THE GREAT ATTENTION EXPERIMENT: 2 - THE DAY HUMAN ATTENTION BECAME A COMMODITY
On 27 October 1994, visitors to a new website called HotWired encountered something the World Wide Web had not shown them before.
An advertisement.
It was a brightly coloured rectangular banner created for the telecommunications company AT&T. Its message asked whether the person viewing it had ever clicked their mouse in a particular place.
Then it confidently predicted that they would.
The advert was unsophisticated by modern standards.
It did not know the viewer's age.
It did not know their location.
It did not know which websites they had visited that morning.
It had not analysed their previous purchases, estimated their income, compared their behaviour with similar people or calculated the probability that they might be considering a new phone contract.
It could not follow them around the internet.
It could not select a different message for every person who saw it.
It could not trigger a global auction among advertisers before the page had finished loading.
But it could do something most advertising before it could not.
It could know whether someone clicked.
That small piece of information changed everything.
The banner itself was not the beginning of the attention economy. Advertising had competed for human attention for centuries.
But it represented the beginning of a profound transformation.
Attention had always been valuable.
Now it was becoming visible.
Then measurable.
Then traceable.
Then targetable.
Then tradable.
Eventually, individual opportunities to influence individual human beings would be bought and sold automatically, billions of times a day, in less time than it takes someone to notice that a webpage has loaded.
There was no single day when human attention became a commodity.
There was a sequence.
The first banner advert offers us a useful place to begin.
Advertising Had Always Sold Access to People
Long before the internet, newspapers sold more than journalism.
They sold readers.
A newspaper attracted an audience by publishing information people wanted.
Advertisers paid to place commercial messages in front of that audience.
The cover price funded part of the operation.
Advertising funded another.
Magazines worked the same way.
So did commercial radio and television.
The content gathered people together.
The advertising paid for access to them.
This arrangement was not hidden.
Readers could see the advert printed beside the article.
Television viewers knew when a commercial break had begun.
Radio presenters introduced sponsors.
The separation between content and advertising was not always perfect, but the underlying exchange was broadly understood.
Advertisers were not usually buying knowledge of one identifiable person's behaviour.
They were buying access to a crowd.
A business selling fishing equipment might advertise in an angling magazine.
A toy company might advertise during children's television.
A luxury car manufacturer might choose a publication whose readership was thought to have sufficient disposable income.
The targeting existed.
It was based largely upon context, demographics and estimates.
The publication acted as a rough indication of the people likely to be paying attention.
The advert was selected because of where it would appear.
Not because an invisible system had assembled a detailed behavioural profile of the person looking at it.
Traditional advertising therefore traded in probability.
A newspaper might have a circulation of several hundred thousand.
A television programme might attract millions of viewers.
Some proportion of them might notice the advertisement.
A smaller proportion might remember it.
A smaller proportion still might eventually buy the product.
The advertiser could measure overall sales.
It could commission research into brand recognition.
It could compare one campaign with another.
But the path between a specific advertisement and a specific person's behaviour was often difficult to see.
The audience was valuable.
The individual remained largely opaque.
The internet changed that relationship.
The Click Closed the Gap
The first banner adverts introduced a new possibility.
An advertisement could become interactive.
A person could move directly from seeing a message to visiting the advertiser.
More importantly, that movement could be recorded.
The advertiser no longer had to rely entirely upon circulation figures, audience estimates or later sales.
It could count the response.
The impression showed that an advert had been loaded.
The click showed that someone had acted.
The person might not buy anything.
They might leave immediately.
They might even click accidentally.
But the distance between advertising and behaviour had narrowed.
HotWired launched the first widely recognised web banner adverts in October 1994, including the AT&T campaign. Wired's own retrospective account describes the banner as one of the features through which the young commercial web began experimenting with how online publishing might be funded.
The banner solved an immediate problem.
Publishing online still cost money.
Journalists, designers, developers and infrastructure had to be paid for.
Advertising offered a way to provide content without charging every reader directly.
There was no obvious reason to consider that arrangement dangerous.
It resembled the model newspapers and broadcasters had used for decades.
But the click introduced a fundamentally different feedback loop.
A publisher could see which adverts attracted responses.
