Dwell Time and SEO: Do Engagement Signals Affect Rankings?

If you have read enough about SEO, you have seen the claim: keep visitors on your page longer and Google will reward you with better rankings. The metric usually attached to that claim is "dwell time." The direct version of that claim does not hold up. No metric called dwell time, as Google Analytics or a rank-tracking tool reports it, has ever been confirmed as a ranking input, and Google's own search advocates have said so repeatedly for years. That is not the end of the story, though.

What changed the conversation is not a blog post or a leaked algorithm update, but sworn testimony and unsealed internal documents from the 2023 U.S. v. Google search antitrust trial. Those materials described a real, shipping system that uses click behavior on Google's results page to help rank results. It is related to what SEO writers mean by dwell time, but it is a narrower, more precisely defined signal, and the difference matters for how you should actually design a page. This article goes deep on that one signal, rather than cataloguing every dwell-time myth in circulation.

Three Different Things People Call "Dwell Time"

People use the phrase "dwell time" to describe at least three different measurements, and only one of them has anything to do with what surfaced in the antitrust trial.

The first is the one search engineers actually work with: the interval between a searcher clicking your result on Google and, if they return to that same results page, clicking again. This measurement happens entirely inside Google's own logs. Nothing on your website generates it, and nothing on your website can read it back.

The second is what most SEOs mean when they say "dwell time": the session duration or engagement time reported by Google Analytics, or the "time on page" estimated by a heatmap tool. This number is calculated from timestamps your own tracking code records. It has no technical connection to Google's ranking systems: Analytics and Search are separate products, and a visitor's session clock on your site is never transmitted into a ranking pipeline.

The third is a derived metric some SEO platforms publish as "estimated dwell time," built from their own crawler or panel data. This is that tool's own estimate of the first concept, built from indirect signals, not a readout of an internal Google number.

Confusing the second and third for the first is where most of the folklore comes from. When a blog says "increase dwell time to rank higher," it almost always means the second: make people spend longer on the page, as measured by your own analytics. That is not what surfaced in court.

What Google Has Actually Said About Dwell Time

Across many years of public office-hours videos, forum threads, and conference answers, Google's search advocates, including John Mueller and Gary Illyes, have repeatedly denied that dwell time, time on site, or bounce rate function as a named, direct ranking factor. That denial has been consistent and specific: Google does not have an internal metric called "dwell time" that a page's ranking is checked against.

That denial and the trial evidence are not actually a contradiction, once you separate terminology from mechanism. A spokesperson can accurately say "we have no factor called dwell time" while a separate, differently named system uses click-return behavior for a related purpose; the denials were addressing the Analytics-style metric SEOs kept asking about, not every conceivable use of click data anywhere in the ranking stack. The opposite overreach is just as common, though: treating a years-old denial as proof Google ignores click behavior entirely. Both extremes miss what is now documented.

What the Long-Click Patents Show, and Why a Patent Is Not a Feature

Since the mid-2000s, Google has filed multiple patents describing how a "long click" (a searcher selects a result and does not return to the results page, or returns only after a long interval) could be distinguished from a "short click" (a rapid return, often followed by another click) and used to adjust rankings over time. These filings describe the mechanism in detail: aggregate click patterns for a given query, compare results against each other, and treat a pattern of short clicks on one result and long clicks on another as evidence about relative satisfaction.

A patent is not proof that the described system shipped, or shipped unmodified. Companies patent ideas defensively and patent variations they never build. What the patents establish is that the mechanism has been engineering-feasible and actively considered inside Google for close to two decades, which is why, when testimony describing a live click-based ranking system became public, it read as confirmation of a long-suspected mechanism rather than a surprise.

What the Antitrust Trial Actually Documented

In the search-antitrust case the U.S. Department of Justice brought against Google, which went to trial in 2023, Google's own witnesses and internal exhibits described a production ranking system referred to internally as Navboost. Pandu Nayak, a Google vice president responsible for search ranking, testified about how it works.

Reporting on the unsealed testimony and exhibits describes Navboost as a system that aggregates historical click and query data, reportedly drawing on a window of roughly thirteen months, into a click-based score for a given query-and-result pair. That score reflects the long-click and short-click distinction from the earlier patents: results that reliably keep searchers from returning to try something else accumulate a different score than results searchers routinely abandon.

Testimony also described this signal as one input among the hundreds combined by further layers of the ranking system, sometimes referred to in reporting as "twiddlers," that adjust a base relevance score before results are shown. Executives characterized click-based signals as most useful for queries with enough search volume to build a reliable click history, and less useful for rare or freshly emerging queries where little or no click history yet exists.

