2000. Google launches AdWords

The moment

Mountain View, California, 23 October 2000. Google is two years old, privately held, answering more than 20 million searches a day, and it puts out a press release [1]. The product is called AdWords. Any business can go to adwords.google.com, pick keywords, write a short text ad, and watch it appear beside the search results within minutes. There is no sales call and no agency in between. Around 350 businesses and agencies signed up during the beta earlier in the month. The ad shows only when a user types the words the advertiser bought; the release calls the result a “prequalified visitor”. Nobody calls it the end of the media department.

What came before

Banner advertising on the web was sold the way magazine pages were, by the thousand impressions, on the hope that the right people were looking. Google’s first ads, the premium sponsorships of August 2000, went through a sales team at a cost per thousand, and AdWords at launch kept that logic: $15, $12 and $10 per thousand impressions for the top, middle and bottom positions [1].

The idea that a search engine could auction its results had already been tried, and booed. In February 1998 Bill Gross of the Idealab incubator stood on the TED stage and demonstrated GoTo.com, where advertisers bid for position and paid only when someone clicked. Gross remembered the logic years later: “The whole point of Internet advertising, I thought, was accountability. You could measure it, unlike with print ads.” GoTo went public in June 1999, won the AOL syndication deal in September 2000, renamed itself Overture in 2001, and by early 2003 had 88,000 active advertisers [14, 15, 12].

Gross had not patented it, and said so: “We didn’t patent the idea. So if we don’t patent it, they can copy it.” The disputed first is therefore clear enough. Paid placement by auction was Overture’s. The self-service keyword ad, matched to a query and bought without a human, was Google’s. Google put the two together sixteen months later [14].

What it changed

On 20 February 2002 Google announced AdWords Select: cost per click instead of cost per thousand, and an auction in which position was set by “a combination of ad performance (click-thru rate) and how much an advertiser agrees to pay per click” [2]. The second clause is the invention. An ad nobody clicked sank however much its owner bid, which made relevance a pricing input.

The day after, a marketer could do three things that had never been possible together. Buy a stated need rather than an audience that might have one. Pay only when the need became a click. Change price, copy and keyword at two in the morning with no insertion order and nobody to take to lunch. Philip Kotler and Kevin Lane Keller describe it in Marketing Management as a “continuous auction on search terms that serve as a proxy for the consumer’s product or consumption interests” [25].

The money followed fast. Google’s advertising revenue was $66.9 million in 2001, $410.9 million in 2002 and $1.42 billion in 2003, and advertisers supplied 94 to 99 percent of revenue through the first decade [3, 4, 5]. In 2025 Google advertising revenue was $294.7 billion, and Alphabet still took “more than 70% of total revenues from online advertising” [11].

2001$0.07

2002$0.41

2003$1.42

2004$3.14

2005$6.07

2006$10.49

2007$16.41

2008$21.13

2009$22.89

2010$28.24

2011$36.53

2012$43.69

2013$51.07

2014$59.62

2015$67.39

2016$79.38

2017$95.38

2018$116.32

2019$134.81

2020$146.92

2021$209.50

2022$224.47

2023$237.86

2024$264.59

2025$294.69

The line never bends down; it is the one advertising series of the century that grew in 2001, 2009 and 2020.

Source: Google Inc. and Alphabet Inc., Forms S-1 (2004) and 10-K (2004 to 2025), US Securities and Exchange Commission; 2001 from the S-1 net revenue basis, 2002 onward from the 10-K gross basis, 2013 and 2014 as restated in the 2015 10-K, 2016 onward "Google advertising" (Search and other, YouTube ads, Network). Redrawn by Brave New Media.

Unit: $ billions

Show the data

Year	Google advertising revenue ($ billions)
2001	0.07
2002	0.41
2003	1.42
2004	3.14
2005	6.07
2006	10.49
2007	16.41
2008	21.13
2009	22.89
2010	28.24
2011	36.53
2012	43.69
2013	51.07
2014	59.62
2015	67.39
2016	79.38
2017	95.38
2018	116.32
2019	134.81
2020	146.92
2021	209.50
2022	224.47
2023	237.86
2024	264.59
2025	294.69

Overture sued in 2002 [14]. Yahoo bought Overture in 2003 for about $1.63 billion [12]. On 9 August 2004, nine days before Google priced its IPO, the two settled: Google took a licence to US Patent 6,269,361, Yahoo dropped the suit, and Google issued 2.7 million Class A shares to Yahoo for the patent and an older warrant dispute [13].