An advertiser could compare messages.
Designers could alter wording, colour, position and imagery.
One version could be tested against another.
The more behaviour became measurable, the more advertising could be optimised around the behaviour it produced.
The audience was no longer merely sitting in front of the advertisement.
It was answering back.
The Internet Needed a Memory
The early Web had a limitation.
It was largely forgetful.
A browser could request a page from a server.
The server could return it.
When another request arrived, the website did not automatically possess a continuous memory of everything the person had previously done.
That created practical problems.
A shop needed to remember which items someone had placed in a basket.
A website needed to recognise that a user had already logged in.
Preferences had to persist between pages.
The solution was the HTTP cookie.
Cookies allowed state information to pass between a web server and a browser. The 1997 Internet Engineering Task Force specification described them as a mechanism for creating stateful sessions through information carried in Cookie and Set-Cookie headers.
Cookies were not invented as instruments of mass surveillance.
They solved a real technical problem.
Without some form of memory, many ordinary website functions would become awkward or impossible.
A shopping basket that forgot its contents every time someone opened another page would be useless.
A website that repeatedly demanded a new login would become intolerable.
A service unable to remember language or accessibility preferences would be less useful.
The technology itself was not the moral problem.
What mattered was how far that memory could travel.
A cookie created by the website someone was deliberately visiting could help that website function.
A cookie connected to an advertising network operating across many unrelated websites could do something else.
It could begin assembling a history.
The same advertising company might deliver adverts on a travel website, a newspaper, a sports blog and a shopping site.
If its identifier appeared across those places, activity that once looked separate could be connected.
The person might remain unnamed.
But their browser no longer appeared as a stranger.
It could be recognised as the same browser that had searched for flights yesterday, read a particular article that morning and viewed a product without buying it.
The internet had acquired a memory.
Advertising would soon learn how valuable that memory could become.
Context Became Behaviour
There is a crucial difference between contextual advertising and behavioural advertising.
Contextual advertising looks at what someone is viewing now.
Behavioural advertising considers what they have done before.
An advert for running shoes beside an article about marathon training is contextual.
The subject of the page determines the likely relevance of the advert.
An advert for those same shoes appearing later while the person reads the news may be behavioural.
The advert has followed an inferred interest rather than remaining attached to the original context.
The difference sounds small.
It changes the object being analysed.
Contextual advertising evaluates the page.
Behavioural advertising evaluates the person, or at least the profile attached to their device.
By 2000, the United States Federal Trade Commission was already examining what it called online profiling. The FTC described network advertising companies gathering data across unrelated websites, primarily through cookies, in a process that was largely invisible to the people being observed. It noted that profiles were often linked to an identification number rather than a person's name.
That last detail matters.
People often imagine personal data as information that explicitly identifies them.
Name.
Address.
Telephone number.
Date of birth.
But advertising systems do not always need to know who you are in the traditional sense.
They need to know enough to distinguish you from other opportunities.
They need a persistent identifier.
A history.
A category.
A prediction.
The commercial question is not necessarily:
Who is this person?
It may be:
What is this browser likely to respond to next?
That is sufficient to create value.
A profile does not need your name to influence your choices.
Search Revealed Intention
Behavioural tracking made it possible to infer interests over time.
Search advertising introduced something equally powerful.
Intention.
A person reading a motoring article may be interested in cars.
A person searching for a specific car model, dealership or finance arrangement may be closer to making a decision.
Search queries reveal what someone wants at the precise moment they express it.
That made search advertising extraordinarily valuable.
Google introduced its first advertising programme in 2000 and launched AdWords later that year. Advertisers originally paid according to how often their adverts appeared. In 2002, Google moved AdWords to a cost-per-click model, meaning that advertisers generally paid when a user clicked. By early 2004, advertising represented 96 per cent of Google's net revenue for the preceding quarter.
The model aligned payment with observable behaviour.
An advertiser could select words connected to a commercial intention.
A person could search.
Relevant advertisements could appear.
The advertiser paid when the person acted.
Google's system also used auctions to determine the placement and cost of advertisements. Its 2004 registration filing explained how bids, click-through rates and expected usefulness helped determine which adverts appeared and what advertisers paid.
