The short version
When you open a page that carries ads, there is often a very quick process behind the scenes that decides which ad fills each slot. The slot is offered to advertisers, who bid according to how valuable they think you are as an audience, and the winning ad is delivered. All of this typically happens while the page is loading.
Not every ad is sold this way. Some are booked directly between a publisher and an advertiser, and some show whatever the site’s owner chooses. The auction-based route, often called programmatic advertising, is the part most closely tied to personal data.
The players
- Publishers are the sites and apps with space to sell.
- Advertisers want to reach an audience, such as a shop promoting a product.
- Ad networks, exchanges and platforms act as middlemen. Some represent publishers, and others represent advertisers.
- Data providers supply audience information to improve the targeting.
Large platforms that run both content and advertising, such as social media services and search engines, can run the whole process within their own systems, using the information that people provide while using them.
From page load to screen
The sequence is roughly as follows.
- The page loads and requests an ad. Code on the page signals that a slot is available.
- A bid request is sent. The request describes the opportunity: the page or app, the type of slot, the device and location in broad terms, and an identifier if one is available.
- Advertisers evaluate it. Their systems consider whether this slot matches a campaign and how much to bid.
- An auction picks a winner. The highest valid bid, adjusted by factors such as predicted quality, typically wins. The details differ between exchanges.
- The ad is delivered. The winning creative is sent to your browser or app, and the page shows it.
- Measurement follows. Impressions, clicks and sometimes later purchases are recorded to report results and to refine the next round.
A single page with several ad slots can trigger many of these exchanges. Because the bid request can be sent to many companies, including those who do not win, information about the visit may reach more parties than the one whose ad you see. That wide sharing is a recognised privacy concern.
Signals that influence the choice
Different campaigns use different signals, and they can be combined.
- Context. The topic of the page. A hiking article may attract outdoor-gear ads whether or not anything is known about the reader. This works without personal data.
- Basic technical details. Device type, language, and approximate location derived from the network address or from phone settings.
- Time and place. Ads for a restaurant near lunchtime in the right city.
- Interest profiles. Categories based on past activity, such as “interested in travel”, built from online tracking across sites and apps or from activity within one platform.
- Retargeting. Ads for something you viewed before, such as a product left in a basket. This is the reason an item seems to follow you around.
- Purchased or matched data. Customer lists or segments from data brokers can be matched to your identifier or email address.
- Lookalike audiences. Advertisers ask a platform to find people who resemble their existing customers, which relies on statistical models and machine learning.
Many of the assumptions in a profile are predictions. A person who researched a gift once may be shown ads for that category for weeks, which suggests the systems are less precise than they can feel.
Do the phones listen to you?
People often suspect their device is listening to conversations when a product they only mentioned suddenly appears in ads. Permission settings do give some apps microphone access, and misuse would be a serious matter. In practice, though, many such experiences can be explained by other signals: shared networks, searches by people in the same household, location, or simple coincidence combined with how often we notice ads that match. It is hard to be sure in any individual case, and reviewing microphone permissions is a reasonable precaution.
Why advertisers use it
Targeting is intended to make advertising cost less per useful result and to fund free services. Critics point out that it relies on extensive data collection, can allow exclusion or manipulation in sensitive areas, and can be difficult for people to see or challenge. Some jurisdictions restrict targeting based on sensitive information or on minors, and many require consent for certain tracking, in line with data protection laws.
The same logic shapes content as well. The ranking of what appears in a feed has similarities to recommendation algorithms, although those aim to keep you engaged and ads aim to get a response.
Controls you have
- Use the ad settings in your accounts to review interest categories, turn off personalisation or remove topics. Names and options vary by service.
- Reset or limit the advertising identifier on your phone.
- Block or limit third-party cookies and use tracking protection, or a private browser with it enabled.
- Respond to consent banners according to your preference, bearing in mind that declining personalisation usually does not remove ads, only makes them less tailored.
Turning off personalisation changes which ads you see, not how many. It also does not erase data already collected.