Google Search

• 1. Introduction

  ›

  Google Search began in 1996 in the most unglamorous of settings: the garage of Susan Wojcicki's home in Menlo Park, California. Larry Page, Sergey Brin, and Scott Hassan were the three people who built it, and what they created would eventually become the most-visited website on the entire internet. By 2025, Google Search holds a 90% share of the global search engine market. That is a number almost too large to process. Nine out of every ten online searches on earth flow through one company's servers. How did a garage project reach that scale? And what does it mean for the rest of us that it did? Those are the questions this documentary will pursue.

• 2. The PageRank Idea

  ›

  Before Google, search engines ranked results by counting how often a search term appeared on a page. It was a crude measure, easy to game and easy to flood with noise. The research project at Stanford that would become Google took a different approach. The team's early nickname for the technology was BackRub, because it worked by tracing backlinks: the more important pages that linked to a site, the more important that site was assumed to be. The algorithm that formalized this logic was called PageRank, and Larry Page filed a patent for it in 1998.

  PageRank was not invented in a vacuum. It was influenced by an earlier ranking method called RankDex, developed by Robin Li in 1996. Page's 1998 patent filing explicitly cites Li's prior work. Li would go on to found Baidu, the dominant Chinese search engine, in 2000. So the intellectual lineage of the two largest search engines in the world traces to a shared root.

  Over the years, Google added many other signals on top of PageRank. The full set of ranking criteria is reported to include over 250 different indicators, and their specifics are kept secret. The secrecy is deliberate: publishing the formula would invite manipulation. In a 2007 interview with the Financial Times, Google's then-chief executive Eric Schmidt described a much more ambitious horizon for search, saying the goal was for users to be able to ask questions like 'What shall I do tomorrow?' and 'What job shall I take?' He reaffirmed that vision in a 2010 interview with The Wall Street Journal, saying that most people don't want Google to answer their questions but to tell them what they should be doing next.

• 3. Indexing the Web

  ›

  Google indexes hundreds of terabytes of information from web pages. Beyond standard HTML, the engine can surface PDFs, Word documents, Excel spreadsheets, PowerPoint presentations, certain Flash content, and plain text files. Despite that breadth, sources generally estimate that Google indexes less than 5% of the total internet. The rest belongs to the deep web, inaccessible to its crawlers.

  In August 2009, Google invited web developers to test a new search architecture it had codenamed Caffeine. The upgrade made no visible changes to the interface but added significant speed improvements and a rebuilt indexing infrastructure underneath. Google announced the completion of Caffeine on the 8th of June 2010, claiming results were 50% fresher because the index was now updated continuously rather than in batches. Behind the scenes, Caffeine moved the indexing system away from MapReduce and onto Bigtable, Google's distributed database platform.

  Mobile usage reshaped indexing priorities in October 2016, when Gary Illyes, a webmaster trends analyst at Google, announced that the engine would build a separate primary index dedicated to mobile devices. Desktop would be served by a secondary, less current index. The rollout of that change began in December 2017. Nearly 60% of Google searches now come from mobile phones, a figure that makes the shift hard to argue with.

• 4. Algorithms in Motion

  ›

  Search engine optimization grew into a full industry precisely because Google's ranking decisions carry financial weight. An ecosystem of consultants developed to help businesses improve their positions in results, working both on-page factors like title elements and body copy, and off-page factors like the anchor text of external links. Google has published guidelines for website owners who use legitimate optimization services, but the arms race between the company and those trying to game the system is ongoing.

  The story of the DecorMyEyes website illustrated how strange that dynamic can become. Numerous consumer complaints about the company generated mentions on news websites, and those mentions, according to Google, were what drove its rankings up rather than the complaints themselves. Google fixed the underlying issue with an undisclosed algorithm change.

  In 2013, Google overhauled its search algorithm under the name Hummingbird, a reference to the speed and accuracy of the bird. The change was announced on the 26th of September 2013, though it had already been running for a month. Hummingbird placed greater weight on natural language and conversational queries, looking at context and meaning rather than individual keywords. It also sent web developers a practical signal: write for humans, not for keyword density.

  In August 2018, a broad core algorithm update was named Medic by industry analysts at Search Engine Watch and Search Engine Land, after it disproportionately affected health and medical websites. Google's framework for evaluating those pages centers on the concept of YMYL, short for Your Money or Your Life: pages where misinformation could cause real financial, physical, or emotional harm face especially high quality thresholds. In March 2024, Google announced an update expected to eliminate 40% of all spam results, with the rollout confirmed complete on the 20th of that month.