This was a remarkable improvement over much traditional advertising.
A small local company no longer needed to buy a large newspaper advertisement and hope the right people noticed it.
It could pay to appear when someone actively searched for the service it provided.
Advertising became more accessible.
More relevant.
More accountable.
More efficient.
It also moved closer to the moment of decision.
The commodity was no longer simply the possibility that an audience might notice an advert.
It was access to an expressed human intention.
From Buying Space to Buying Outcomes
Traditional advertisers primarily bought space.
A page in a magazine.
Thirty seconds of television.
A section of a billboard.
A slot on the radio.
The location was agreed in advance.
The message remained broadly the same for everyone who encountered it.
Digital advertising gradually shifted the transaction.
Advertisers began paying for impressions.
Then clicks.
Then leads.
Then installations.
Then purchases.
Then increasingly sophisticated predictions about what a particular person might do.
The closer payment moved towards observable outcomes, the greater the pressure to measure everything preceding those outcomes.
Which advert was shown?
Where did it appear?
Which person or device saw it?
What had they done previously?
What happened afterwards?
Did they purchase immediately?
Did they return three days later?
Did they encounter another advert somewhere else before buying?
Did the campaign produce revenue?
Advertising became less interested in merely displaying a message.
It became interested in constructing a chain of attribution.
This produced genuine commercial benefits.
Waste could be reduced.
Campaigns could be adjusted quickly.
Businesses could identify which messages worked.
Publishers could demonstrate the value of their audiences.
Consumers might see advertising more closely connected to their needs.
The same process also meant that more human behaviour had to be observed.
Better attribution required more data.
Better targeting required more classification.
Better predictions required more history.
Advertising efficiency and behavioural visibility grew together.
The Advertisement Separated From the Page
Early web advertising remained closely connected to the publisher selling it.
A website created space.
An advertiser bought the space.
The advertisement appeared.
As the online market expanded, intermediaries entered the relationship.
Advertising networks aggregated inventory from many publishers.
Advertisers could reach audiences across multiple websites without negotiating separately with every publisher.
Ad servers decided which advertisement to display.
Data-management systems assembled information.
Demand-side platforms helped advertisers buy opportunities.
Supply-side platforms helped publishers sell them.
Ad exchanges brought the two sides together.
The advertisement was no longer chosen solely by the publisher of the page.
It could be selected by an external system using information collected elsewhere.
This separation was essential to the development of behavioural advertising.
The same intermediary could observe activity across many participating websites and use that information when deciding what to display next.
DoubleClick became one of the most important early companies in this infrastructure.
In 2007, Google agreed to acquire DoubleClick for $3.1 billion. The Federal Trade Commission investigated the transaction for eight months before deciding not to block it on competition grounds. The Commission explicitly acknowledged that the merger had raised consumer privacy concerns, while noting that those concerns extended across the wider online advertising marketplace and fell outside the antitrust basis upon which it was reviewing the acquisition.
One FTC commissioner dissented, warning that the combined company could play a transformative role in the future of online advertising and that the privacy interests of consumers had not been adequately addressed.
The importance of the acquisition was not that Google suddenly invented targeted advertising.
It connected major parts of the machinery.
Search advertising.
Display advertising.
Publisher tools.
Advertiser relationships.
Audience information.
Measurement.
Infrastructure.
The market was moving beyond adverts placed on individual websites.
It was becoming an interconnected system through which the behaviour of users, the inventory of publishers and the objectives of advertisers could be matched at enormous speed.
The Auction Nobody Sees
Open almost any advertising-funded website today and a remarkable process may take place before the content has fully appeared.
The page creates an opportunity to show an advertisement.
Information connected to that opportunity is assembled into a bid request.
Potential advertisers evaluate it.
Automated systems decide whether the person appears relevant to their objectives.
Bids are submitted.
A winner is selected.
The advertisement is delivered.
The user sees a page.
Behind the page, a market has operated.
This process is known as real-time bidding.
The UK's Information Commissioner's Office describes real-time bidding as the buying and selling of advertising inventory on an impression-by-impression basis, normally through an auction conducted in the time it takes a webpage to load.
The word inventory is revealing.
From the publisher's perspective, the empty advertising position is inventory.
