The search engine wars

There was not one war

"The search engine wars" is usually told as one contest that Google won, which flattens a decade of separate fights into a single story. It is more useful as four overlapping conflicts, each with different combatants and different winners.

• The portal war, roughly 1996 to 2001, over the front page: whose site people opened first. Search was treated as a feature of a destination rather than as the product.
• The index war, roughly 1995 to 2004, over who actually crawled the web. Fewer companies were in this one than the brand names suggest, because most engines rented their results.
• The auction war, 1998 to 2005, over how search made money — won, in the sense of being invented, by a company almost nobody now remembers.
• The distribution war, from about 2004 to the present, over defaults: which engine a browser or phone uses when the user does not choose. It is the only one still running.

The portal war and the index war were fought simultaneously by the same companies, which is precisely why so many of them lost.

The portal war: whose front page

The founding assumption of the late 1990s was that a search engine is a leaky bucket: every successful search sends the visitor somewhere else, and a business measured in page views cannot survive doing that. The prescribed cure was to give people reasons to stay — email, news, weather, stock quotes, shopping, chat.

Consolidation ran fast. Excite launched in October 1995, bought Magellan in June 1996 for around $18 million and WebCrawler from AOL that November, with AOL taking about 20% of Excite. Lycos, spun out of Michael Mauldin's Carnegie Mellon project that went online on 20 July 1994, held an April 1996 IPO that was the fastest inception-to-offering in NASDAQ history to that point, and bought Wired Digital — and with it HotBot — in October 1998. Disney bought 43% of Infoseek in 1998 and the rest in 1999.

The peak numbers explain the wreckage. Excite merged with the broadband provider @Home Network in January 1999 in a deal valued at $6.7 billion; Terra Networks bought Lycos in October 2000 for $12.5 billion. Both failed completely: Excite@Home filed for Chapter 11 on 1 October 2001 and the portal sold that December for around $10 million, Terra sold Lycos in August 2004 for $95.4 million, and Disney announced on 29 January 2001 that it was discontinuing GO.com.

The most instructive casualty was not a portal by origin. AltaVista, launched by Digital Equipment Corporation on 15 December 1995, had the best index on the web; under CMGI from 1999 it became a general-purpose portal at exactly the moment relevance ranking became the thing that mattered.

The index war: whose crawler

Behind a dozen brands there were only ever a handful of crawlers — the most consistently misreported fact of the period.

AltaVista's crawler, Scooter, written by Louis Monier at DEC's Palo Alto labs, was multi-threaded — it kept large numbers of fetches in flight at once rather than crawling serially — which is why it could build a full-text index an order of magnitude larger than its contemporaries. The project existed to demonstrate DEC's 64-bit AlphaServer hardware, which had the memory to hold that index. The most important search engine of the mid-1990s began as a hardware advertisement.

Inktomi, out of the University of California, Berkeley, took the opposite approach: it never ran a consumer destination of consequence and sold its index wholesale. HotBot, launched by HotWired in May 1996, was Inktomi's technology with a magazine's branding on it, and MSN Search in 1998 used third-party results including Inktomi, LookSmart and AltaVista. Yahoo — the largest portal of all — never crawled anything in the 1990s: its hand-built directory sat over web results from AltaVista, then Inktomi, then, from 2000 to February 2004, from Google. For four years the dominant portal's results were supplied by the company that would end its search business.

Then consolidation. Overture bought AltaVista and FAST's web search division within a week in February 2003. Yahoo bought Inktomi in December 2002 and Overture in 2003, and in February 2004 dropped Google for its own crawler, Yahoo! Slurp — the only period in its history with an independent index, given up under the Microsoft search alliance of July 2009. Microsoft went the other way, building its own crawler in 2004–05 and rebranding to Bing on 3 June 2009.

The auction war: what GoTo.com invented

The fight that determined the economics of the modern internet was not about relevance at all.

Bill Gross presented the idea at the TED8 conference on 21 February 1998, to a reception contemporaneous accounts describe as confused and in places hostile: a search engine in which advertisers openly bid for placement against a keyword and paid only when someone clicked. GoTo.com, a spin-out of Gross's Pasadena incubator Idealab, launched that February with bids starting at one cent, ranked highest bid first. The company's own analogy was the Yellow Pages.

Three components made it work, all novel in combination: a keyword auction setting the price rather than a rate card; pay-per-click pricing, which moved risk from advertiser to publisher; and self-serve tooling, shipped on 1 June 1999, which let tens of thousands of small advertisers run campaigns without a salesperson. Overture reported more than 88,000 advertisers in the first quarter of 2003.

What GoTo did not have was an index. Because a page of pure advertising only covers commercial queries, it backfilled unpaid results from Inktomi from 15 June 1998 and never built a crawler. This is the most misstated fact about the company: GoTo.com ran an advertising marketplace on somebody else's index.

