While most consumers know their online behavior is used to target them with advertisements, it’s not always top of mind—until the barrage of personalized ads come in to serve as a reminder, making viewers feel uncomfortable. But there are clear incentives for marketers to continue walking the razor’s edge between delighting and disturbing audiences to meet the demand for meaningful interactions.

The key is the type of information a brand uses. Private information—such as sexual orientation, political beliefs or sensitive search topics—makes consumers feel as if someone is talking about them behind their back, according to an experiment conducted by the Harvard Business Review. More general information, such as name, general location or shopping history, is deemed less offensive.

Harvard Business Review also found that consumers don’t seem to mind brands collecting information, so long as they maintain a level of transparency. When trust levels were high, disclosing acceptable flows actually boosted click-through rates.

In one part of the experiment, shoppers were told that an advertisement was based on their activity on the site and click-through rates increased by 11 percent. Time spent viewing the advertised product rose by 34 percent and revenue from the product grew by 38 percent as a result of the transparency.

But purchase interest declines when consumers realize that their personal information is flowing in ways they dislike. Consumers felt more confident viewing ads based on information they voluntarily provided either through a form or through activity on the website. When a company claimed to have inferred the information from outside sources, purchase intent dropped 17 percent.

A new study by customer experience (CX) analytics firm InMoment found that at least 75 percent of consumers surveyed think most forms of ad personalization are at least “somewhat creepy.” And consumers don’t keep this information to themselves: one in five respondents tells friends about marketing experiences that they consider creepy, and one in 10 shares “Big Brother-type experiences” on social media, the study found.

Meanwhile, 40 percent of brands admitted that they commonly use personalization techniques that make consumers feel uncomfortable. This reflects a significant disconnect between what brands and consumers think is an acceptable use of private information—something upcoming GDPR requirements are attempting to fix.

Contrary to what this type of backlash might imply, studies show that consumers prefer their advertising to be custom-tailored, so long as it’s done right. A 2016 study by Adlucent found that 71 percent of consumers prefer this type of treatment, and 44 percent were willing to provide select personal information to receive it. A similar 2018 study by Adobe found that 42 percent of respondents get annoyed when their digital content experience isn’t personalized.

Major ad providers are each addressing the issue of personalization control in different ways. Google recently started offering a way to report an ad for “knowing too much,” as well as the ability to block individual advertisers and specific ads on Google searches, YouTube, Gmail and independent sites.

Amazon is letting brands tap into its recommendation engine technology by offering its own database service called Neptune. Amazon Neptune has been designed to handle billions of relationships and run queries within milliseconds to better understand consumer behavior. Gartner predicts that by 2020, smart personalization engines used to recognize customer intent will enable digital businesses to increase their profits by up to 15 percent.

Spotify closed out 2016 and 2017 with campaigns that highlighted listener statistics in a light-hearted way that felt less weirdly obtrusive and more entertaining.

When Netflix tried to replicate Spotify’s sense of humor, the response was less friendly.

Perhaps the thought of being watched while they binge TV at home is more unsettling than knowing what kind of playlists you create.