PromptRadar finds AI platforms favor men, U.S.-based experts and stale data

PromptRadar’s 2026 AI Spokesperson Index analyzed more than 18,000 AI responses across ChatGPT, Perplexity, Gemini and Google AI Overviews and found major gaps in gender balance, geography and accuracy. The findings suggest AI answers are shaping reputation with a narrow and sometimes outdated set of experts.

Why it matters: - AI platforms are becoming a new gatekeeper for expert visibility, which can affect who gets cited, trusted and contacted. - The study found a strong gender imbalance, a U.S. bias and frequent use of outdated job information in AI-generated recommendations. - PromptRadar says communications teams now need to manage AI visibility as part of reputation strategy.

What happened: - PromptRadar published the AI Spokesperson Index 2026 on Oct. 12, 2026. - The study analyzed 18,348 responses to 1,529 prompts across ChatGPT, Perplexity, Gemini and Google AI Overviews. - The analysis covered expert mentions across nine industries. - The full report is available in the AI Spokesperson Index 2026.

The details: - Women accounted for 21% of the 25,029 expert mentions returned by the AI platforms. - In a benchmark sample of trade-press coverage from the same industries, women accounted for 28% of named expert quotes. - The gender gap was widest in defence, where women made up 11.9% of expert mentions. - Cybersecurity also showed a large gap, with women at 17.3%. - Healthtech had the highest share of women at 37.9%. - When prompts did not specify a region, 74% of the experts named were based in the US. - Even on UK-focused prompts, 28% of the people named were US-based. - For Middle East-focused questions, only 45% of the experts named came from the region. - 16.9% of expert mentions included a role or company that was no longer current. - Defence and AI had the highest shares of outdated information, at 22.2% and 20.5%. - Perplexity produced the lowest share of outdated roles or affiliations, at 9.8%. - Across nine industry top-20 lists, the average overlap between AI and trade press was 37%. - Of experts quoted at least three times in the six-month trade-press sample, 58% were not named by any of the four AI platforms. - Only 38.4% of names returned consistently across all three runs. - Just 9% appeared on all four platforms for the same prompt. - Perplexity surfaced the widest pool of names and the highest proportion of women. - Google AI Overviews drew on the narrowest and most repetitive pool. - Gemini was more likely to answer with organisations or general advice instead of naming an individual.

Between the lines: - The findings suggest AI systems are not drawing from a stable list of widely recognized authorities. - Visibility in traditional media does not automatically translate into visibility in AI answers. - The study also implies that AI outputs can reinforce older biases if the underlying training or retrieval signals are narrow. - A PromptRadar spokesperson said, “The shortlist matters as much as the answer.” - The spokesperson said communications teams need to know whose expertise AI can see, whose it leaves out and whether the information is still accurate.

What’s next: - PromptRadar recommends machine-readable spokesperson profiles, consistent titles and affiliations across company websites, LinkedIn and biographies, and region-specific evidence for local experts. - PromptRadar also recommends tracking AI visibility regularly instead of relying on a single response. - Kamal Leadbitter, Marketing Manager at PromptRadar.ai, said organisations that want more accurate and broader expert visibility need to make those experts clearly identifiable and consistently cited. - The company has published an interactive data explorer alongside the industry scorecards in the full report.

The bottom line: - AI-generated expert recommendations are influential, but the study shows they remain skewed, repetitive and sometimes out of date.