Search became the largest category of US internet advertising and has led it since: 35 percent of the total in 2003, 40 percent in 2004 [16a]. The share has slipped as video and social grew, from 48 percent in 2016 to 38.8 percent in 2025, while the dollars climbed from $34.6 billion to $114.2 billion [16b, 16h].

Search is still the largest format at $114 billion, but video and social have taken nine points of share from it in ten years.

Source: IAB and PwC, Internet Advertising Revenue Report, full-year editions 2017 to 2025 (search includes mobile search in every year shown; earlier editions reported mobile separately). Redrawn by Brave New Media.

Unit: % of US internet ad revenue

Show the data

Year	Search share of US internet ad revenue (%)	Search revenue ($ billions)
2016	48.0	34.6
2017	46.0	40.6
2018	45.1	48.4
2019	43.9	54.7
2020	42.2	59.0
2021	41.4	78.3
2022	40.2	84.4
2023	39.5	88.8
2024	39.8	102.9
2025	38.8	114.2

In 2019 US television advertising was still $70.6 billion against $54.7 billion for paid search, in the same IAB and PwC tables [16d]. By 2021 search had reached $78.3 billion and television, on eMarketer’s count, $65.7 billion [16e, 17]. The crossing sits in 2020 or 2021; we could not reach the one table that would settle it.

Paid search ($ billions)Television ($ billions)

2017$40.6$70.1

2018$48.4$71.0

2019$54.7$70.6

2021$78.3$65.7

2022$84.4$66.6

Television was still $16 billion ahead in 2019; two years later search was $13 billion ahead, and the gap has only widened.

Source: Search: IAB and PwC, Internet Advertising Revenue Report, full-year 2017, 2018, 2019, 2021 and 2022. Television 2017 to 2019: PwC Global Entertainment and Media Outlook as published in the same IAB reports; television 2021 and 2022: eMarketer, US Ad Spending 2023 (May 2023), broadcast and cable, excluding digital. 2020 omitted: the full-year 2020 IAB report was not reached. Redrawn by Brave New Media.

Unit: $ billions

Show the data

Year	Paid search ($ billions)	Television ($ billions)
2017	40.6	70.1
2018	48.4	71.0
2019	54.7	70.6
2021	78.3	65.7
2022	84.4	66.6

Google said in October 2000 that it handled more than 20 million searches a day; in August 2012, 100 billion a month; in May 2016, “trillions” a year; in March 2025, “more than 5 trillion searches on Google annually” [1, 20, 21, 19]. One engine has held around nine in ten of the world’s searches in every year StatCounter has counted [18].

October 2000As stated by Google: more than 20 million searches per day0.02

August 2012As stated by Google: 100 billion searches per month3.3

May 2016As stated by Google: "trillions" per year, at least 2 trillion (floor)5.5

January 2025As stated by Google: more than 5 trillion searches per year (floor)13.7

Google publishes a volume figure about once a decade; each one is a floor, and each is a different order of magnitude from the last.

Source: Google press release, 23 October 2000; Google search press briefing, 8 August 2012, as reported by Search Engine Land; Google Performance Summit, 24 May 2016, as reported by Search Engine Roundtable and later confirmed to Search Engine Land as at least 2 trillion; Google, The Keyword, 3 March 2025 (Google internal data, January 2025). Monthly and annual figures converted to a daily rate by Brave New Media (divided by 30.4 and 365); 2016 and 2025 are floors. Redrawn by Brave New Media.

Unit: Searches per day, billions

Show the data

Statement	Searches per day (billions)	As stated by Google
October 2000	0.02	more than 20 million searches per day
August 2012	3.3	100 billion searches per month
May 2016	5.5	"trillions" per year, at least 2 trillion (floor)
January 2025	13.7	more than 5 trillion searches per year (floor)

Google (%)Bing (%)Yahoo (%)

Sixteen years of data and the range is four points: the auction has had one house the whole time.

Source: StatCounter Global Stats, Search Engine Market Share Worldwide, all platforms, yearly 2009 to 2025, page-view sample across StatCounter's tracking network. Redrawn by Brave New Media.

Unit: % of searches

Show the data

Year	Google (%)	Bing (%)	Yahoo (%)
2009	90.28	2.30	4.78
2010	90.91	3.46	3.93
2011	91.07	3.57	3.59
2012	91.85	3.09	3.01
2013	90.24	3.36	3.12
2014	89.81	3.63	3.57
2015	90.61	3.02	3.40
2016	92.01	2.79	2.62
2017	92.09	2.76	2.12
2018	91.40	2.82	2.15
2019	92.63	2.44	1.83
2020	92.08	2.70	1.63
2021	92.01	2.58	1.49
2022	92.07	3.19	1.36
2023	92.38	2.97	1.16
2024	90.67	3.74	1.22
2025	89.91	4.02	1.37

The Mad Men thread

Readers who arrive through Mad Men come for the copywriters. The record says the money came from somewhere else. From about 1920 the American media granted recognised agencies a 15 percent commission on the space and time they placed, a rate the American Association of Advertising Agencies had helped fix in 1918; the client paid the gross rate and the agency kept the difference [30, 31].

