Starting in Q2 2025, TUYA Digital observed a repeatable pattern in GA4: visitors landing from AI assistants, such as ChatGPT, Perplexity, Copilot, or Gemini, spent significantly more time on the page than users coming from Google Search. This applies across various websites in the legal, tourism, e-commerce, and publishing sectors. In some cases, AI-sourced sessions were twice as long.
How come? Why has ChatGPT, a newly released generative engine as compared to the search giants, successfully outperformed Google in terms of time spent by users on the page?
We all know about the effect of AI on organic traffic, and we are aware that the latest drop was as much as 30% due to generative engines. However, the time spent on pages was one of Google’s strongestholds, which is now about to fall.
Why do AI assistant visitors behave differently?
Unlike classic search, generative assistants cite content only after analysing and summarising it. If your page is listed, it’s already been vetted as applicable. And the user clicking through isn’t guessing, they’ve already seen a summary and want more. These are high-intent readers, not casual scrollers.
The interface also changes behaviour. Most assistants open your page in a new tab or preview pane while the user continues chatting. That results in more time on page, deeper scroll, and stronger engagement signals.
A perfect match for Google’s July 2025 update
The content that earns AI citations overlaps closely with what Google’s July 2025 Algorithm Update now rewards. Structured answers, mobile performance, author clarity, and genuine expertise are at the heart of both. Thin, manipulated, or AI-spammed content has begun to fall. Experience-rich content has risen.
In this way, AI citation logic and Google’s new ranking logic are converging.
Case Signals from TUYA Digital’s Own Portfolio
The discovery around AI referral traffic wasn’t born from theory. It came from tracking real behavior across multiple client sites.
Between April and July 2025, TUYA Digital segmented AI assistant sources (chat.openai.com, perplexity.ai, copilot.microsoft.com, and gemini.google.com) across active GA4 properties in the legal services, tourism, e-commerce, and B2B SaaS sectors. In doing so, we began to see consistent engagement patterns that diverged from what we typically see from Google organic search traffic.
Here are a few representative examples that led us to investigate the trend further:
- In the legal vertical, an article by an international law firm on immigration law was cited by ChatGPT. Users who clicked through from Perplexity spent an average of nearly 4 minutes on the page, compared to less than 2 minutes for Google organic visitors. They were also twice as likely to interact with the contact form, suggesting a more serious intent.
- On an e-commerce platform focused on family care products, a product page began appearing in ChatGPT’s source citations. The engagement difference was immediate: users referred from ChatGPT sessions stayed on the page over 3 minutes on average and explored more product variants. Scroll and gallery interaction were noticeably higher.
- Another clear signal came from Visit Bucharest, where an article on things to do at night was cited in Microsoft Copilot results. Visitors arriving from that AI layer spent an average of more than 3.45 minutes on the page and were more likely to continue to the City Pass offer page, a valuable conversion step for those who know how to optimize for Bing.
- We saw similar behavior on a B2B SaaS site, where a long-form guide about OCR tools for legal professionals was cited in both Perplexity and ChatGPT. The average time spent on the page exceeded 6 minutes. Demo requests from these sessions more than tripled compared to the Google organic baseline.
These examples didn’t occur all at once. They surfaced gradually. But when analyzed together, the pattern became too consistent to ignore.
It wasn’t the traffic volume that stood out. It was the quality. The scroll depth. The engagement signals. And the fact that the AI platforms had already pre-selected these pages as trustworthy, comprehensive, or relevant before sending the user our way.
These aren’t one-off cases. They’re signals, and they’re growing stronger.
Yes, Google might still dominate the search traffic, in terms of sessions. However, in terms of trust and engagement, users tend to trust AI-generated answers more.
Why you need to track AI traffic separately
These users behave differently. They scroll more. Bounce less. Engage longer. But unless you explicitly segment AI sources in GA4, chat.openai.com, perplexity.ai, copilot.microsoft.com, gemini.google.com, you’ll miss the signal entirely.
Once separated, you’ll be able to identify which of your pages are being cited and why. That insight is more actionable than classic keyword data ever was.
We selected the Best Generative Engine Analytics Tools – it might get more and more useful these days for website owners worldwide, impacted or not by the fall of the classic search engines.
GEO is not the future – it’s already here
We’ve entered the age of GEO: Generative Engine Optimization. It’s not about tricking search. It’s about building content that’s so complete, clear, and trustworthy that AI assistants choose to cite it. The overlap with SEO is strong, but the quality bar is higher.
Your best-performing pages in the AI landscape are likely your most genuinely helpful ones.
What to optimize starting now
First, identify the pages already being cited. Strengthen them. Structure them. Make them skimmable without losing depth. Then review your analytics setup to make AI performance visible. Without that, you’re blind to one of the fastest-growing traffic sources in the ecosystem.
Finally, produce new content with intent. That means utilising genuine authorship, providing comprehensive answers, including firsthand insights, and eliminating unnecessary noise.
For those interested in optimizing their websites for generative search engines, such as ChatGPT, Perplexity, and others, we have prepared a detailed article that could become a strong pillar of their business in the future.
TUYA Digital helps brands adapt early
We’re helping clients in travel, SaaS, legal and ecommerce restructure their SEO and content around GEO realities. AI assistant traffic is small today, but growing fast, and already outperforms classic channels in quality metrics that matter.
We track, optimize, and build for this shift in user behavior — not for vanity, but for business results. If you’re not already measuring AI discovery performance, now is the time to start.
What matters isn’t just that AI is citing you.
What matters is what users do when they click.