Much of the user excitement in the early years of the Internet came from the medium’s powerful way in which it eliminated constraints on human communication. You could find your friend from high school, let the world know about an irritating customer experience, become well-known, and even learn skills from experts you didn’t know – and best of all, it was all for free! Those magical experiences have disappeared – to be replaced with what?
Researchers at Amazon Web Services examined the consequences of free or low-cost AI tools, such as machine learning for translation, and came to an astonishing conclusion. Over half the sentences on the web have been translated into two or more languages, with sentences aimed at consumers in much of Africa and the Global South being machine-translated into click-bait garbage, presumably to garner advertising revenue.
They found that much of the Internet (57.1% of all sentences) was translated, and that “low resource” countries were far less likely to generate enough training data using those languages to be accurate. The result? Most of what a typical user in much of the world encounters is inaccurate at best, misleading or outright wrong at worst.
The dead internet
It’s partly conspiracy theory, partly science fiction, and today reflective of an increasingly disorienting internet experience. It refers to an Internet in which most of the interactions are not between human users but between AI-enabled bots of varying degrees of skill. In its more conspiratorial incarnations, the dead Internet idea holds that governments have formed a giant mind-control cabal and used the power of the Internet to supply only the information that they want us to have. The idea of our largely dysfunctional government summoning enough cooperation to actually accomplish something on this scale is farfetched. Another less monolithic force, however, may bring the same outcome into being.
This is the race between technology providers and those with the talent to game their systems. At one point, in 2013, employees at YouTube were concerned that a large group of people were gaming likes and views on the platform, to the point where their algorithms would reach “the inversion” – the switch in which the algorithms decide that the fake material was real, and the human-provided content wasn’t. The cat-and-mouse continues to this day, as people conjure up fake followers, likes, comments, and more. A recent study of Instagram for instance found that “nearly two out of three (58.5 per cent) Instagram profiles in India have spurious or fake followers exceeding 60 per cent.”
Are advertisers getting what they are paying for?
In 2018, the Justice Department unsealed indictments against 8 people accused of siphoning off $36 million in ad spending that essentially fooled the advertisers into thinking that their ads were both attracting views and appearing on the pages of high-quality, legitimate publications such as Vogue or The Economist. The indictment revealed a truly toxic mix of abusive practices from the point of view of the advertisers. Spoof websites intended to fool advertisers into thinking their ads were on legitimate sites. Fake views generated by human impersonators. Faked clicks, mouse movements, and social network login information masquerading as engaged human consumers. Manipulation of geolocation information associated with the IP addresses under their control. Special case countermeasures against code from over a dozen different ad tech companies.
Determining whether information provided about activity taking place on the Internet seems subject to fakery. This is not a new problem – in 2016, a group of advertisers sued Facebook for inaccurately providing information about user behaviour on the platform – overstating by a significant amount the time users spent on it.
The obfuscation continues. Ad transparency tools were mandated by the European Union’s Digital Services Act, which requires that large tech platforms maintain ad libraries and other tools, such as application programming interfaces (APIs), to be used by researchers and the public. The idea is that people should be able to use the tools to search for information about ads they see or specific companies’ ad campaigns, including ad content, target audience, the ad’s reach, and the call to action. Unsurprisingly, researchers have observed that the big tech firms are dragging their feet. As one observer noted, “None of the results were great, and some were ‘a major disappointment.’ ”
Some are arguing that the proliferation of Internet advertising without a commensurate ability to know what effect the ads are having is creating a bubble. As Paris Marx writes, the market has reached an “irrational level of confidence,” a regular feature of financial crashes going back hundreds of years. This is because beliefs about the underlying value of an asset and its real value become disassociated. As he writes, “even as the value of the digital ad industry was continuing to rise, the average clickthrough rate on Google’s display ads fell to 0.46% in 2018, ad fraud was expected to jump 21% to $42 billion in 2019, and a Google study found 56% of its display ads may not even be seen by a human.
Vice, once valued at $6 billion, went bankrupt. BuzzFeed News shut down. As former New York Times editor Jill Abrahamson observed, “It turned out that advertising was a bad bet. With the change of an algorithm, Facebook and Google slashed Vice and BuzzFeed’s massive audiences and hoovered up the bulk of digital advertising payments. Their young, hip followers were not willing to pay for their periodic scoops.”
Media observer Nick Manning argues that the beginning of the end of the Internet ad bubble is already evident. As he puts it, “We are reaching the end of a long period of meteoric but unrealistic growth for some online media properties as a combination of higher costs and lower revenues brings hard reality to a sector that has been overblown for several years, buoyed up by insane levels of adspend not justified by effectiveness nor efficiency.” The stocks of ad-tech companies are way down, cookies are disappearing and as he puts it, “personalisation at any level is erratic, and personalisation at scale is a myth.”
