listing and with each other at the same time.
For brands, the implication is not that AI search is unusable for product discovery. It is that the product facts reaching shoppers through AI answers are not yet stable, and the variation is large enough to change how a fragrance is perceived. A single added note can shift expectations, especially when the note suggests a sweetness, freshness, or warmth the official scent does not promise.
Scento cautioned that an unlisted note is not necessarily incorrect, and that caution deserves emphasis. A note pyramid is a published summary, not a complete chemical analysis, and a fragrance can carry nuances that the official listing does not name. An engine that adds a note may be surfacing a genuine impression, a widely repeated description, or a reasonable inference rather than inventing something from nothing.
The study was also limited to a single round of English-language testing in the United States, so its figures describe one snapshot rather than a permanent verdict on any platform. Engines change, listings change, and a repeat test could produce different numbers. The honest reading is narrower and still useful: on this test, on this date, the engines differed, and Gemini differed most.
The practical lesson for fragrance brands is to make the official note pyramid easy for both search engines and AI systems to find, quote, and verify. That means publishing the full note list on the product page itself, in plain text, rather than burying it in images or leaving it to retailers to state inconsistently. When the canonical version is clear and consistent everywhere, an AI answer has a better anchor to reproduce, and shoppers have a better place to check what they were told.
It also means monitoring how AI engines describe your products, not just how they rank your pages. If an engine consistently adds the same unlisted note to one of your fragrances, that pattern is worth knowing about, whether it turns out to be a fair description or a recurring error. The study gives the industry a starting point for that kind of monitoring, and a reason to treat AI product descriptions as a reputation surface in their own right.
Sources: Perfumer and Flavorist reporting on the Scento study, October 30, 2026.
Linkbaits helps you create content that earns citations and backlinks. Start with the free tools and build your first linkable asset this week.
Start free trial