It is an asset available for sale.
But an empty rectangle on a page has limited value by itself.
Its value depends upon who is about to see it.
A luxury travel company may pay more to reach someone believed to be researching expensive holidays.
A local tradesperson may value someone located nearby.
A retailer may want to reach a person who viewed a product but did not buy it.
A political campaign may seek a particular location or demographic.
The opportunity becomes more valuable as the system becomes more confident about the person behind it.
The ICO found that bid requests could include identifiers, IP-address information, browser and device details, location, time zone, language and audience segmentation. Other connected information could include search queries, previous sites, mouse movement, scrolling, clicks, media views, session time and demographic information.
Not every bid request contains every possible category.
Not every advertiser receives a named biography of the user.
Not every piece of information is accurate.
But the commercial process no longer asks merely:
How much is this advertising space worth?
It asks:
How much is the opportunity to influence this particular type of person, in this particular context, at this particular moment, worth to us?
That is a fundamentally different market.
What Is Actually Being Sold?
The phrase your data is being sold is often used loosely.
Sometimes organisations do sell or license datasets.
Often the transaction is more complex.
A platform may retain the underlying data while using it to offer advertisers targeted access to categories of people.
An advertiser might never receive a person's complete browsing history.
It may simply ask the platform to show an advert to people whose behaviour resembles a desired audience.
The platform keeps the profile.
The advertiser purchases the opportunity.
This is why saying that users themselves are the product can be misleading.
The person is not handed to an advertiser.
Nor is attention placed into a container and transferred like oil.
What is sold is access.
The right to place a message in front of someone.
The selection of that person from a larger population.
The timing of the intervention.
The prediction that they may respond.
The measurement of what they do next.
The commodity is therefore not attention alone.
It is a commercially packaged opportunity to direct attention and influence behaviour.
That opportunity may contain several layers of value.
There is the space on the screen.
The likelihood that the person will notice it.
The information used to select them.
The prediction that the advertisement is relevant.
The possibility of changing a decision.
The ability to measure whether the intervention worked.
Advertising has always attempted to influence human behaviour.
Digital advertising industrialised the selection, timing and measurement of that influence.
The Profile Does Not Need to Be Correct
A profile can be commercially useful without being perfectly accurate.
Suppose an advertising system classifies someone as interested in running.
Perhaps they searched for trainers.
Read a marathon article.
Watched a fitness video.
Visited a sports retailer.
The person may genuinely be a runner.
They may also be buying a present.
Researching an article.
Helping a friend.
Clicking out of curiosity.
The system does not need certainty.
It needs a probability greater than random selection.
This is another reason the word identity can distract from what is happening.
The advertising industry does not always need to understand you as a human being.
It needs a model that performs slightly better than showing the same message to everyone.
A category that improves response by a small percentage can become enormously valuable when applied across billions of advertising opportunities.
That profile may be incomplete.
It may be wrong.
It may contradict how you understand yourself.
None of that prevents it from affecting what you see.
The profile becomes part of your environment whether or not it accurately represents your life.
That creates a strange asymmetry.
The system forms conclusions about you.
You may never know what those conclusions are.
It changes what appears in front of you.
You cannot easily inspect why.
It learns from your response.
You receive the result of an invisible interpretation.
The person is known statistically while remaining largely ignorant of how they have been classified.
Anonymity Did Not Remove the Privacy Problem
For years, the advertising industry often distinguished between personally identifiable information and data linked only to a browser, device or pseudonymous identifier.
The implication was reassuring.
The system might know that identifier 78294 had viewed particular pages.
But it did not necessarily know the person's name.
Therefore, the person remained anonymous.
The FTC warned that the distinction was becoming increasingly difficult to sustain.
Its 2009 report defined behavioural advertising as tracking an individual's online activities to deliver tailored advertising. It noted that the practice was commonly invisible, that profiles could be attached to a computer rather than a traditional identity, and that information not initially considered personally identifiable could later be combined with names, addresses or other identifying data.
The report also distinguished contextual advertising from broader behavioural tracking. Advertising based only on the page currently being visited or a single search query, without retaining a wider history, was considered less invasive than tracking activity across time and sites.
This distinction remains valuable.
