The pregnancy of 30-year-old tech entrepreneur Sally was announced to the world through her husband's Facebook page, after he tagged her in a photograph showing a positive home pregnancy test . Two months later, while Sally was browsing Facebook, she noticed a Huggies ad.

Sally, who wouldn't let Ad Age use her real name because she does business with Facebook, had never "liked" Huggies or any baby-related posts or pages. Nor had she posted about her pregnancy, so she figured Facebook had connected the dots between her husband's status update and his relationship with her.

Did Facebook and its client, Huggies, know she was pregnant? It depends on whom you ask.

According to Facebook and Huggies parent, Kimberly-Clark, Sally's browsing experience resulted from blind luck. The ad, featuring a cherubic little one wearing a lei, was the subject of a two-week test targeting parents of young children, Huggies fans and their friends -- as well as a three-day subtest of women ages 18 to 34. This is probably why Sally saw it, according to Meg Way, Kimberly-Clark's global director of digital planning and strategy.

Facebook, for its part, said it rarely uses the content of status updates as a signal for ad targeting.

But plenty of marketers that target pregnant women believe they're identifying them, at least in part, by their status updates. Some marketers say they have been told so by Facebook. The confusion over what exactly Facebook is doing is indicative not only of the opacity of the social network's ad-targeting algorithms but also the privacy tightrope it walks, offering marketers the precision they crave while assuaging users that their every utterance isn't being mined for ad targeting.

Here's what we know, based on what Facebook has told Ad Age : Marketers can reach pregnant women on Facebook with near-surgical precision, mixing and matching a variety of targets, such as those interested in baby products and people who like children's music, which taken together produce a high likelihood of hitting the mark.

But Facebook is careful to note that it doesn't use the content of status updates to target pregnant women.

"If your goal is to reach something we don't have in our system, you can still reach the people you want to reach; you'll just reach additional people as well," said Brian Boland, director of product marketing at Facebook.

Café Mom VP-Marketing Kristina Tipton said her team has identified a Facebook audience of more than a million women who are likely to be pregnant or may have recently been so by anonymously targeting specific keywords that show up in users' conversations like "morning sickness," "ultrasound" and "pregnancy test ." Those terms surface with a hashtag beside them in Facebook's ad tool, and Ms. Tipton has been told by her Facebook rep that this process includes people who have mentioned the terms in their posts as well as users who have added those terms to their profile.

'Like' morning sickness?

But according to Facebook spokeswoman Annie Ta, audiences are created for precise interest targets solely based on what users have added to their profiles as "likes" or interests. In other words, if on Facebook you liked the topic "morning sickness" or listed "pregnancy tests" as an interest, you may have been targeted by Café Mom.

It's hard to imagine too many women liking morning sickness or even pregnancy tests. In a search of Facebook's ad tool, #morning sickness has an audience of 2,000, with an audience of 268,000 for #ultrasound and 54,000 for #pregnancy test -- presumably all people who have added those terms as a like or interest, according to Facebook's explanation. Facebook technically does have a way to reach "expecting parents" in its ad tool: a "broad category target " of 63,920 men and women in the U.S. But this is hardly a compelling number for national advertisers. Facebook says that the signal that pulls users into this group is a "child: expected" profile setting under "relationships and family" that it didn't include in the new "timeline" format, which explains why the audience is small and dwindling.

But when it comes to that group, too, there's confusion over whether status update content is used as a signal. One Facebook advertising partner who works closely with the company told Ad Age of asking a Facebook rep what data points inform "expecting parents" targeting and being told that there isn't a discrete profile point that someone could use to define herself as an expecting parent; it works by targeting user activity including relevant status updates, likes and interests.

Tech-savvy consumers may already assume that their status updates are a key part of the targeting recipe since Facebook's own "data use policy" states that "key words from your stories" are used to deliver ads. But according to Facebook's Ms. Ta, key words in status updates are used only rarely for real-time targeting. (A hypothetical example is a user who has posted "I could go for some pizza tonight" being served an ad with a coupon from Domino's Pizza.)

The big question raised by these differing accounts of how targeting works is whether Facebook needs to be more transparent about how it targets users. Certainly there's a gap between what marketers say they are being told and Facebook tells a journalist on the record.

Still, Facebook wouldn't be the only company with an opaque targeting system on the market. Google for years has closely guarded its Adwords algorithm, offering best practices to marketers but never detailing the exact recipe for why and when you see a particular ad. Google hasn't been penalized by marketers for that opacity, despite its algorithm remaining an enigma even to the search marketers whose livelihood depends on it.

Facebook certainly has a reason to remain ambiguous about targeting, given the sensitivity of the issues involved, especially with health topics. Target learned this lesson when The New York Times Magazine reported that the retailer had been inferring which customers were pregnant based on their recent purchase history and marketing to them accordingly.

Purposefully vague

While it remains to be seen if Facebook users lash out, marketers and agencies, for the time at least, are assuming the best -- that the targeting algorithm is ever changing as it is continually refined for the benefit of marketers and consumers.

"For me, it doesn't seem like they're trying to be deceptive," said Justin Oh, senior digital planner at 22squared, who noted that Facebook has refused to provide details about how targeting categories are constructed because they're always changing. "It seems like they don't want to say anything definitive and be held to that , almost."

Indeed, Facebook's Mr. Boland remained vague when asked whether Facebook would someday use real-time conversations to make the "expecting parents" target more robust for advertisers, who might want more precise targeting. He said he didn't know that Facebook had the capability.

"Not all advertisers are created equally in terms of how they define privacy as opposed to how we define privacy," he said.