In recent years, Google has greatly expanded its delivery of search results that are meant to answer users’ queries right on the search page, with no need to click through, using information that Google scraped from the web or collected from partners. It has also expanded results that highlight Google-owned products, such as YouTube, Google Images, Google Maps, Google Flights, and a seemingly never-ending stream of “related” search queries that take users deeper and deeper into Google’s most valuable product, Search. These types of results, which we’re calling “modules,” often appear in boxes and have typically been visually distinct from Google’s traditional ranked search results—though those distinctions are starting to blur.
We designed an experiment to measure the quantity and placement of these Google-created and self-referential search results and how they compare to other types of results and links. In particular, we wanted to compare them against results and links for websites not owned by Google. Here we will refer to that bucket of outward-leading content in search results as “non-Google.”
Because there is no current, randomized sample of Google search queries that is publicly available, we created a sample of 15,269 searches from all topics appearing in Google Trends between November 2019 and January 2020. We used Google’s division of searches into root words and ran each of those through Google Search on a mobile emulator for an iPhone X.
We then used a novel technique to measure how much space Google apportioned to various types of results and links on the first page of search results. (Details in Appendix 1.)
We classified search results into four categories: Google, non-Google, ads and AMP (originally an acronym for “accelerated mobile pages”), which are pages written by third parties (often news sites) in a markup language created by Google and cached on Google’s servers for quick loading on mobile. We categorized as “Google” those results or links that send users to Google sites and YouTube, as well as text inside Google “modules” that does not link out. (See more in the Categorization section below.)
We found that Google results dominate the first screen, taking up 62.6 percent of it in our sample. That percentage drops when considering the entire first page, where Google’s share was 41 percent. By comparison, Google apportioned 44.8 percent of the first results page to non-Google websites, 8.7 percent to AMP pages, and one percent to ads.
In addition, we found that non-Google results were pushed down to the middle and lower middle of the page, while Google gave its own results the choicest locations at the top of the search results, as shown in this plot.
In this plot, the x-axis is the percentage of the page section occupied by each category. (Note that Google is divided into two categories.) Each horizontal grid line is 10 percent down the page. The wider a color, the more area occupied by that category. Source: 15,269 tending searches
We found that in more than half of the searches in our sample, Google content took up at least 75 percent of the first screen. In one in five searches, non-Google content was entirely absent from the first screen.
Placement on the search page matters. Data from the search engine analytics tools from software firms Advanced Web Ranking and Sistrix shows that the click-through rate drops off dramatically from the top to the bottom of the search results page on mobile devices. (The same dynamic takes place on desktop.)
The effects of Google’s actions on web browsing behavior cannot be overstated. It’s estimated that in the United States, Google serves nearly nine in 10 web searches. The company says it receives more than 63,000 queries each second.
It’s important to note that we used a novel categorization method to determine which results to consider “Google” and which to consider “non-Google.” There are no universally agreed-upon standards.
Google spokesperson Lara Levin said that because our sample is not random, our results may contain more Google “answers” and AMP results than a truly random sample would have.
She criticized our categorization choices, saying that AMP content should be counted as non-Google and stating that not all results that we labeled “Google” benefit the company. “Providing feedback links, helping people reformulate queries or explore topics, and presenting quick facts is not designed to preference Google. These features are fundamentally in the interest of users, which we validate through a rigorous testing process.” (See further comments in the Google Response section.)
When we measured the effects of multiple potential interpretations of Google and non-Google results, our classification system turned out to be neither the strictest nor the most generous for either, instead falling in the middle. In every definition option we explored, Google still gave itself the most space on the coveted first screen.
In 1997, Sergey Brin and Larry Page registered Google.com as the home for an innovative search engine that would help people find what they were looking for on the web.
It quickly claimed market share from existing search engines, such as Yahoo and MSN. By the end of 2004, Google was the most popular search engine in the U.S., and it now captures 88 percent of the U.S. search engine market, according to Statcounter.
In 2004, the year Google went public, Page laid out a vision for Google as a pure reference tool: