The modern internet and social media is saturated with advertisements. Even for those of us who use adblockers, there are some mediums where the ads are inescapable. Many of the ads we see fade into the background noise of content that we are exposed to throughout our day. However, every now and then we might see an ad for something we could swear we only thought about, only spoke about with a close friend offline, or only just became aware of outside the context of the internet. Ads like this give us the sense that we’re somehow being watched, that our innermost thoughts and conversations are readable by the algorithm and useful for serving up ads. While precognition and telepathy may not yet be possible for the Silicon Valley giants, they’ve developed some clever ways to come close.
The fuel for modern advertising is user generated data. Entire industries have emerged over the past twenty years that focus on collecting, consolidating, enriching, and selling data about each and every one of us. The more “plugged in” we are, the more data can be harvested about us. Where once advertisers only knew our age and profession, they now can see our income, family and friend connections, shopping habits, interests, likes and dislikes, political leanings, life events (like marriage, parenthood, promotion, etc), and much more. This data leaks out of us through the devices we wear and carry, and the apps, services, and websites we use every day.
Companies like Meta, Google, and Amazon have access to enormous amounts of data. They use that data, along with machine learning and AI, to analyze patterns and behaviors. This helps them identify which groups of people fit a particular advertising need and decide when, where, and how to show those people ads.. This method of “calculating” where to put ads falls within the larger (and growing) field of “computational advertising”.
There are several core targeting techniques at work in advertising today. We’ll start with those that are more straightforward and relatable, then work our way into the more abstract methods.
Demographic targeting is the most straightforward because it most closely matches the way advertising has traditionally functioned. If you know the age, gender, and income of the people who typically buy your product, you’ll want to advertise to more people with similar characteristics in order to grow your business. So if you sell a hiking boot that seems to be popular with men, aged 25-30, with a median income, who live near the mountains, you’ll want to target your advertisements to reach more of this group.
With a smartphone in your pocket or a smartwatch/fitness tracker on your wrist, your GPS location is almost always available to companies looking to advertise to you. This allows companies to target their advertisements based on where you are and what time it is. Typically, this means that the company has considered when and where you will most likely engage with the ad. For example, a company that sells breakfast sandwiches likely won’t flood your social media feed or the websites you visit with ads if you’re at home in the evening. However, if you’re on the move, say commuting to work in the morning, the company would use location targeting to show ads that might bring you to one of their restaurants.
Relevance and timeliness is key in advertising. Advertisers don’t want to waste their dollars displaying their ads to people who aren’t thinking about or aren’t interested in their product. Contextual targeting detects the content of the article or site you’re browsing using natural language processing, then uses this information to only display ads that match the content they appear alongside. For example, if you’re reading an article on classic cars, you would likely see ads for automotive tools, car auction sites, or classic car specialist mechanics.
Online tracking mechanisms enable behavioral targeting. “Cookies” are the most familiar term for mechanisms that track us online, but the industry has developed a suite of tools and techniques that enable this. To target based on our behavior, advertisers are particularly interested in our browsing history, search keywords, clicks, likes, and past ad-engagements. This information gives a sense of what each of us individually will engage with and when. Social media has revolutionized this type of targeting by offering free services where users are encouraged to voluntarily disclose their personal information, interests, preferences, and opinions. This creates a situation where advertisers have a wealth of highly personalized information to target individuals with.
With known data on their current customer base’ demographics and behaviors and sophisticated tracking information on the demographics and behaviors of other internet and social media users, advertisers can create “Look-alike” audiences for targeting. This type of targeting goes well beyond what can be achieved with demographics alone, and allows advertisers to identify the people who think, behave, browse, and buy in similar ways to their current customers.
A key component that brings the rich troves of data and powerful targeting mechanisms into a creepy crescendo is real-time bidding (RTB). Real-time bidding is how any site or app that serves ads actually earns money from their ad space. When you scroll through a social media app or a website, the next ad you see isn’t determined until you scroll up to it. As you use the service, an auction is continuously taking place where advertisers bid to show you their ads. If an advertiser wins their bid, they get to display their ad. Advertising platforms take this a step further by allowing advertisers to retarget, refine, and optimize their advertising campaigns.
So what is behind those moments of uncanny timeliness/precognition in advertising? Is it an artifact of a single point of data or a single targeting technique? Or is it a combination of factors that emerge periodically, almost like deja vu, to create moments of strangeness?
The reality is that advertisers and social media platforms have access to unimaginable amounts of data on each of us and can leverage immensely powerful computational techniques to model how we will behave day to day. When we discuss a new topic with a friend, or encounter a topic offline, our minds perceive this as a serendipitous, novel encounter. To the machine brain of modern advertising though, these interactions all exist on a scale of predictability. What we might or might not do is, in some ways, more known to advertisers than it is to ourselves. Advertisers don’t read our minds – they predict them, using patterns we can’t see.
This system isn’t perfect though. Our behavior as individuals is hugely complex and hard to predict. This means that, most of the time, we see advertisements that make sense and feel natural within the context of what we’re interested in and what we’re browsing. On rare occasions though, hundreds of factors align and computational advertising hits closer to home than feels natural.
Moments of creepy advertising online are uncomfortable to experience, but valuable to acknowledge. Silicon Valley and the digital advertising industry have morphed into a machine that collects and knows far more about each of us than it should. The fact that their algorithms sometimes strike us as creepy is a good reminder that they are, indeed, very creepy. When we consent to use a free app, is it fair that we’re by proxy consenting to such an intense level of data collection, profiling, and targeting? Would any of us agree to use these apps if we truly understood how much they know about us?
Asta. (2025). Assisted research and analysis for article preparation. Allen AI. Retrieved October 14, 2025, from https://asta.allen.ai
Google (2025). [Fig.1 Revealing identity through data]. Images generated by Gemini. Gemini. Retrieved October 14, 2025, from https://gemini.google.com/