Not all digital advertising depends upon extensive surveillance.
A website can display advertising connected to its content.
A search engine can show an advert related to a current query.
A publisher can sell space directly.
The attention economy did not become intrusive simply because advertisements moved online.
It became intrusive as the system increasingly sought to observe people across contexts, retain those observations, combine them and use them to predict future behaviour.
The issue was not advertising alone.
It was the construction of the profile.
The Market Began Operating Faster Than Human Understanding
Real-time bidding compressed the entire advertising transaction into milliseconds.
The user opens a page.
The market evaluates them.
The advert appears.
They may never know an auction occurred.
They may not understand which organisations received information.
They may not know why that particular advert was selected.
They may not remember agreeing to anything beyond clicking a button on a consent banner they wanted to remove as quickly as possible.
The ICO concluded in 2019 that a single real-time-bidding request could result in personal data being processed by hundreds of organisations. It also described detailed profiles being repeatedly shared without individuals' knowledge and raised systemic concerns about transparency, security, consent and data protection compliance.
The scale matters.
A human being cannot realistically investigate hundreds of intermediaries before deciding whether to read an article.
The transaction is too fast.
The ecosystem is too complex.
The information is too fragmented.
Even organisations participating in the system may struggle to identify every party and every flow of data. The ICO reported that the complexity of the ecosystem made it difficult even for market participants to remain clear about which organisations performed which roles.
Consent under those conditions becomes philosophically difficult, even before considering the legal standard.
A person may technically receive a privacy notice.
That does not mean they understand the market opening behind the page.
There is a difference between being given information and being meaningfully informed.
The Benefits Were Real
It would be easy to describe this entire development as theft.
That would be emotionally satisfying.
It would also be incomplete.
Targeted digital advertising created genuine benefits.
It helped fund search engines, social networks, journalism, entertainment, maps, email and countless other services without requiring every user to pay directly.
It allowed small businesses to reach customers without purchasing expensive national media campaigns.
It made advertising more relevant.
It reduced some forms of waste.
It enabled publishers to generate revenue from audiences spread across the world.
It helped people discover products and services they genuinely wanted.
It allowed campaign performance to be measured and improved.
The FTC repeatedly acknowledged these benefits while examining behavioural advertising. Its reports noted that targeted advertising could support free content, improve relevance and allow businesses to spend advertising budgets more effectively.
The ICO also recognised that advertising technology could help advertisers reach new audiences, reduce campaign costs and improve measurement, while allowing publishers to attract more buyers and generate greater revenue from available advertising space.
The problem is not that the system produced no value.
The problem is that value became dependent upon a progressively deeper ability to observe, categorise, predict and reach human beings.
The same infrastructure that makes advertising more useful can make people more transparent to commercial systems.
The same measurement that reduces waste can encourage more measurement.
The same personalisation that improves relevance can make manipulation more precise.
The benefits and risks are not separate stories.
They grew from the same machinery.
The Publisher's Product Quietly Changed
A traditional publisher created content and sold advertising around it.
A digital publisher still does that.
But something changed once every advertisement could be individually selected, measured and auctioned.
The publisher was no longer selling only space around content.
It was producing a stream of advertising opportunities.
Each page view created inventory.
Each returning user created another opportunity.
Each additional minute might generate more impressions.
Each click created more behavioural information.
Each piece of information could help make future opportunities more valuable.
The content still mattered.
It attracted the audience.
But the audience's behaviour increasingly became part of what made the business commercially valuable.
This changed the economic relationship between publishers and users.
A useful article satisfied the reader.
A successful advertising product also needed to create monetisable exposure.
Sometimes those objectives aligned.
A good article attracted more readers.
More readers created more advertising revenue.
The money funded more journalism.
Everyone benefited.
Sometimes they diverged.
A page divided into multiple sections could generate more advertising impressions.
A sensational headline might attract more clicks.
A deliberately incomplete story might encourage another page view.
An autoplaying video might create additional watch time.
An endless feed might remove the moment at which the person would otherwise decide to leave.
These later design choices will be examined elsewhere in this series.
For now, the important change is economic.
Once attention became measurable and monetisable, consuming more was no longer merely something users did.
It became something businesses had reason to encourage.
