Behavioral targeting: what it is and when contextual is safer

Behind every paid campaign is a targeting decision. Should the ad follow people who have already shown interest, or should it appear where the right people are reading and watching? The first approach is behavioral targeting. It can make each impression more relevant, but it depends on tracking data that browsers, platforms and privacy laws increasingly restrict.

This guide explains what behavioral targeting is, how it differs from contextual targeting and retargeting, and what the research says about its value. It then sets out the legal and platform limits a marketing leader needs to know, when each approach is the better choice, and how to test them against each other before moving budget.

DisclosureinBeat, which is mentioned in this article, is part of Fieldtrip, a group of marketing agencies in which Jake Jorgovan is a partner.

What behavioral targeting is

Behavioral targeting chooses who sees an ad, or what a website shows them, based on what they have done: pages viewed, products added to a cart, searches, purchases or earlier responses to ads. The data is grouped into segments, such as people who viewed the pricing page but did not buy, or customers who have not ordered for a while, and each segment gets its own message.

Four forms are worth separating:

• On-site personalization. A website changes what a visitor sees based on that visitor's activity on the same site, such as recommending products related to the ones already viewed.
• Cross-site (network) targeting. An ad network or platform builds a profile from a person's activity across many sites and apps, and advertisers buy segments from it.
• Retargeting. Ads reach people who have already interacted with your site, app or ads. One campaign can focus on people who did not buy or sign up; another can re-engage existing customers. Search retargeting does the same for people who searched for related terms.
• Customer lists and lookalikes. You upload a list of your own customers or leads to a platform to reach them again, or to find people who resemble them.

On-site personalization and customer-list campaigns can start from data your business collects directly, called first-party data. That describes where the data comes from, not permission to use or share it for advertising. Cross-site targeting depends on tracking a person across other companies' sites and apps, which is where most of the legal and technical restrictions below apply.

Behavioral vs contextual targeting

Contextual targeting places an ad according to the content around it. An ad for running shoes appears beside a marathon training article, whoever is reading it. It needs no profile of the person and no record of where they have been.

Image

Zoom

The approach described in this article, shown as a visual guide.

The two can be combined. One way uses contextual placements to reach people who are new to the brand, and behavioral signals from a company's own site and customer data to follow up with people who have already engaged.

What the research says about its value

In 2010, a study commissioned by the Network Advertising Initiative (opens in a new tab), an industry body, found that in 2009 behaviorally targeted ads earned 2.68 times as much revenue per ad for twelve ad networks as untargeted "run of network" ads. A smaller subset of networks reported that people who clicked a behaviorally targeted ad converted at 6.8%, against 2.8% for untargeted ads. The price figure measures what networks were paid. The conversion figure covers only people who clicked, on a subset of networks.

Later studies disagree about the size of the premium:

• Google's 2019 experiment. When Google disabled cookies for a small random share of users on its programmatic ad platform for 96 days, the top 500 publishers' revenue from that traffic fell 52% on average (opens in a new tab), a median decline of 64% per publisher.
• A university study of one media company. Researchers at Minnesota, UC Irvine and Carnegie Mellon found that when a user's cookie was available, the publisher's revenue increased by only about 4% (opens in a new tab), or about $0.00008 per ad (May 2019 draft).
• People who opt out. A 2020 study in Marketing Science (opens in a new tab) found that 0.23% of the American impressions in one ad exchange's data came from users who had opted out of behavioral advertising through the industry's AdChoices program. Their ads fetched 52% less revenue than comparable ads.

Those figures describe what publishers and networks receive. For an advertiser, the more useful evidence is whether the targeted ad changes what customers do. A 2013 field experiment with an online travel firm (opens in a new tab) by Anja Lambrecht and Catherine Tucker found that retargeted ads showing the exact products a person had viewed were, on average, less effective than generic brand ads shown to the same visitors. They stopped underperforming when the person's browsing, such as visiting review sites, suggested their preferences had become more specific.

Contextual advertising has its own record. Ster, the ad sales house of the Dutch public broadcaster NPO, described its move away from cookies in December 2020 (opens in a new tab). Only one in ten visitors to NPO's sites consented to advertising cookies. In a 2019 study with Omnicom, Ster ran campaigns with half the budget on inventory with cookies and half without, and concluded that cookie data was not needed for good campaign performance. Its contextual targeting then beat run-of-network and demographic targeting on site traffic and conversions in its own analysis, and Ster stopped using advertising cookies on January 1, 2020. Ster also credits fewer intermediaries taking a share of the budget, so its results do not isolate the effect of targeting, and it is one publisher's reported experience.

Taken together, the studies that find a large premium measure what ad sellers are paid. The travel experiment shows why timing matters: product-specific retargeted ads did worse than generic ones on average, and stopped underperforming when the buyer's browsing suggested their preferences had narrowed. Behavioral targeting is worth testing before you assume it pays.

Where behavioral data is shrinking

Much of the tracking that cross-site targeting relied on is now blocked, restricted or opt-in:

• Browsers. Safari has blocked third-party cookies by default since March 2020 (opens in a new tab). Firefox gives every site its own separate cookie store by default, which stops cookies following people from site to site (opens in a new tab) (rolled out from June 2022). Chrome took a different route: in April 2025 Google kept its approach of leaving third-party cookies to each user's settings (opens in a new tab), and in October 2025 it announced plans to retire Topics, Protected Audience and other Privacy Sandbox technologies (opens in a new tab), citing low adoption.
• Apps. Apple requires apps to ask permission through App Tracking Transparency (opens in a new tab) before tracking a user across other companies' apps and websites. Without it, the device's advertising identifier is returned as all zeros.
• Sensitive categories. Google Ads does not let advertisers in sensitive interest categories (opens in a new tab), such as health, use their own audiences: customer lists, website visitor segments or lookalikes. Google's predefined audiences, which exclude sensitive signals, remain allowed. Meta removed detailed targeting options (opens in a new tab) relating to health, race or ethnicity, political affiliation, religion and sexual orientation from January 19, 2022.