The Cory Doctorow Maxim
From a user’s point of view, relentless automated ad proliferation is also causing major problems. A 2022 article with the provocative title “Is Google Getting Worse? Why critics say ads, spam sites are killing search,” concluded that the answer for many users is “yes.” An influential blog post by Dmitri Brereton concludes that “Google search results are clearly dying…. most of the web has become too inauthentic to trust.” Daniel Gross, a Silicon Valley entrepreneur and keen observer of the tech scene observes that “In 2000, Google got popular because hackers realised it was better than Lycos or Excite. This effect is happening again. Early adopters aren’t using Google anymore.”
What they are doing is appending “reddit” to their Google searches. As Brereton says, “You append “reddit” to your query (or hacker news, or stack overflow, or some other community you trust).” This behaviour invalidates the advertising-based model for search entirely, as people now turn to trustworthy communities populated by real people. Kate Lindsay of The Atlantic argues that “something went terribly wrong with Internet ads.” She concludes that we are facing “an internet adpocalypse that has become impossible to escape.”
Cory Doctorow, science fiction author, internet privacy activist, and keen observer of the tech scene memorably coined the phrase “enshi#!fication” to characterise the decay (for users) of technological platforms over time. As he writes, “So what’s enshi#!fication and why did it catch fire? It’s my theory explaining how the internet was colonised by platforms, why all those platforms are degrading so quickly and thoroughly, why it matters and what we can do about it. We’re all living through a great enshi#!fication, in which the services that matter to us, that we rely on, are turning into giant piles of shit. It’s frustrating. It’s demoralising. It’s even terrifying.”
He argues that the process goes through 3 predictable stages. In stage 1, platforms are good to their users and gain their trust. In stage 2, they gradually start abusing that user trust to make more money from their business customers. In stage 3, they then start abusing the business customers, too, using monopoly or pseudo-monopoly power to claw back all the value for themselves. His case study on Facebook is illustrative. Doctorow argues that this process will only stop when 4 forces are harnessed: regulation, competition, self-help, and worker activism.
Whither the web?
Bots talking to bots. A degraded user experience. A growing lack of trust in what the Internet is offering us. More and more valuable content behind paywalls. Where is this likely to lead?
One trend that is already evident for advertisers is presence in omni-channels – combining both the real world and the digital one. Taylor Swift’s recent approach to marketing her new album, The Tortured Poet’s Department, is being hailed as an example of marketing genius. Departments that used to be budgeted separately (such as digital out-of-home and physical in-store displays) are now being joined up by advertisers in new combinations.
In advertising-supported search, we are already seeing evidence of new competitors such as Perplexity, promising to create a better experience than that offered by competing search engines.
Subscriptions are rising in popularity in some segments and have proven a durable basis for long-term success. The New York Times, for example, now derives the bulk of its revenue from people who pay to subscribe – taking control into its own hands and radically reducing its dependence on advertising and algorithms.
More radical ideas include thinking of the evolution of the Internet in the same way governments thought of the evolution of electricity provision. Given how essential electricity was, many determined that it was unsuitable to remain in private hands, and electricity companies became regulated utilities. Without generous ad revenue, that may be another way that Internet services are provided (not a popular idea with current private-sector owners of so much of this critical infrastructure). Like railroads, highways, or the air traffic control system, the Internet is no longer optional for modern life. Many feel this is a foregone conclusion. A Consumer Reports study found that 80% of consumers they surveyed said that broadband is as essential as water and electricity.
We are also likely to see the rise of curated peer-to-peer or trust-based communities. Reddit is a current example, but we can imagine many more curated communities whose members trust one another to become prized by advertisers seeking to have authentic communications with them. After all, as Daniel Pink has long argued, selling at its best is a deeply human experience.
Finally, we may see digital services take on different structural forms, now that the seemingly endless spigot of growth-at-all-costs VC money seems to have run out (for everything except AI, of course). Imagine if Uber drivers enjoying shared ownership of the platform that generates their income? You bet they would implement different policies than shareholder-return-driven owners. A company called The Drivers Cooperative is already taking on Uber and Lyft, claiming that their drivers make 8-10% more than drivers for those firms and that profits are shared by the driver-owners.
We’ve also seen the power of well-designed programems of ownership. It’s been demonstrated by cases such as that of CHI Overhead Doors, a novel approach pioneered (very profitably, I would add) by KKR.
We may perhaps see a blend of all these approaches. It seems very clear, however, that the weak signals are getting increasingly strong that the advertising-based model of the Web that we’ve become so used to is on the brink of a major inflection point.