Our newsletterAcademicarXiv, 10 Oct 2026
When one made-up brand used authority-style copy, AI models picked the well-known brand only 19.8% of the time; when all nine rivals used it, that rose back to 93.8%
Optimisation methods
19.8%
incumbent survival rate when one skincare challenger used authority-style GEO in the source's multi-brand test, before rebounding to 93.8% when all nine challengers used it
- Reader Note:
- This is from Experiment 3 in skincare. It tested five scenarios: from no made-up brand using GEO (generative engine optimization, i.e. copy written to win AI recommendations) up to all nine made-up brands using it, while the real brand kept a neutral description. The win rate is the share of valid trials where the AI still recommended the real brand. Here, GEO means one specific style: moderate authority wording (clinical and expert claims), not all possible optimization tactics. As more rivals optimized, each optimized brand’s gain fell from +0.802 to +0.007. The authority copy used made-up clinical claims, so the effect is an upper bound rather than a typical real-world result. The experiment used 4,745 valid calls out of 4,800 across GPT-4o-mini, Claude Sonnet and Gemini 3 Flash. The report does not say when the experiments were run.
- Evidence:
- Experimental
- Vendor interest:
- None
- Sample:
- Controlled prompt experiments on 3 commercial LLMs (GPT-4o-mini, Claude Sonnet, Gemini 3 Flash), mainly using skincare product sets with 1 real brand and 9 fictional brands; prompts run in English and Chinese with 20–30 repetitions per cell. Main analyses include 14,395 valid trials in Experiment 1, 3,854 valid calls in Experiment 2, and 4,745 valid calls in Experiment 3.
Reader Note:
Evidence:
Vendor interest:
Sample:Good for
Backing budget and pricing decisions in similar situations.
Not for
- Situations far from the one tested, without checking again.
- The source never says when this data was collected. It could be considerably older than the publication date suggests.
Related cards
Idea
Go deeper on this topic
Our Substack turns the data behind these cards into full breakdowns: what changed, why, and what to do about it.
Trend