Platforms can also automate audience selection. Meta's Advantage+ audience (opens in a new tab) finds the audience from signals such as past conversions, Meta Pixel data and interactions with earlier ads, and Meta recommends testing it for almost all campaign types except retargeting. That leaves the advertiser's own data as the part it controls: the conversions, site activity and customer lists it collects with permission and shares with the platforms.

The legal lines that matter

The rules below do not ban behavioral targeting. They govern how the data is collected and shared: on a lawful basis, with notice, and with a way to opt out. This is a summary of three sets of rules from the regulators' own pages, not legal advice.

California

The CCPA gives consumers the right to opt out of the "sale or sharing" of their personal information. The California Attorney General's guidance (opens in a new tab) explains that sharing means sharing for cross-context behavioral advertising: targeting based on a person's activity across many websites. For covered businesses, a browser signal such as Global Privacy Control must be honored as an opt-out, and a business must wait at least 12 months before asking an opted-out consumer to opt back in. Sephora's 2022 settlement required it to pay $1.2 million in penalties (opens in a new tab) over allegations that it sold customer information to third-party tracking companies without telling customers and did not process Global Privacy Control opt-outs.

Compliance is uneven. In a 2025 test by Wesleyan University and Consumer Reports (opens in a new tab), 12 of 40 online retailers appeared to keep serving retargeted ads after receiving the opt-out signal.

The European Union

Two layers of rules apply. Placing or reading advertising cookies on a device is governed by each country's law implementing the ePrivacy Directive. Cookies that require consent cannot be set without it, consent must be a positive action by the user, and legitimate interest cannot be the legal basis for placing or reading them. The GDPR governs what is done with the data afterwards, as a 2023 report by the European Data Protection Board's cookie banner taskforce (opens in a new tab) explains. The GDPR lists six legal bases for processing (opens in a new tab), including consent and legitimate interest; the same report said it could be considered that no overriding legitimate interest exists for activities such as "Select personalised ads". Collecting the data on your own site does not remove these requirements. Fines under the GDPR can reach EUR 20 million or 4% of global annual turnover (opens in a new tab), whichever is higher, for the most serious infringements.

Health data in the US

The FTC has acted against companies that sent health information to advertising platforms. In February 2023 GoodRx agreed to a $1.5 million civil penalty under a proposed order banning it from sharing health data for advertising (opens in a new tab). In March 2023 BetterHelp agreed to pay $7.8 million to consumers (opens in a new tab), and the proposed order banned it from sharing personal information with certain third parties for retargeting. Both companies had promised to keep the data private, and both used it as customer lists. GoodRx uploaded contact details of users who had bought particular medications to Facebook to target them, and BetterHelp used customers' email addresses to have Facebook find similar people.

How people perceive it

People notice the difference too. In online experiments with US and European consumers, Kinshuk Jerath, Klaus Miller and Daniel Sokol (opens in a new tab) found that behavioral targeting produced a high sense of privacy violation, while contextual targeting significantly lowered it.

When to use each

The approaches can serve different parts of one campaign. The question is how much of the budget depends on tracking that may be blocked, refused or legally risky, and whether the extra precision is paying for itself.

How to decide with a test

1. 1

   Map the data you hold and how it was collected

   List the behavioral signals your campaigns use: pixels, conversion data sent to platforms, customer lists and third-party segments. For each, record the collection notice, the permission or legal basis for the intended use, and how your site handles opt-out signals such as Global Privacy Control.

2. 2

   Check the category rules before you build audiences

   If you sell in a sensitive category, confirm what Google and Meta allow before planning retargeting or customer-list campaigns. On Google, that leaves contextual placements and Google's predefined audiences.

3. 3

   Run behavioral and contextual side by side

   Compare behavioral and contextual prospecting with the same creative, offer and landing page, keeping geography and timing comparable. Test retargeting separately, so that people's earlier interest does not decide the result. Agree the budget, the success measure and the review date before launch, allowing for your sales cycle. Judge on confirmed sales or qualified leads and the cost of each, not only on the platform's attributed results; where you can, a controlled lift test (opens in a new tab) measures the conversions the ads actually caused. I'd give a test like this about a quarter before moving serious budget.

4. 4

   Watch retargeting for decay

   Retargeting draws on people who already interacted with your brand, so a small audience can see the same ads many times. Exclude people who have already converted, limit how long each person stays in the audience, cap how often they see an ad, and plan new creative before results fade. The guide to user fatigue in social ads covers the signs to watch.

In a 2020 essay on outbound, I explained Andrew Chen's law of shitty clickthroughs (opens in a new tab) this way: "every marketing tactic is great when it starts, but then everyone starts doing it and people become immune to the tactic." Chen's essay cites an analysis in which ad creative left unchanged lost clicks over time, which is the practical risk in retargeting. inBeat, part of Fieldtrip where I am a partner, describes the problem in its Hopper case study (opens in a new tab), which names fast ad fatigue and higher acquisition costs as the challenge on TikTok. inBeat's answer was a creator content process that keeps new ads coming.

Before scaling a winning behavioral audience, check that its advantage holds against your own sales or lead records, and that it stays profitable as frequency rises. Which of your current audiences would still be there if every tracking cookie disappeared tomorrow?