• 5. Beyond the Search Box

  ›

  Universal Search launched on the 16th of May 2007, merging results from different content types into a single page. Before that change, a standard search returned only website links. After it, the results page could include news, images, maps, videos, and blogs all at once. Marissa Mayer, then vice-president of search products and user experience, described the goal as breaking down the walls that had separated Google's various search properties.

  The Knowledge Graph arrived in May 2012, beginning in the United States and expanding internationally before the end of the year. It draws information from a range of sources and presents it in a box to the right of search results. Within seven months of launch, its coverage had tripled in size. By May 2016, it could answer roughly one-third of the 100 billion monthly searches Google processed at that time. The Knowledge Graph's information is also used to generate spoken answers in Google Assistant and Google Home.

  In June 2017, Google expanded search results to include job listings aggregated from major job boards and company homepages. A Personal tab launched in May 2017, allowing users to search content within their own Google accounts, including Gmail messages and Google Photos. Google Discover, a personalized feed of articles and videos, launched in December 2016 and was renamed from Google Feed in September 2018. The "OK Google" voice search feature was announced at the Google I/O developer conference in May 2013 and, as of May 2016-20% of mobile search queries were being made by voice.

• 6. AI Overviews and the New Search

  ›

  At the Google I/O event in May 2023, Google unveiled a feature called Search Generative Experience, an experimental tool that produced AI-generated summaries in response to search queries. The feature was available through Google Labs. Its development was driven in part by the alarm caused inside Google by the launch of ChatGPT, which the source describes as sending Google executives into a panic over the potential threat to the search business. At I/O in 2024, the feature was upgraded and renamed AI Overviews.

  AI Overviews rolled out to users in the United States in May 2024. The first weeks were rough. Errors went viral: the tool suggested adding glue to pizza, advised eating rocks, and incorrectly described Barack Obama as Muslim. Google characterized these as isolated examples. Within two weeks, the company made technical changes and scaled back the feature, pausing it for some health-related queries and reducing its reliance on social media posts as sources. Scientific American criticized AI Overviews on environmental grounds, noting that such a search uses 30 times more energy than a conventional one.

  The feature expanded to the United Kingdom, India, Japan, Indonesia, Mexico, and Brazil in August 2024. On the 28th of October 2024, it reached 100 more countries, including Australia and New Zealand. In March 2025, Google introduced AI Mode, an experimental feature using the Gemini 2.0 model that accepts complex multi-part queries and supports text, images, and voice. Initially, AI Mode was available only to Google One AI Premium subscribers in the United States through the Search Labs platform. Danielle Coffey, CEO of the News/Media Alliance, said when AI Overviews were announced in May 2024 that the feature would be 'catastrophic' to publisher traffic.

• 7. Monopoly, Privacy, and Consequences

  ›

  On the 5th of August 2024, a US judge in D.C. Circuit Court found that Google held an illegal monopoly over internet search, in violation of Section 2 of the Sherman Act. Judge Amit Mehta concluded that the company maintained its dominance by paying large sums to phone makers and browser developers to install Google as the default search engine. A November 2023 disclosure during the antitrust trial revealed that Google pays Apple 36% of all search advertising revenue generated through the Safari browser. An economics professor at the University of Chicago made that figure public. The revenue from Safari users had been kept confidential, but the percentage implies the total is likely in the tens of billions of dollars. Google's lead attorney reportedly cringed visibly when the number was disclosed.

  In April 2025, a trial began to determine what remedies the Department of Justice could impose, potentially including breaking up the company or restricting its ability to use its data to build dominance in the AI sector. Separately, the EU Court of Justice ruled on the 10th of September 2024 that Google's treatment of rival shopping searches was discriminatory, upholding a fine of 2.4 billion euros.

  Privacy concerns have accumulated alongside the company's growth. In 2012, the US Federal Trade Commission fined Google 22.5 million dollars for violating an agreement not to intrude on the privacy of users of Apple's Safari browser. Google's use of long-term cookies has drawn criticism for enabling it to track search terms and retain data for more than a year. Searches have also triggered geofence warrants and keyword warrants in which information is shared with law enforcement. In 2023, the Colorado Supreme Court upheld the use of search history requests to identify suspects in a 2020 arson case. The court noted that its ruling was not a broad proclamation and that the warrant lacked individualized probable cause, a distinction that left the question of scope far from settled.