Bounce rate and SEO: what the number measures now, and how to use it properly

Bounce rate is a diagnostic for your pages, not an input to your rankings. Google's own explanation of how it ranks results describes meaning, relevance, content quality, usability and context, and your analytics bounce rate is not among the things it names. What bounce rate is genuinely good for is finding the pages where the traffic you already earn arrives and leaves immediately, which is a content and intent problem you can fix. It also means something different now from what it used to mean: in Google Analytics 4, bounce rate is simply the inverse of engagement rate, so reading it without knowing how an engaged session is defined produces the wrong conclusion. This page covers the definition, why it is not a ranking signal, and how to use it properly.

What bounce rate measures now

Google's analytics help page on engagement rate and bounce rate defines bounce rate as the percentage of sessions that were not engaged. An engaged session is one that meets any one of three conditions: it lasted longer than a set threshold, it included a key event, or it included at least two page or screen views. The threshold defaults to ten seconds and is adjustable up to a minute in the property settings.

	Old definition	Google Analytics 4
What counts as a bounce	A session with a single page view	A session that met none of the engagement conditions
Someone who reads for four minutes and leaves	Counted as a bounce	Not a bounce. The session was engaged
Someone who clicks two pages in three seconds	Not a bounce	Not a bounce, for a different reason
Relationship to engagement rate	None	Bounce rate is one hundred per cent minus engagement rate

That change matters. Under the old definition, a page that fully answered a question in one screen scored terribly. Under the current definition, the same page scores well, which is a far more honest reflection of whether it did its job. If you are comparing today's number to a figure from an old report, you are comparing two different metrics.

Why it is not a ranking factor

Two reasons, and both are structural rather than a matter of opinion. First, bounce rate lives in your analytics property, which is your own installation of your own tool, configurable by you, with an adjustable engagement threshold. A ranking system built on a number the site owner can change in a settings screen would be trivially gamed. Second, Google's published description of how results are ranked sets out the categories of signal it uses, and site analytics figures are not among them.

What does exist, and what people are reaching for when they talk about bounce rate and SEO, is a real relationship between satisfying the searcher and ranking over time. A page that answers the query attracts links, gets recommended, earns repeat visits and gets refreshed, and all of those are things a search engine can see. Bounce rate is a thermometer for that, not the mechanism.

The number that is actually about search

For diagnosing the search side, use Search Console rather than analytics, and read two numbers together.

• Click through rate per query. High impressions and a low click through rate means your title and description are not matching the query's intent. That is a snippet problem, fixed in the title.
• Clicks against engaged sessions on the same landing page. Plenty of clicks and a low engagement rate means the page wins the click and then disappoints. That is a content problem, fixed above the fold.

Those two cover almost every real case. The first is a promise problem, the second is a delivery problem, and they need completely different fixes.

What a high bounce rate is usually telling you

1. Intent mismatch. The page ranks for a query it does not answer. Check the top queries for that URL in Search Console and either answer them on the page or stop targeting them.
2. The answer is below the fold. The visitor arrived with a question and found an introduction. Put the direct answer in the first paragraph.
3. The page is slow on a phone. People leave before the content paints. This one shows up as a bounce rate that is far worse on mobile than desktop.
4. An interstitial or consent layer blocks the content. A popup on arrival is a measurable bounce generator.
5. The traffic is not your audience. Common on pages that rank for an ambiguous phrase, and on paid traffic bought on a broad match.
6. Tracking is broken or duplicated. A missing tag on a template, or a tag firing twice, produces numbers that are nothing to do with the content. Always rule this out first.

How to read the number without fooling yourself

Never read a site wide bounce rate. It averages pages that have entirely different jobs, and the average moves whenever your traffic mix moves. Segment it four ways instead.

Page type	What a healthy figure looks like	What to do about a poor one
Blog post answering one question	Lower engagement expectations. One good read is a success	Add the obvious next step, and link two related pages in the body
Service or product page	Should be strong. People arrived to evaluate	Add proof, pricing guidance and the enquiry route higher up
Home page	Should be strong. Navigation is the point	Make the three main paths obvious within one screen
Contact or location page	Often high and fine. They got the address and left	Nothing. Judge it on calls and directions instead
Paid landing page	Judge on conversion, not engagement	Tighten match types and the message match between ad and page

Then compare each page against its own previous month rather than against a benchmark from an article. Your own trend is the only honest comparison, because device mix, traffic source and the engagement threshold all differ between sites.

This page covers bounce rate, what it measures and how to use it in search work. For the reports to set up and read, see Google Analytics for SEO and the Search Console guide. For the speed work behind a mobile bounce problem, see Core Web Vitals. For turning engaged visits into enquiries, see conversion rate optimisation.

Where people get stuck on bounce rate and SEO

Does bounce rate affect SEO?

It is not a ranking input. It is a useful signal that a page is not satisfying the people search already sends it, and fixing that is what improves the page's performance over time.

What is a good bounce rate?

The right answer is page type by page type, compared with your own previous period. A contact page and a blog post should not be held to the same figure, and a single site wide target drives bad decisions.

Why did my bounce rate change when I moved to GA4?

Because the definition changed. It is now the inverse of engagement rate, with engagement defined by time, key events and page views. Treat the new figure as a new series and start the comparison from the switch.

Should I change the engagement time threshold?

Raise it where your pages are long reads and ten seconds is too generous to be meaningful. Set it once, write down the date you changed it, and leave it alone so your trend stays readable.

Does a high bounce rate on a blog post mean the post is bad?

Often it means the opposite. The reader asked a question, the first paragraph answered it, and they left satisfied. Judge those pages on whether they are being found, read and linked.

What should I fix first on a page with poor engagement?

The first screen. Put the direct answer at the top, remove anything that blocks the content on arrival, and check the page on a mid range phone on a mobile connection.

The reader your analytics cannot count

Bounce rate describes people who arrive. Some buyers now get what they came for without arriving: they ask an assistant, it assembles a reply from pages, listings and reviews, and it names a short list of businesses. Nothing in your report records that, so a quiet week in analytics can sit beside a loud one in assistant answers, or the reverse. Before you read a falling number as a failing page, ask the questions your page answers in two assistants and see whether it is the source. That count is the one AI Knows Us takes, and your analytics never does.