Dwell time is how long someone spends on your page after clicking a search result, before returning to the results. Google has never confirmed using it, no tool can measure it, and the industry has discussed it for over a decade regardless.
It’s worth understanding properly, because the confusion around it produces some genuinely bad advice.
Where the term came from
Not Google. Dwell time entered the search vocabulary through Bing, whose engineers wrote publicly about measuring how long a searcher stayed on a result before coming back. Duane Forrester, then at Bing, described it in those terms around 2011.
From there it migrated into SEO conversation as a Google ranking factor, a leap nobody ever justified. Google has not confirmed it, and its representatives have pushed back on the idea repeatedly over the years.
Nobody outside Google can measure it
This is the part that should settle most arguments. Dwell time is measured from the search results page: the clock starts when someone clicks and stops when they come back. Your analytics cannot see the return to Google, because that happens on Google’s property.
What your analytics measures is time on page or engagement time, which is a different quantity taken from a different vantage point. The 2 can diverge wildly. Someone who reads your page for 4 minutes and then closes the tab satisfied never returns to the results at all.
So any tool claiming to report your dwell time is reporting something else and calling it dwell time. Treat that as a reason to distrust the tool.
What is actually established about user behaviour
Here’s where the instinct behind dwell time turns out to be partly right, and where the evidence actually sits.
Click behaviour is used, and this is confirmed
During the US antitrust proceedings, Google’s Pandu Nayak confirmed under oath that Navboost is one of Google’s important ranking signals. Navboost re-ranks results using aggregated click data over a rolling window of roughly 13 months. That is sworn testimony describing Google’s own system, which puts it in the strongest evidence class available.
The leaked documentation describes click quality, not duration
The 2024 Content Warehouse leak referenced fields distinguishing good clicks from bad clicks, along with squashed and unsquashed variants. Commentators reasonably read that as Google evaluating whether a click led to satisfaction rather than just counting it.
What that documentation does not do is confirm a duration threshold, a dwell time metric, or how any of it is weighted. An attribute existing in internal documentation tells you Google can measure something. It does not tell you the measurement reaches live ranking or how much it counts.
The honest summary
| Claim | Status |
|---|---|
| Dwell time is a confirmed Google ranking factor | Not established. Google has never confirmed it |
| Bing discussed dwell time publicly | True, and it’s where the term originated |
| Google uses aggregated click behaviour to re-rank | Confirmed under oath (Navboost) |
| Google distinguishes good clicks from bad clicks | Named in leaked internal documentation, unweighted |
| Your analytics can measure dwell time | False. It cannot see the return to the results page |
| A specific number of seconds is the threshold | Invented. No source exists for any figure |
What to do about it
Precisely what you’d do anyway, which is why this rarely deserves the argument it gets. Answer the question the searcher asked, near the top, in the format they expected. Make the page easy to read on a phone. Don’t make people hunt.
What not to do is optimise for the metric directly. Padding an article with 600 words of preamble to increase time on page makes the page worse for every reader in exchange for a number nobody outside Google can see. If click satisfaction is being measured, a slower route to the answer is the thing you’d be punished for, not rewarded.
Buying traffic or click services to manipulate these signals runs directly into the fraud detection the same leaked documentation describes, and into Google’s spam systems. It’s an expensive way to achieve nothing.
Common mistakes
- Reporting time on page as dwell time. Different measurements, different vantage points, different meanings.
- Quoting a threshold. Any specific number of seconds you’ve seen quoted was invented by whoever wrote the article.
- Padding content to inflate it. Optimising an invisible metric by damaging a visible experience.
- Confusing it with bounce rate. Related confusions, both unmeasurable by Google from your analytics.
- Dismissing user signals entirely. The opposite error. Navboost is real and confirmed; it just isn’t your dwell time metric.
Related
Bounce rate is the same confusion wearing different clothes. Google’s ranking systems covers Navboost and the undocumented layer. The Navboost testimony is in the research library with the evidence graded.