Ahrefs created a fictional brand and tested its visibility across ChatGPT, Perplexity, Google AI Mode and Gemini. 96% of AI visibility came from branded searches. The experiment reveals how AI search actually discovers and ranks brands.
96% of AI visibility came from branded searches. If nobody is searching for your name, AI search will not find you.
Ahrefs did something clever. They created a completely fictional brand, built it a website, seeded it with content and backlinks, then tracked how it appeared across the major AI search platforms. The experiment answers a question every marketer is asking: how do AI systems actually discover and recommend brands?
The headline finding: 96% of the fictional brand's AI search visibility came from branded queries. People (or in this case, testers) searching for the brand by name. Not generic category searches. Not "best tools for X" queries. Direct brand searches.
That number should make every marketer rethink their AI search strategy.
Why it matters
The experiment tested four AI search platforms and each behaved differently. Google AI Mode was the most stable and predictable. It treated the fictional brand similarly to how traditional Google Search would, drawing from indexed pages and structured content. Perplexity was the fastest to discover and index the brand. It picked up new content within hours, not days.
Gemini was the most unreliable. It misidentified the fictional brand approximately 60% of the time, confusing it with other brands or generating incorrect information. ChatGPT fell somewhere in the middle, generally accurate but slower to update.
The implications for Australian businesses are direct. If you are investing in "AI SEO" or "GEO" (generative engine optimisation) as a channel, the Ahrefs data suggests the biggest lever is not technical optimisation of your content for AI crawlers. It is brand awareness. People need to be searching for you by name for AI systems to surface you reliably.
Of the fictional brand's AI search visibility came from branded queries, not generic searches
This inverts the traditional SEO playbook. In classic search, you could rank for category terms without strong brand recognition. A well-optimised page about "best accounting software Australia" could compete regardless of brand awareness. In AI search, the data suggests that brand recognition is the prerequisite, not the outcome.
What to do about it
Three things become clear from this experiment.
First, brand building is not just a top-of-funnel luxury. It is the primary driver of AI search visibility. If your marketing mix is 100% performance and 0% brand, AI search will not find you.
Second, structured data and clear entity markup matter more in AI search than in traditional search. AI systems are trying to build knowledge graphs of brands and their attributes. Make it easy for them. Schema markup, consistent NAP information, clear product descriptions and unambiguous brand naming all help.
Third, monitor which AI platforms are getting your brand right and which are getting it wrong. Gemini's 60% misidentification rate means that a meaningful share of AI-assisted searches are returning incorrect information about brands right now. If that is happening to your brand, you need to know.
Of the time Gemini misidentified the fictional brand, confusing it with others
The Ahrefs experiment is the most useful piece of AI search research published this year. It replaces speculation with data. The data says: if you want to win in AI search, make sure people know your name first. Everything else is secondary.