User Behavior and SEO: What Google Says About Clicks, What a Federal Court and a Leak Revealed, and What to Do About It

What Google says about clicks on its own pages

Start with the one source nobody disputes. Google's public explainer on how results are ranked says: "We also use aggregated and anonymized interaction data to assess whether search results are relevant to queries." The next sentence says that data is turned into signals for its machine-learned systems. That is a plain admission that what searchers do on the results page feeds ranking, phrased so it promises nothing about any single click.

Google's page on testing changes to Search adds the other half. It runs live traffic experiments "to see how real people interact with a feature before launching it to everyone," and reports 16,871 of them in 2023, and 4,781 launches that year. Its human quality raters, by contrast, grade sample results to benchmark quality, and Google says those ratings do not directly impact ranking.

Our guide to Google ranking factors files clicks under "partly confirmed" for exactly this reason. The rest of this page covers what came out when Google's own engineers testified under oath, and what leaked in between.

What the antitrust trial put on the record about Navboost

In United States v. Google, the Justice Department and a group of states argued that Google kept its search monopoly partly through scale: more searches mean more user data, and more data means better results. To weigh that argument, the court needed to know how Google actually uses that data. Its memorandum opinion of August 5, 2024, which found Google liable for monopolizing general search, sets out the answer in its findings of fact, drawn from testimony by Google engineers and executives and from internal documents.

The opinion also quotes Google's own internal documents on the logic: "If a document gets a positive reaction, we figure it is good." And it records testimony that newer language models did not replace Navboost in ranking. A judge's findings are a summary of evidence, not Google's engineering manual, and they describe systems as of the 2023 trial. Still, they are the most authoritative public account that exists, made under oath and tested by the other side.

Note what the record does not say. It gives no weight for Navboost, no click-through threshold and no list of what counts as a good click. It describes a system that learns from the behavior of millions of searchers over months. That is a very different machine from the one an SEO vendor describes when promising that a few hundred clicks next week will move a page.

The 2024 documentation leak, and how far to trust it

On May 27 and 28, 2024, the consultancies iPullRank and SparkToro published reports on internal API documentation for what appeared to be Google's Content Warehouse: more than 2,500 pages describing 14,014 attributes, uploaded to a public GitHub repository on March 27, 2024 and removed on May 7. On May 29, Google confirmed to The Verge that the documents were real, while cautioning against assumptions based on "out-of-context, outdated, or incomplete information."

Both analysts said so plainly in their own posts: an attribute in a schema is evidence that Google stores something, not proof that it ranks pages with it. Read the leak as a strong confirmation of the court record, especially on clicks, and as a weak guide to anything the court did not also describe.

Bounce rate, dwell time and pogo-sticking, sorted

Three terms dominate forum arguments about user behavior. Only one of them describes something Google can see.

The practical upshot: stop reporting bounce rate as an SEO number. Engaged sessions and key events by landing page, the measures in our SEO KPIs guide, tell you whether a page satisfied the visitor. Google's view of the same visit lives on its results page, where you cannot see it.

Why paying for fake clicks is a bad bet

Every revelation about clicks produces a new product: services that send bots or paid crowds to search your keyword and click your listing. Google's spam policies, last updated August 28, 2026, list machine-generated traffic, defined as automated queries and other automated access to Google Search without permission, and say it violates both the policies and Google's terms of service. A click bot is automated access by design.

• The record is long. A system trained on 13 months of data across all of Google's searches is built to absorb noise. A few hundred bought clicks are a rounding error unless they look nothing like real behavior, which makes them easy to spot.
• Clicks are filtered. The leaked fields suggest Google sorts good clicks from bad and normalizes the rest, which is what any engineer would build once click data became valuable.
• The page still has to deliver. A bot cannot call you. If the page fails real searchers, they return to the results page, and that is the behavior the record says Google measures.
• You carry the risk. Google's page on hiring an SEO says you are responsible for the actions of any companies you hire.

Our guide to white hat SEO covers how Google enforces its spam policies and what a penalty looks like.

Satisfying the search: the behavior you can actually change

If Google learns from whether searchers come back to the results page, the job is to make coming back unnecessary. Google's guide to helpful, people-first content puts the test in its own self-assessment questions: will someone "leave feeling like they've had a satisfying experience," and does your content leave readers feeling they need to search again? Here is that test turned into work on a single page.

1. 1

   Read the results page first

   Search the query in a private window. If Google shows service pages and a map, the searcher wants to hire. If it shows guides, they want to learn. A page of the wrong type sends people back no matter how well it is written.

2. 2

   Put the answer where the eye lands

   Price range, service area, hours, the next available slot: whatever the search asked, answer it in the first screen on a phone. Our mobile SEO guide covers what Google's smartphone crawler sees.

3. 3

   Keep the snippet's promise

   A title promising same-day service on a page that says call for availability wins the click and loses the visitor. Our title tags guide covers writing ones the page can honor.

4. 4

   Make the next step obvious

   A tappable phone number, a short form, a price list. A visitor who finds what they need and acts has no reason to return to Google. That is conversion rate optimization work as much as SEO.

5. 5

   Read your own signals

   In GA4, compare engaged sessions and key events by landing page for organic visits. A page with steady rankings and few engaged sessions is the one to rewrite first.

Google cannot see your bounce rate. It can see whether people came back to look for someone else.

None of this is new advice, which is the point. The trial and the leak did not reveal a secret tactic. They confirmed that the old instruction, answer the search better than anyone else, has a measurement system behind it. Writing pages that do is the subject of SEO content and the on-page SEO checklist.

Shared click data and AI answers: what changes next

The case did not end with the liability ruling. On September 2, 2025, the same court ordered remedies, and the Justice Department's announcement says Google must make certain search index and user-interaction data available to rivals and potential rivals. Both sides appealed to the US Court of Appeals for the D.C. Circuit, where the government filed its brief in July 2026, so how much data moves, and when, is still unsettled as of October 2026.

Our reading, not a court's: if competitors gain access to user-interaction data, the behavior of real searchers may come to shape more than one engine's results. A page that satisfies people would then be rewarded in more places, and a page that gets clicks and sends people back would be marked down in more places too.

AI answers are the other change. When an AI Overview answers the question on Google's page, fewer people click at all, and the clicks that remain come from searchers who want more than a summary. Our page on AI Overviews taking your clicks shows how to measure that, and how AI chooses businesses covers what the assistants weigh instead of clicks.

To see whether your pages give searchers a reason to stay, the free site audit checks speed signals, SEO, mobile, security and content and emails the fixes. If you would rather hand it off, DataCram's SEO programs work on pages, speed and the next step together, starting at $575 a month with no long contract; the exact fee follows complexity, setup and budget, per our SEO pricing. The rest of the subject is mapped in the SEO guide, starting with how search engines work.