The Signal Everyone Talks About and Almost Nobody Measures Right
Dwell time — the duration between a user clicking a result and returning to the SERP — gets cited constantly as a ranking signal. Google has never confirmed it as a direct input, and the indirect evidence for how it actually functions is significantly more complicated than “longer = better.”
We’ve been tracking this closely since 2010. The pattern that keeps emerging: dwell time discussions almost always conflate three distinct things that Google treats very differently.
Three Signals People Conflate Into One
1. Pogo-sticking (fast return clicks)
A user clicks your result, hits the back button in under 10–15 seconds, and clicks a different result. Clearest negative signal. Google can observe this through Chrome usage data and the Search app, and there’s a reasonable case — supported by former Googler Paul Haahr — that rapid back-and-forth from a specific result correlated with lower satisfaction scores during quality rater evaluations. That correlation eventually feeding into ranking models is plausible, not proven.
But pogo-sticking on a navigational query where someone confirmed they’d found the right site? Different story. Google is not naive about query context.
2. Last-click satisfaction
If a user clicks a result and doesn’t return to the SERP at all during that session, that’s a potentially strong positive signal — not because they stayed on your page for a long time, but because the SERP problem was solved. A user landing on a phone number page, finding the number immediately, and closing the tab has a dwell time of maybe 8 seconds. That’s a success, not a failure.
Optimizing for raw dwell time on that kind of query is actively wrong. You’re adding friction to inflate a metric that was never designed to measure what you’re using it for.
3. Session depth as a proxy
Sometimes the return-to-SERP pattern matters less than what the user does next on your site. If they navigate deeper — to a related article, a product page, a contact form — that session behavior is visible to Google in aggregate through Chrome and Android usage data. This is the closest thing to “dwell time done right” as an SEO signal, because it’s actually a proxy for content utility, not content length.
Why Query Category Changes Everything
Here’s the framing most dwell time analysis misses: Google normalizes these signals per query type, not globally.
For an informational query like “how does compound interest work,” a 4-minute read with no back-click is a strong signal. For a transactional query like “book flight JFK to LAX,” a 45-second session followed by a conversion on your booking engine is perfect. For a navigational query like “Chase bank login,” a 6-second click with no return is ideal.
If Google is using return-to-SERP signals at all — and the evidence suggests something in this family is in use — it’s almost certainly comparing your result’s return rate against the baseline for that specific query cluster, not against some universal threshold. Your dwell time can be 90 seconds and still send a negative signal if every other result on that SERP averages 4 minutes.
What Google Has Actually Said (Carefully Parsed)
Google’s official position has consistently been that they don’t use “dwell time” as a ranking factor. What they have said — through patents, through DOJ antitrust trial documents that surfaced in 2023, and through statements from engineers — is more nuanced than a flat denial.
The Navboost system, disclosed in the antitrust trial, processes click signals including “good clicks,” “bad clicks,” and “last longest” clicks. That last category is Google’s internal label for the last result clicked in a session — the one that apparently resolved the query. These aren’t dwell time measurements. They’re click quality classifications that correlate with, but are distinct from, raw time-on-page.
The practical implication: Google is measuring click quality, not clock time. Dwell time is a rough correlate of click quality in some query categories. It is not the signal itself.
Where This Actually Shows Up in Rankings
The clearest place we see return-to-SERP behavior having a real effect is on queries with high result-set volatility — where Google is actively testing different result combinations. If a URL gets consistently high pogo-stick rates relative to competing URLs on the same query, it tends to drift down over weeks, not overnight.
It also matters more on queries where Google doesn’t have a settled understanding of which result type users prefer. On stable head terms with clear winner patterns, the signal has less room to move rankings because positions are already calibrated. On mid-tail informational queries — the kind where content quality varies wildly — getting your content structure right has the most visible impact on position stability. That’s where we focus attention first.
The Audit Move That Actually Helps
Rather than trying to instrument dwell time directly (you can’t see Google’s data; your GA4 engagement metrics are an approximation at best), focus on the leading indicator: content-to-intent match at the paragraph level.
Pull your top 20 traffic pages by impression from Search Console. For each one, identify the primary query cluster driving impressions. Then read the first 150 words of the page and ask: does this page answer or credibly begin to answer that query within those first 150 words, or does it delay?
Pages that bury the answer — intro paragraph restating the question, background section nobody asked for, actual content starting 400 words in — are structurally inviting pogo-sticks. Not because users are impatient in some abstract sense, but because they’re comparing your page against three other tabs they have open.
This is a content architecture problem more than a writing style problem. Fix the structure first.
One Genuine Caveat Worth Sitting With
Everything here is working from inference, leaked documents, and observed correlation. Nobody outside Google has access to Navboost’s actual weighting model.
The DOJ documents gave us the best confirmed view yet of how Google uses click signals — but “confirmed that a system exists” is not the same as “confirmed how much weight it carries versus link signals, topical authority, and freshness in any given ranking decision.” Anyone selling a specific number or a guarantee around dwell time optimization should be treated with skepticism.
What’s safe to act on: reduce friction for users who arrived with clear intent, structure pages so the answer is findable fast, and stop measuring content quality with a stopwatch. Be the result that solves the query. Dwell time is one imperfect window into whether you’re hitting it.
For your highest-impression queries right now — are you optimizing for users, or for metrics that proxy user satisfaction? Those can point in very different directions.