Every AI assistant reads a different web. I counted the citations in 141 answers
I put the same 47 buying questions to ChatGPT, Perplexity and Google AI Mode on the same day, and classified every citation across the resulting 141 answers. The engineering-relevant part of the result is not the ranking. It is how differently the three engines behave as clients of your site. Citations per answer. Perplexity 16.7, Google AI Mode 3.7, ChatGPT 1.6. Perplexity cited provider…
Three leading AI assistants—ChatGPT, Perplexity, and Google AI Mode—were each queried with the same 47 buying-related questions on the same day. Every citation across the resulting 141 answers was meticulously analyzed. The key finding: the three engines display markedly different behaviors when accessing your website. Perplexity cited provider websites the most, doing so 667 times across its 47 answers.
ChatGPT, on the other hand, cited them only 47 times, almost precisely once per answer—limiting the potential influence your site can have on its responses. This wide divergence highlights a critical engineering brief: the information your site provides can significantly impact the answers generated by Perplexity, but not by ChatGPT.
Against Perplexity, your page is actively engaged and its content is genuinely in play. However, with ChatGPT, selection happens before your page is even read, making on-page work less impactful. Notably, Google AI Mode referenced Google Maps in 23 out of 47 answers, suggesting a potential benefit from a verified business listing.
However, all 507 Maps links were of the generic form "google.com/maps/search/Name,+City," lacking actual listings. This misinterpretation could be avoided by scrutinizing link structures before drawing conclusions based solely on hostnames. An unexpected revelation during the examination process was the behavior of JSON-LD structured data in Next.js applications.
When rendered through next/script, it does not appear in the server HTML at all. Instead, it is queued into self.__next_s and injected after hydration. Google's JavaScript execution eventually reveals it, whereas most fetchers behind assistant answers do not. For optimal results, use a plain script tag with type=application/ld+json and suppressHydrationWarning for JSON-LD markup.
This critical detail can make or break whether non-JS readers can access your structured data. The study's full methodology, tables, sample size considerations, and a section detailing what the data does not support are available at https://www.bandziuk.com/blog/ai-assistant-recommendations-study.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
