Vendor lineage
Ranked pages by who linked to them rather than how often they said the word - and was nearly sold to AltaVista for a million dollars.
Larry Page and Sergey Brin met at Stanford in 1995 and began a research project in January 1996 called BackRub, which analysed which pages linked to which. The insight was that a link is a vote, and that votes from well-linked pages should count for more. They called the resulting algorithm PageRank.
Search engines at the time largely ranked results by how often a search term appeared on a page, which is trivially easy to game and was being gamed heavily. PageRank measured something a page's own author does not control: who else thought it was worth linking to.
In 1998 Page and Brin tried to sell PageRank to AltaVista for $1M, intending to go back to their studies. AltaVista did not buy it. They registered google.com on 15 September 1997 - the name a misspelling of googol, the digit one followed by a hundred zeros - and incorporated Google Inc. on 4 September 1998 in a friend's garage in Menlo Park.
The first cheque came from Andy Bechtolsheim, a co-founder of Sun Microsystems, followed by $25M from Kleiner Perkins and Sequoia in 1999. The IPO in 2004 raised $1.67B.
The homepage was deliberately bare while every competitor was becoming a portal. That was partly conviction and partly circumstance - neither founder was much of an HTML author - but it landed at exactly the moment AltaVista and its peers were adding shopping, email and news to the front page of a search engine.
A footnote worth keeping: RankDex, built by Robin Li in 1996, was exploring link-based ranking on a similar timeline. Li patented that work and later founded Baidu. PageRank was not the only idea of its kind; it was the one that got built into a company that wanted to be a search engine.
The timeline
- Three rejections before the famous one
In the spring they approached Excite, Infoseek and Lycos with the technology and were turned down by all three, reportedly because the incentives did not align: a search engine selling advertising against page views has no obvious reason to want visitors to find what they came for and leave.
- And the fourth, with a stated reason
The offer to AltaVista at a million dollars was refused because its parent, Digital Equipment Corporation, was not receptive to technology developed outside it. Four companies in the search business were offered the thing that would end their industry, and the last of them declined it on a point of principle about where ideas should come from.
- A cheque to a company that did not exist
Andy Bechtolsheim's $100,000 was written to an entity that had not yet been incorporated. The patent on PageRank was filed on 1 September; the company was incorporated on the 4th. By then the service was handling ten thousand queries a day and straining Stanford's network.
- AdWords, and the auction
Advertising sold by auction against keywords rather than by negotiated placement - which is the invention that paid for everything, and a larger commercial idea than the ranking algorithm it sits on.
Flagship products and solutions
- The index and the rankingThe original product, still the core one, and now built on hundreds of signals of which the original link analysis is only the oldest.
- AdWords and AdSenseThe auction that monetises the query, and its extension onto other people's pages. The search engine is the funnel; this is the business.
- The clean pageWorth listing as a product decision rather than a design one. Everything a competitor added to its front page took attention away from the box, and the box was the entire proposition.
Key innovations
- Measuring what the author cannot controlKeyword frequency is written by the person who wants to rank. Inbound links are written by everybody else. Moving the signal from the page to the network around it is the whole idea, and every ranking system since has been an argument about which uncontrollable signals to use next.
- Publicly funded, privately enormousThe prototype was developed with support from a National Science Foundation programme on digital libraries. Public research money produced the ranking work behind one of the most valuable companies ever built, which is worth stating plainly whenever the question of who should fund basic research comes up.
- Selling attention by auctionBefore AdWords, advertising was negotiated: a salesperson, a rate card, a placement. Running an auction for each query set the price automatically, scaled without salespeople, and let the smallest advertiser buy the same inventory as the largest. The ranking made the audience; the auction is what turned it into money.
- Refusing the portalEvery competitor concluded that a search engine was a doorway and that the doorway should be decorated. This one concluded that a search engine is a place people want to leave quickly, and built for that. The bare page was partly, as the entry above notes, a lack of interest in writing HTML - but it was the correct answer for the wrong reason, which happens more often than the histories admit.
Main markets
General web search, at roughly 90% of the global market as of 2025 - a share that has been challenged in courts rather than in the market, and whose durability is now a legal question as much as a technical one.
Its competitors are the other general engines, the regional ones that hold their own markets, and increasingly the assistants and answer engines that remove the click the whole business model depends on.
Analyst standing
- There is no meaningful competitive assessment to offer of a service with nine tenths of a market. The assessments that matter are regulatory, and they are ongoing in several jurisdictions.
- The observation this timeline is better placed to make is about the four refusals. Excite, Infoseek, Lycos and AltaVista each had the chance to buy the ranking method that displaced them, and each had a reason not to that made sense from inside. Incumbents rarely fail to see a technology; they see it and correctly identify that adopting it would damage something they already have.