The Scale Reveals What Became Valuable
In 2025, digital advertising revenue in the United States alone reached approximately $294.6 billion, according to the Internet Advertising Bureau and PwC.
Alphabet reported that more than 70 per cent of its 2025 revenue came from online advertising.
Meta reported that its advertising revenue increased by more than $35 billion during 2025, driven by increases in both the number of advertisements delivered and the average price of each advertisement. The company also linked growth in impressions partly to increases in users and engagement across its products.
These figures do not prove that every use of these products is harmful.
They do not prove that every advertisement is manipulative.
They do not establish that technology companies secretly sell complete personal dossiers to every advertiser.
They demonstrate something simpler.
The ability to place commercial messages in front of human beings, selected and measured with extraordinary precision, has become one of the most valuable businesses ever created.
The resource underpinning that market is not the screen.
Screens are cheap.
It is not the advert.
Adverts are abundant.
It is not information.
Information is almost limitless.
The scarce element is the person capable of noticing.
The human mind with a finite number of waking hours.
The customer who might act.
The child who might form a preference.
The voter who might change an opinion.
The viewer who might remain for another video.
The reader who might open another page.
The user who might return tomorrow.
The market does not manufacture attention.
It competes for access to the attention that already exists.
Commodification Changes the Question
Calling something a commodity does not mean it is identical to wheat, copper or oil.
Human attention cannot be extracted, stored in a warehouse and sold later.
The term describes what happened when attention became standardised into units that markets could price and exchange.
An impression.
A click.
A completed view.
A conversion.
A minute watched.
A user segment.
A bid request.
A predicted probability of action.
These measurements transformed something subjective and human into something commercially legible.
Once attention could be represented through units, those units could be compared.
One placement performed better than another.
One group responded more frequently.
One moment produced more purchases.
One platform delivered a higher return.
One person appeared more valuable to a particular advertiser than someone else.
Commodification did not require companies to understand the meaning of a person's attention.
It required them to measure its commercial effect.
That distinction is central.
A father listening to his daughter gives attention.
A reader becoming absorbed in a novel gives attention.
A scientist studying an unexplained result gives attention.
A person compulsively refreshing a feed gives attention.
Humanly, these experiences are profoundly different.
Commercially, they may be reduced to time, activity, retention and response.
The market does not naturally distinguish between attention that enriches a life and attention that merely remains available for monetisation.
It measures what it can count.
No Single Moment Completed the Transformation
So when did human attention become a commodity?
Was it 1994, when the first banner advertisements made online response visible?
Was it when cookies allowed browsers to be recognised across visits?
Was it when advertising networks began connecting behaviour across unrelated websites?
Was it when search auctions attached prices to expressions of human intention?
Was it when cost-per-click connected revenue directly to action?
Was it when companies began constructing behavioural profiles?
Was it when advertising spaces became auctioned automatically for individual users?
Was it when engagement itself became a central commercial metric?
There is no single answer.
The transformation occurred in stages.
Attention became an audience.
The audience became an impression.
The impression became a click.
The click became a history.
The history became a profile.
The profile became a prediction.
The prediction entered an auction.
The auction placed a price upon the opportunity to influence what someone did next.
Each step appeared rational.
Most offered genuine benefits.
Together, they changed the economic structure of the internet.
The human being ceased to be merely the reader, viewer or user of the service.
Their behaviour became part of the infrastructure funding it.
The Hidden Price of Free
People often say that online services are free.
Many are free at the point of use.
That is not the same as operating without an exchange.
The user receives communication, information, convenience or entertainment.
The company receives opportunities to display advertising, observe behaviour, improve prediction and create further advertising opportunities.
No money needs to leave the user's bank account.
Value still moves.
This does not automatically make the exchange unfair.
A person may reasonably prefer to see advertising rather than pay a subscription.
A small business may benefit enormously from targeted marketing.
A publisher may be unable to survive without advertising revenue.
The deeper problem is that most users were never invited to negotiate the terms of the exchange.
The infrastructure developed gradually.
The tracking expanded invisibly.
The number of intermediaries increased.
The privacy policies lengthened.
Consent became another obstacle to clear before reaching the thing someone actually wanted.
