Attention Metrics and SEO Rankings in 2025

Attention Is the New SEO — But Not the Way Most People Tell You

The Advice That Does Not Survive Contact With Reality

Search for "engagement SEO" and you will be told, confidently, that Google rewards long dwell time, punishes bounce rate, and watches your scroll depth.

There is a problem with all three: Google cannot see any of them.

Scroll depth, time on page, and bounce rate are numbers produced by your analytics script running on your site. They live in your Google Analytics property — a product Google has repeatedly stated is not wired into Search ranking. On a site with no analytics installed at all, those metrics do not exist, and yet that site ranks perfectly well. Google's own search advocates have spent years saying bounce rate and time on page are not ranking factors, and on this point the evidence agrees with them.

So the standard advice asks you to optimize numbers the ranking system never receives.

What Google Actually Measures

Here is the part that is real, and it is more interesting than the myth.

Google does not watch what you do on the page. It watches what you do in its own search results — and it has done so since roughly 2005 through a system called NavBoost.

NavBoost stopped being a rumour during the U.S. antitrust trial. Under oath, Google VP of Search Pandu Nayak described it as one of Google's important ranking signals: a re-ranking system built on click data, segmented by location and device, running on a rolling window of roughly the last 13 months. His description of the mechanism was deflating in the best way — it is not some inscrutable neural net. It is, in his words, essentially a big table recording that for a given query, a given result got a certain number of clicks.

The 2024 Content Warehouse documentation leak lined up with that testimony, surfacing field names like goodClicks, badClicks, and lastLongestClicks.

Notice what those names imply. Not all clicks count the same. A click where the searcher lands, gets what they wanted, and stops searching reads differently from a click where they bounce straight back to the results page and try the next link. The second one is the pattern Google can see clearly — because both events happen on Google's own property.

That is the whole distinction. Google does not know you scrolled 80% of the page. It knows whether you came back to the search results ten seconds later and clicked something else.

Why This Distinction Changes What You Do

If you believe Google measures on-page engagement, you chase things like autoplay video loops, artificial page-length, and interaction gimmicks that inflate session duration.

If you understand Google measures search-result behaviour, the job becomes much simpler and much harder:

Your title and description are a promise. They set the expectation the click is made on. A title engineered purely for clicks, that the page then fails to deliver on, produces exactly the return-to-search pattern that hurts you. Overselling in the SERP is self-defeating in a way that plain clickbait analysis misses.

Answer the actual question, near the top. The searcher who finds their answer in the first screen and stops searching has produced the strongest signal available. That is the opposite of the "bury the answer to increase time on page" advice — which, if Google truly measured dwell time, would work. It does not, and it annoys people.

Match the page type to the intent. Someone searching a price question wants a price, not a consultation form. Give them the wrong page shape and they will bounce back to the results no matter how well written it is.

Speed still matters — for a different reason than you think. Fast INP response times do not feed a magic engagement score. They stop people abandoning before the page is usable, which prevents the return-to-SERP event. The mechanism is indirect, and it is real.

Internal links earn the second question. Connecting related guides does not work because it pads session duration. It works because someone who finds their next answer on your site is not going back to Google to find it somewhere else.

The Honest Caveat

NavBoost is one system inside a pipeline with many. Nobody outside Google — including anyone selling you engagement optimization — knows its exact weight relative to relevance, links, or page quality, and the trial testimony did not settle that. Click data also cannot lift a page that was never in contention: you have to rank well enough to be clicked before clicks can help you.

Treat this as a reason to stop optimizing invisible metrics, not as a new lever to game. Google has been explicit that trying to manipulate click signals directly is a losing exercise, and the 13-month window means anything artificial has a long time to be noticed.

The Sitora Take

The reason we keep pointing at speed, structure, and clear answers is not that they feed a secret engagement score. It is that the only user behaviour Google reliably sees is whether a searcher's problem got solved — and there is no shortcut to that. It is a direction each core update has pushed further for small business sites.

Most "attention metrics" advice is optimizing a dashboard Google never reads. Write the page that ends the search instead.

👉 Learn more about our Custom Digital Solutions at Sitora