Its real business was syndication. Rather than fight for consumer attention, it sold its paid listings into other engines' results and split the revenue — a "GoTo-in-a-Box" agreement with Microsoft dated 15 April 2001, featured listings on Yahoo from 20 November 2001, AOL alongside them. It renamed itself Overture Services on 8 October 2001 and in 2003 bought AltaVista and AllTheWeb. Yahoo — its largest customer — announced the acquisition on 14 July 2003 at $1.63 billion and completed it on 7 October 2003; $2.2 billion is reported at completion, consistent with an all-stock deal moving with Yahoo's share price. The brand ended on 18 April 2005.

The patent, and what the record actually says

The litigation that followed is the part of search history most often distorted, so separate what is established from what is not.

Established. GoTo filed a patent application on 28 May 1999, granted on 31 July 2001 as US 6,269,361, covering bid-for-placement ranking with per-click charging. Google launched AdWords in October 2000 priced per impression and replaced it with AdWords Select in February 2002 — a keyword auction in which advertisers bid for higher placement. Overture sued FindWhat.com in January 2002 and Google in April 2002. In August 2004, immediately before Google's IPO, the Google case settled: Google took a perpetual licence to the '361 patent and related Overture patents, issued 2.7 million shares to Yahoo, and Yahoo dismissed the suit.

Not established, and not assertable. Google never admitted infringement, and no court ever ruled that AdWords infringed — a licence-and-settlement is a commercial resolution, not a finding. The patent's validity was never adjudicated either: the only case to reach a jury, Overture against FindWhat, deadlocked into a mistrial in May 2005, with an unresolved allegation that the patent was unenforceable for inequitable conduct, then settled in August 2005 for $8 million plus royalties. Nor should the Google settlement be called a payment of a dollar figure; shares were issued, and their value depends on which price and date you apply.

The influence claim is stronger than the patent claim. Google's founders had publicly opposed letting advertising drive results, and its first advertising product was not an auction. Google adopted the auction after GoTo's model had visibly worked — then made the change that decided the war: it ranked ads by bid combined with quality and clickthrough, not bid alone. A more relevant ad gets clicked more often, so quality weighting produced better results and more revenue per search. Overture ranked purely by bid, which maximises revenue per click and degrades the page, and Yahoo did not adopt quality weighting until Panama in February 2007. By then the market had gone.

The distribution war, which is still running

Once relevance and monetisation were settled, the remaining variable was placement. Most people never change their default search engine, so whoever owns the default owns the query.

Microsoft's attempt to buy its way in came first: an unsolicited bid of roughly $44.6 billion for Yahoo in February 2008, which Yahoo rejected. Eighteen months later Yahoo signed the alliance that handed its web results to Bing anyway. Yahoo paid Mozilla for the Firefox default in the US from 2014; Mozilla returned to Google in November 2017.

The payments that mattered most were Google's, and they became the subject of United States v. Google LLC, filed on 20 October 2020. On 5 August 2024 Judge Amit Mehta ruled that Google "is a monopolist, and it has acted as one to maintain its monopoly", specifically that its exclusive default-placement contracts were unlawful. The remedies decision of 2 September 2025 rejected structural relief — no Chrome divestiture — but banned exclusive distribution agreements, ordered Google to share index and click-and-query data with qualified competitors, and required it to license results and text ads on five-year terms initially capped at 40% of queries. Google may still pay for default placement if the deals are non-exclusive and no longer than a year. The remedies took effect on 3 February 2026; Google appealed, the Department of Justice cross-appealed, and as of 19 August 2026 the D.C. Circuit had not ruled.

Who won, and what it cost

Google won all four fights, but not in the same way and not for the same reason.

It won the portal war by refusing to fight it. While Excite, Lycos, AltaVista and Infoseek added horoscopes, Google shipped a page with a logo and a text box and let people leave. The metric everyone else was optimising — time on site — was the wrong one.

It won the index war on engineering and then on inheritance: by 2004 barely any independent Western crawlers were left. It won the auction war by copying an idea it did not invent and fixing the flaw its inventor never fixed. And it has been winning the distribution war with money, the fight that finally attracted a court.

The cost is visible in the market-share figures. StatCounter put Google at 91.31% worldwide across all devices in July 2026 — a proxy measured from tracked page views rather than a census of queries, but the standard public number. The 1990s had a plurality of crawlers; 2026 has Google, Bing, a handful of national engines and a small independent tier. The current war is over AI answers rather than blue links.

Frequently asked questions

Sources

• en.wikipedia.org/wiki/GoTo.com
• patents.google.com/patent/US6269361B1/en
• computerworld.com/article/1720542/yahoo-to-buy-overture-for-1-63b.html
• newsbreaks.infotoday.com/NewsBreaks/Overture-Acquires-Two-Major-Web-Search-E…
• theregister.com/2005/05/17/overture_ad_patent_mistrial/
• en.wikipedia.org/wiki/Excite_(web_portal)
• everycrsreport.com/reports/LSB11362.html
• gs.statcounter.com/search-engine-market-share