The market became more sophisticated than the public understanding surrounding it.
We learned to use the services.
We did not necessarily learn how the services used us.
Final Thought
Human attention did not become valuable when the internet arrived.
It had always been valuable.
The internet made it measurable.
The click connected exposure to action.
The cookie connected one action to another.
The profile turned those actions into an inferred person.
The auction allowed advertisers to compete for access to that person in real time.
At first, advertising occupied space on a webpage.
Eventually, the webpage became the entrance to a market.
Behind it, systems gathered information, classified behaviour, calculated value and selected which commercial message should appear before which human being.
All before the human being had finished looking at the page.
That is the moment the attention economy became more than advertising.
The commodity was never the person.
It was never quite the data.
It was not even attention by itself.
It was the increasingly precise opportunity to capture that attention, direct it towards a commercial objective and measure whether the intervention succeeded.
Once that opportunity could be priced...
Human attention had entered the market.
Sources & Research Gaps
Principal Sources
Wired and the First Banner Advertisement
Wired's retrospective accounts document HotWired's October 1994 launch and the appearance of the first widely recognised web banner advertisements, including the AT&T campaign.
Internet Engineering Task Force, RFC 2109
The 1997 technical standard explains how cookies enabled stateful sessions by passing information between web servers and browsers. The source establishes their technical purpose but does not imply that cookies were originally designed for cross-site behavioural surveillance.
Federal Trade Commission Reports on Online Profiling and Behavioural Advertising
FTC reports and testimony from 2000 and 2009 describe online profiling, cookies, network advertising, behavioural targeting, consumer benefits, privacy concerns and the increasingly unclear distinction between personally identifiable and pseudonymous information.
Google 2004 Registration Statement
Google's SEC filing provides primary evidence concerning the introduction of its advertising products, the move to cost-per-click pricing, the role of auctions, the importance of click-through rates and the proportion of company revenue generated by advertising at the time.
Google and DoubleClick
Google's announcement and the Federal Trade Commission's investigation document the $3.1 billion acquisition of DoubleClick, the competition analysis and the privacy concerns raised at the time.
UK Information Commissioner's Office Report on Adtech and Real-Time Bidding
The ICO's 2019 investigation explains real-time bidding, bid requests, advertising intermediaries, data categories, auction speed and the regulator's concerns about transparency, consent, security and the processing of personal data throughout the adtech ecosystem.
IAB/PwC Internet Advertising Revenue Report
The report documents United States digital advertising revenue of approximately $294.6 billion during 2025.
Alphabet and Meta Corporate Filings
The companies' 2025 reporting provides current evidence concerning their continuing dependence upon advertising revenue, advertising impressions and user engagement.
Research Gaps and Limitations
The history of online advertising does not follow a single clean timeline. Different technologies, companies and pricing systems developed in parallel.
The first banner advertisement is used as a symbolic starting point. It was not the first time human attention had been sold commercially, nor did it create behavioural advertising on its own.
The presence of a cookie does not automatically indicate tracking or advertising. Cookies remain necessary for many ordinary website functions.
Behavioural profiles vary enormously in quality and depth. Some are based on extensive cross-site histories, while others use limited contextual or first-party information.
Real-time bidding is one form of programmatic advertising. Not every digital advertisement is purchased through an open real-time auction.
A bid request can contain many categories of information, but this does not mean every request contains every possible field or that every participant receives identical data.
The ICO's detailed findings relate to the adtech market it examined in 2019. Technologies, legal interpretations and industry practices have continued to evolve, although the fundamental auction and profiling mechanisms remain commercially significant.
Corporate revenue figures establish the financial importance of advertising. They do not prove that every individual product decision is designed solely to maximise attention.
More research is needed into:
- the modern shift from third-party cookies towards first-party data, device fingerprinting and platform-controlled identity systems;
- how accurately behavioural profiles describe individuals;
- the economic value of personalised advertising compared with contextual advertising;
- the number of organisations receiving or processing information during contemporary advertising auctions;
- the effect of privacy regulation upon actual tracking rather than stated compliance;
- whether users meaningfully understand the exchange they enter when using advertising-funded services;
- how artificial intelligence is changing the creation, targeting and optimisation of advertising.