This is the clearest public confirmation to date that some click-derived, dwell-time-adjacent signal genuinely operates inside live ranking. What it is not is confirmation of "dwell time" in the sense most SEO content means: it is a query-level pattern aggregated across many different searchers' click behavior on the results page, not a per-visitor stopwatch running on your website.

Pogo-Sticking: The One Signal Worth Understanding in Depth

The mechanism worth learning in detail is pogo-sticking, because it is the part of this system that a page's design can actually influence, and because its definition is precise enough to reason about.

Pogo-sticking happens when a searcher clicks a result, returns to that same results page within a short window, and then clicks a different result or changes the query. The entire event is observable to Google without your site doing anything unusual: it happens in Google's click logs, driven by the searcher's browser hitting back or reopening the search tab.

This has a consequence that folklore about dwell time usually gets backwards: pogo-sticking is not a measurement of how long someone stayed on your page, it is a measurement of whether they came back to try again. A visitor who clicks through, gets what they need in under a minute, and then closes the tab, bookmarks the page, or continues elsewhere on your site never generates a pogo-stick, no matter how short their visit was by the clock. Conversely, a visitor who leaves a tab open for a long time, walks away, and eventually returns to the results anyway can register as a short click despite a long raw duration, because what is measured is the return, not the elapsed time.

That is also why a page can have a shorter average time-on-page than a competing result, by any analytics tool's measurement, and still not be at a disadvantage on this signal: satisfied visitors who leave the browser altogether do not pogo-stick, regardless of how quickly they were satisfied.

The structural causes of pogo-sticking are not proprietary knowledge; they follow directly from a mismatch between what a result promised and what the click delivered:

• A title or meta description that promises an answer the page does not actually contain
• The relevant answer buried below unrelated content, a long preamble, or several ad units
• A login wall, paywall, or forced signup appearing immediately for content the snippet implied was accessible
• A full-page interstitial that blocks any content from being visible on arrival
• A page that loads too slowly for the searcher to wait, so they return before it ever renders

Designing Pages Around What Is Actually Verified

Once the mechanism is understood as return-click behavior rather than a raw clock, the design response changes: the goal is not to inflate a visitor's session length, but to remove the reasons a satisfied person would ever go back and try another result. A few concrete practices follow directly from that distinction:

• Put the direct answer to the query's primary intent in the first visible screen, ahead of backstory or unrelated sections, so a searcher does not scroll past filler to find out whether the page answers their question.
• Keep the title tag and meta description narrowly accurate to what the page delivers. A click earned by an inflated promise ends in a return more often than one earned by an honest description, because the returning visitor is registering a specific mismatch, not a preference for shorter pages.
• Hold interstitials, if you use them at all, until after the core content has been shown; one that blocks the first screen gives a searcher nothing to evaluate before deciding to leave.
• Treat load performance as a satisfaction issue, not just a technical one. Core Web Vitals is a separately documented ranking signal in Google's own published guidance, unlike dwell time, and slow rendering increases the chance a searcher returns to the results before your content is even visible.
• When a query has more than one common sub-intent, a term some searchers use to mean a definition and others to mean a troubleshooting question, for instance, address the dominant intent first and label the others as clearly separate sections, so a minority-intent searcher still finds their answer instead of going back to search again.

What Does Not Move This Signal, Whatever a Blog Post Claims

A few widely repeated tactics are built on the wrong model of the mechanism, and are worth naming so they can be set aside.

Autoplaying a video "to increase dwell time" changes a number in your own analytics dashboard, if it changes anything at all; it does not touch the click-return behavior recorded on Google's results page, which has no visibility into what happens on your page after the click. The same is true of forcing content across several page loads to raise a pageview count: it can make a session-duration report look better while doing nothing to the actual signal, and an annoying version of that tactic can increase the exact return-click behavior it was meant to prevent.

Exit-intent popups and chat widgets aimed at "keeping people on the page" run the same risk in the other direction: deployed against someone who already found their answer, they are more likely to cause an irritated return click than to prevent one.

Finally, a bounce-rate or estimated-dwell-time score from a third-party SEO tool is that tool's own measurement of your visitors, or its own estimate, not a readout of the click-satisfaction system described in the antitrust testimony. Treating it as a direct proxy for your standing on Google's internal signal measures the wrong thing, even when the number itself is accurate.