Particle6 appeared in 37.5% of 64 AI-generated answers about Tilly Norwood and AI actors, more than any other named brand, yet ChatGPT and Claude did not name Tilly Norwood herself in a single one of their 32 combined answers. That gap, measured in the 5W AI Visibility Study on Tilly Norwood, is the clearest evidence yet that a brand's presence in one AI engine says nothing about its presence in another.
For marketers, the study is a working example of a problem bigger than one AI actor: a name can be well covered by press, well documented on Wikipedia, and still return zero mentions on two of the four major answer engines a customer might ask.
Why did ChatGPT and Claude return zero mentions of Tilly Norwood?
Gemini named Tilly Norwood in 8 of its 16 answers and Perplexity named her in 5, but ChatGPT and Claude returned none across their combined 32 answers. Each engine uses a different retrieval tool: Perplexity's native web search, Google Search grounding for Gemini, OpenAI's web search tool for ChatGPT, and Anthropic's web search tool for Claude. A brand that ranks in one engine's search index does not automatically surface in another's, because each system decides independently what to retrieve and quote.
The study's authors measured this with a fixed set of 16 prompts asked once per engine, so the 0-of-16 result describes this specific test, not a permanent verdict. The same prompt set can be rerun after a brand publishes new source material, turning the absence into a measurable target rather than an assumption.
Why does Particle6 lead the tally without "winning" the category?
Particle6's 37.5% appearance rate carries a 95% confidence interval of 26.7% to 49.7%, and that range overlaps the intervals for Synthesia, Xicoia Ltd., Tilly Norwood, and SAG-AFTRA. When intervals overlap this much, a study built on 64 answers cannot distinguish a real gap from ordinary sampling variation between the top five names.
This matters for how a brand should read any share-of-voice number pulled from a small set of AI answers. A single-digit percentage-point lead in a 16-prompt pilot is a starting hypothesis to retest at larger scale, not a marketing claim to repeat.
Why did Wikipedia outrank every brand's own website as a citation source?
Wikipedia was cited in 40.6% of the 64 answers carrying retrieval evidence, ahead of every individual brand's own domain. Deadline reached an 18.8% citation rate through its own reporting, while Xicoia Ltd., Netflix, and Apple each had a 0% owned-domain citation rate in this measured set, despite being named in the answers themselves.
An AI engine choosing what to cite treats a brand's own website as one source among many, competing with reference sites, trade press, and video platforms. A brand's own pages have to be structured for extraction the way a Wikipedia entry is: a direct definitional answer near the top, named facts, and a format the retrieval tool can quote without editing.
How should a brand act on results like these?
The study's own recommendation is to treat every gap as a hypothesis to test, not a verdict to publish. For a name absent from an engine's answers on a specific question, the fix is a source page that answers that exact question directly, followed by rerunning the same prompt set to see whether the appearance rate moves. Kelly Carothers, Director of AI Search at 5WPR Public Relations, notes that brands who wait for quarterly reporting cycles to check AI visibility are working from stale data by the time they act, since engine behavior changes without public notice.
Why it works: ChatGPT, Gemini, Perplexity, and Claude retrieve and quote passages rather than rank pages the way a traditional search engine does. A page that opens with a direct, self-contained answer to a specific question is more likely to be the passage an engine's retrieval tool selects, because the tool is matching the user's question against text that already answers it, not scoring the page as a whole. The 5W AI Visibility Index has documented the same pattern across categories from luxury goods to consumer technology.
What does this mean for AI visibility measurement going forward?
A 16-prompt pilot with one trial per engine is a snapshot, not a trend line. The value of a study like this one is in the retest: publish source material against a measured gap, rerun the identical prompts, and compare the before-and-after appearance rate on the same denominator. That comparison, not the single snapshot, is what tells a brand whether its GEO work changed anything.
What is the main takeaway from the Tilly Norwood study?
The Tilly Norwood AI Visibility Study shows that appearing in AI-generated answers depends on which engine is asked and what source material exists to answer the exact question a customer types, not on how much press coverage a brand has accumulated. Marketers who want this kind of measurement built for their own brand can reach the GEO team at 5W Public Relations directly.
