Short answer: the sources checked for this article do not confirm that ChatGPT, Gemini or Perplexity look at dwell time, scroll depth or bounce rate when they pick a page to cite. What is documented is narrower: Google and Bing both say they use data about how people interact with search results. Authora’s working model adds that some search-enabled chatbot answers draw on search indexes. If that holds, engagement could play an indirect role through the index, but the checked sources do not confirm that link or show how large any such role is.
That distinction matters. If you treat a plausible chain as a proven ranking factor, you risk spending your time on the wrong fixes. This guide separates what the sources say from what remains an open question.
Is there a direct link between engagement and AI citations?
Not one that has been confirmed. None of the sources in this article state that a chatbot reads your analytics data, or that it checks how long visitors stayed on your page before it decides to quote you.
Authora’s own working model points elsewhere. In our guide on hoe AI-chatbots bronnen voor antwoorden selecteren, Authora describes a model in which chatbots approximate credibility through proxy signals from ranking systems, text features and knowledge graphs. The same guide takes the view that passages which answer the exact question cleanly and are easy to extract stand a better chance of being used. Both points are Authora’s practical recommendations, not rules the platforms have documented, and the Google and Bing sources checked for this article do not confirm them.
So the honest position on user engagement and AI citations is this: a direct effect is unproven, and an indirect effect is plausible but unmeasured.
How engagement could reach AI citations through the search index
The plausible route, in Authora’s working model, runs through the search engines that some AI tools may draw on. If a search engine uses interaction data to order its results, and a chatbot pulls sources from those results, engagement could influence citations one step removed. The checked Google and Bing sources confirm only the first part of that chain.
What Google says about interaction data
Google states on its How Search Works pages that it uses aggregated and anonymized interaction data to assess whether search results match queries. It turns that data into signals that help its machine-learned systems estimate how well a result fits a search.
Note what this does and does not say. It confirms that interaction data feeds Google’s ranking systems in aggregated form. It does not name dwell time, scroll depth or bounce rate, and it says nothing about Gemini or AI Overviews.
What Bing says about user engagement
Bing is more specific. In a version of its webmaster guidelines available in its policy archive, Bing lists user engagement as something it considers. It asks whether users clicked through to results for a query, whether they spent time on the results they clicked or quickly returned to Bing, and whether they adjusted or reformulated their query.
That “spent time or quickly returned” question is close to what SEOs call dwell time. It is measured from the search engine’s side, based on behavior around the results page. It is not the same as the bounce rate in your own analytics tool.
Why the index matters for chatbots
This is where the chain becomes plausible. Authora’s current position is that ChatGPT’s search-driven answers use Bing as the underlying index, and that around 87% of its citations come from sites that rank well in Bing. In the same position, Authora describes Gemini as closely tied to Google Search, and Perplexity as favoring authoritative domains, pages that answer questions directly and fresh content. These are Authora’s stated views, not findings from the Google and Bing sources above.
Put those pieces together and you get a reasonable hypothesis: if engagement helps a page rank in Bing or Google, it may make that page more likely to be retrieved by a chatbot that leans on those results. The size of that effect is unknown. If you want the technical side of the Bing connection, see hoe de indexering door Bing de zichtbaarheid van ChatGPT beïnvloedt.
Confirmed versus speculative: a quick overview
| Claim | Status | Basis |
|---|---|---|
| Google uses aggregated, anonymized interaction data in ranking | Confirmed by Google | Google’s How Search Works page |
| Bing considers clicks, time on clicked results, quick returns and query reformulation | Confirmed by Bing (archived guideline version) | Bing policy archive |
| ChatGPT search relies on Bing’s index | Authora’s stated position | Authora client knowledge, not independently confirmed here |
| Engagement indirectly affects AI citations via the index | Plausible, not measured | Inference from the rows above |
| Chatbots read your scroll depth or bounce rate directly | Unconfirmed | No checked source supports this |
| Improving dwell time will get you cited | Unsupported | No checked source supports a causal link or timeline |
What about scroll depth and bounce rate?
These two deserve a separate note, because they are often lumped in with search-side signals.
Scroll depth and bounce rate are metrics you see in your own analytics setup. Neither source here mentions them. Google speaks of interaction data in general terms. Bing describes behavior around its results page: clicking, returning, reformulating. A search engine can observe those actions without seeing how far someone scrolled on your site.
So be wary of advice that says “lower your bounce rate to get cited by ChatGPT.” It stacks two assumptions: that the search engine uses that metric, and that the chatbot inherits it. Neither step is documented.
Should you prioritize engagement-focused UX for this reason?
Not for the citation reason alone. The evidence is too thin to justify a UX roadmap built around a hoped-for AI effect.
There are better reasons to improve engagement, and they hold up without any speculation:
- Clicks that do arrive count for more. When AI answers absorb part of the demand, the visitors who click through deserve a page that answers quickly and leads somewhere useful.
- Search-side satisfaction is documented. If people quickly return to Bing after visiting your page, Bing says it looks at that. A page that solves the query is the sensible response, whether or not a chatbot is involved.
- The same fixes support extraction. A clear answer near the top, well-labeled sections and short, self-contained passages make a page easier for people to use. Authora recommends them for AI answers too, because they make passages simpler to lift; no checked source confirms that AI systems quote such pages more often.
That last point is where the same work may serve both readers and AI answers. Our guide on wat ervoor zorgt dat inhoud geschikt is om door AI te worden geciteerd covers the structural side in detail.
If you want to track how engaged your organic visitors are, that belongs in your reporting setup rather than your citation strategy. We cover that in which metrics replace CTR when AI Overviews reduce clicks.
How to look into this for your own site
You cannot prove the mechanism from the outside. You can collect observations that tell you whether it is worth caring about for your pages.
- Check indexing first. Confirm your key pages are indexed in both Bing and Google. A page missing from a given index cannot benefit through that index-based route.
- Review interaction insights. Bing’s archived guidelines say the Bing Webmaster Dashboard provides insights into how users interact with webpages. Use those insights to spot pages that attract clicks but may not satisfy the query; they are candidates for a clearer answer.
- Run a fixed prompt set. Test the same questions across tools before and after you improve a page, using a consistent method like the one in testopdrachten voor ChatGPT, Gemini en Perplexity.
- Read changes as signals, not proof. If citations change after an update, structure, freshness, rankings and model updates are all possible causes. Engagement is one hypothesis among several.
The practical takeaway: build pages that answer the question fast and well, make sure Bing and Google can index them, and let engagement follow from that. Authority compounds over time, and the pages that solve the query cleanly are the ones worth having in any index. Want that work handled for you across Google and AI chatbots? Ontdek hoe Authora werkt.