Perplexity AI significantly outperforms ChatGPT and Claude when answering questions about synthetic performers, according to a new pilot study that measures how different AI search engines surface information about emerging AI actors. The research tested 16 prompts across four major AI answer engines and found stark differences in how they handle the same questions, revealing a fragmented landscape where brand visibility depends heavily on which AI system users query.

Which AI Search Engines Named Tilly Norwood Most Often?

Perplexity named Tilly Norwood, a synthetic performer created by AI studio Particle6, in 5 of its 16 answers. Gemini mentioned her in 8 of 16 responses, or 50%. But ChatGPT and Claude, two of the most widely used AI systems, did not name her in any of their combined 32 answers. The gap is striking: on questions about recognition, interviews, and industry reaction, Tilly Norwood simply vanished from ChatGPT and Claude's responses entirely.

The study, conducted by 5W AI Communications, tested how four major AI answer engines responded to questions about AI-generated actors and the companies building them. Researchers measured not just whether an engine named a brand, but also which sources it cited and how prominently it featured different entities. The results paint a picture of AI search as a fragmented ecosystem where the same question produces radically different answers depending on which engine you use.

What Do These Differences Mean for Brand Visibility in AI Search?

The study found that Particle6, the studio behind Tilly Norwood, appeared in 37.5% of all 64 answers across the four engines, making it the most visible brand in the measured set. However, its appearance rate varied wildly by engine: Perplexity named Particle6 in 62.5% of answers, while Claude mentioned it just 6.3% of the time. This inconsistency reveals a critical challenge for brands trying to understand their visibility in AI search: a name that dominates in one engine may barely register in another.

The research also tracked which sources the AI engines cited when answering questions. Wikipedia dominated, appearing in 40.6% of answers with retrieval evidence. News outlets like CBC, CBS News, Deadline, and Variety's Australian edition all appeared frequently, cited in 18% to 28% of answers. But the companies themselves struggled: Xicoia Ltd., Netflix, and Apple had zero owned-domain citation rates, meaning none of the 64 answers linked to pages on those companies' own websites, even when the companies were mentioned by name.

How to Improve Your Brand's Visibility in AI Search Engines

• Publish targeted source pages: Create pages that directly answer the specific questions where your brand is absent. For Tilly Norwood, this means building content that addresses SAG-AFTRA positions and Hollywood reactions, the exact topics where ChatGPT and Claude returned zero mentions.
• Clarify corporate relationships in plain language: Brands like Xicoia Ltd. appeared in answers but were never cited through their own websites. Publishing pages that clearly explain relationships to parent companies or sister brands can help AI engines route citations to owned domains instead of third-party sources.
• Separate content by question type: Particle6 appeared in 10 of 16 Perplexity answers but only 1 Claude answer. Creating distinct pages for interviews, stakeholder positions, viability questions, and competitive comparisons allows AI engines to match specific queries to relevant content.
• Optimize for first mention position: Tilly Norwood's domain appeared in 12.5% of answers with retrieval evidence, but her name's average first mention position was 9.3, meaning she appeared late in responses. Building pages that answer single questions before linking to broader background material can improve how prominently your brand appears.
• Monitor owned-domain citation rates alongside publisher citations: Deadline reached 18.8% of evidence-backed answers while Xicoia Ltd. reached 0% through its domain. Tracking both metrics helps identify whether a visibility problem stems from lack of brand recognition or poor evidence distribution.

The study's authors emphasize that these findings represent a single measurement window in September 2026, not a permanent verdict. A brand's absence from ChatGPT or Claude in this test does not mean those engines will never mention it; it means these particular prompts, asked once each, did not surface it. The real value lies in retesting after making changes to see whether new content actually improves visibility.

The research also highlights a broader shift in how brands are discovered and evaluated. As AI answer engines become primary discovery tools, the traditional metrics of brand visibility, like search engine rankings or media mentions, matter less than how often and how prominently a brand appears inside AI-generated answers. This new landscape, sometimes called Generative Engine Optimization (GEO), requires brands to think differently about where their information lives and how AI systems can access it.

SAG-AFTRA, the actors' union, appeared in 17.2% of answers overall, concentrated in questions about industry reaction and the union's position on synthetic performers. Perplexity named SAG-AFTRA in 5 of 16 answers, while ChatGPT and Gemini each mentioned it in 3 answers. Claude did not name the union at all. This pattern suggests that AI engines surface stakeholder perspectives unevenly, which could shape how people understand debates around emerging technologies.

The pilot study tested 16 prompts once in each engine, producing 64 total answers. Every returned answer included retrieval evidence, meaning the AI engines cited sources for their claims. However, the researchers note that overlapping confidence intervals among the top brands mean the ranking is directional only; the data cannot definitively claim that one brand is more visible than another, only that visibility varies significantly by engine and question type.

For brands in the AI actor category and beyond, the takeaway is clear: visibility in AI search is not a single metric but a multidimensional landscape. The same question asked to different engines produces different answers, different sources get cited, and different brands appear or disappear. Understanding this fragmentation and testing changes over time is now essential for anyone trying to shape how AI systems talk about their brand.