The commission paid for copy, research, layouts and media planning, as Time explained in February 1956, when a Justice Department consent decree ended the 4As requirement that members collect 15 percent. The rule went; the habit stayed. In 1960 Shell and Ogilvy’s agency tried a fee instead of 15 percent and made the New York Times for it [29, 31].

So the 1960s media department earned more when the client spent more. Prices were negotiated between people who knew each other, and nothing in the system rewarded buying less.

Mel Karmazin, then chief executive of Viacom, visited Google in 2003 and heard Page and Brin make the case for measuring everything, including the effectiveness of advertising. Ken Auletta records what Karmazin said on his way out, only half in jest: “You’re fucking with the magic!” Auletta’s next line is the one to keep. “Engineers don’t believe in magic.” [22]

The auction did what the consent decree could not. There was no commission to collect and no sales rep to keep sweet, and a client with a credit card could run the account from a laptop. The flight of advertisers from old to new media, Auletta writes in Frenemies, “started in the late 1990s and accelerated in the new century” [23].

The holding companies answered by putting a desk of their own between the client and the exchange: VivaKi at Publicis from 2008, Accuen at Omnicom, Xaxis at WPP’s GroupM in June 2011, which Brian Lesser launched as “a new, global audience buying company” [32]. The agency was back in the room, and this time the margin was its own to defend rather than the medium’s to grant.

Applied: if you are a B2C brand

Say you sell a mattress online and search is your largest line. The AdWords record says you are buying intent that already exists. Someone typed “best mattress for back pain” before your ad appeared, and the auction charged you for being there when they did. Les Binet and Peter Field, in the IPA’s Media in Focus, found paid search and email the most effective activation channels, and warned that activation effects “tend to decay away quickly, and don’t tend to build much over time” [26]. Search harvests what something else planted.

So run two books. In the first, treat every query as a buying situation and manage it like stock: exact and phrase match on the terms that convert, negative keywords for the ones that only cost, and a cost-per-order ceiling you enforce. In the second, accept that how many people search for a mattress this month is set by how many beds are wearing out, and that only brand work changes whose name they type. Kotler and Keller draw the same line, broad terms for brand building and specific ones for leads [25].

We watched it happen on an anonymised residential case, a developer launching a property into a category nobody was searching yet. We built a forty-eight-term keyword architecture in four intent groups, “not to capture demand” but “to find out which door people would come through once demand existed”, and kept the paid search money on the established project, where demand was real. Qualified enquiries for the launch grew every month of the quarter. Search spend follows demand and never leads it [33].

Applied: if you are a B2B brand

Say you sell industrial sensors and your buyers search rarely but specifically. A plant engineer who types “ATEX certified vibration sensor 4-20mA” has told you what she needs, the standard it must meet and the signal it must speak. That query is a specification, and the auction will price it as one. Pay. Write the ad in her words and land her on the datasheet. Measure to the quote request rather than the click.

John Dawes of the Ehrenberg-Bass Institute has the arithmetic for the rest of the budget: firms change providers of services like banking or software “around once every five years on average”, so “only 20% of business buyers are ‘in the market’ over the course of an entire year; something like 5% in a quarter” [27]. The other 95 percent are typing nothing this quarter, and no bid reaches them. Dawes concludes that you advertise to people who are out of the market now, so yours is the name they know when they come in.

So the sensor company runs search narrow, a few hundred exact queries with a cost per lead finance has signed off, and spends the rest on being the name in the engineer’s head when the old sensor fails. The thin query is a door. Somebody still has to make her want to walk through yours.

What to do with it

1. Split the search account into demand you capture (queries people type) and demand you wish existed (queries you hope for), and report them as two lines. If the second line is still spending in month three, move that money to brand work.
2. Pull click-through rate by keyword every month and rewrite or cut the bottom quartile before raising a bid; the auction has priced relevance since 2002.
3. Set one cost-per-enquiry ceiling per intent group and write it into the bid rules, so a competitor’s enthusiasm cannot inflate your line.
4. Put a dated number on the share of this quarter’s budget that reaches people who are not searching; if it is under half, write down why.
5. Once a year read the IAB and PwC format table and Alphabet’s 10-K revenue note; if your search line grows faster than the category, find out who is funding the difference.