Facebook is experimenting with letting brands study people's posts and comments on the network in an effort to better inform their marketing.

The beta test, an extension of Facebook's Audience Insights API marketing tech platform, isn't expected to be widely available until next year, according to people familiar with the offering who spoke on condition of anonymity to discuss something Facebook hasn't announced yet. Early ad partners, which include top agencies and media companies, are searching Facebook's vast history of public posts to see what topics, themes, brands and products are being discussed. Users' identities are withheld.

It's the first time that Facebook, where room for ads in the main News Feed is almost maxed out, is making it possible for advertisers to mine what users post. The new insights tool could help marketers see the social network in a whole new dimension, and even give them a broader understanding of their businesses, with data that informs them about trends in the industry and the consumer mindset.

"On Facebook, you know everything about a person from their profile, what they liked and who they connect with," says one agency executive in the test. "But Facebook is not good at knowing what people are saying, what they're posting."

Advertisers involved said that Facebook has been taking its time developing the new data products because it's trying to balance privacy of users with what it can offer marketers.

"We're always testing out new potential solutions for advertisers, but have nothing to announce at this time," a Facebook spokesman said in an e-mail statement.

Facebook has always been much less interested in the content of posts than how people respond to them. That's partly why it was so blind to the fake news of election season; bad actors posted hoaxes and misleading headlines, but the company didn't see a problem because the content "performed" so well. Some say the audience insights program now in testing could help Facebook get better control over its platform, even if that's not why it was designed. A tool that can sift comments quickly could root out, for example, hate speech.

On the other hand, more data harvesting on consumers opens privacy concerns and could theoretically lead to unintended ad targeting. Facebook was recently embarrassed by automated ad tools that let marketers target anti-Semitic interests.

Advertisers hope that the tool will eventually lead to ad targeting, perhaps compiling groups of people who discuss the same topics and hitting them with highly tailored messaging. Ads could be served to thousands of people who discuss a soft drink, for example. But the research is the most important feature, according to advertisers. An automaker could use the tool to analyze how people reacted to a new car, in theory, or a brand could use public feedback to refine its media plan.

This type of analysis, research and data, based on people's comments, has been hard to come by on Facebook, partly because of privacy concerns that are still very much present, but also because the network was never set up to sift through streams of commentary. That's different than, say, Twitter, where posts are public and a search returns all the accounts discussing that topic. Facebook posts are usually only visible to friends of the person posting, making it harder to track trends and get inside the general mood of the crowd.

Advertisers say tapping into Facebook posts opens whole new opportunities for marketers, letting them approach the social network in ways that just weren't possible. A direct line into what people say can be more informative than what they claim to like and similar cues.

But deciphering Facebook posts can be difficult, too. Unlike Twittter's necessarily brief messages, Facebook lets people ramble. "Most of what we saw was garbage," said one agency exec, who tested the tools. "Nonsensical."

Still, marketers see plenty of potential. "You just have to leverage it in a way that is intelligent and perhaps less obvious," says Emily Kramer, senior director at Merkle, a marketing data and analytics company.

Sure, posts could indicate if someone is in the market for a product, but less intuitively they could also reveal who is not in the market. Perhaps a new car buyer isn't the best candidate for an expensive cruise offer.

And the "creep factor" of scrutinizing personal posts seems to be less of a concern as time goes on. "The world is getting more intelligent about how data is used," Kramer says. "Fewer people say it's creepy and realize the quality of engagement on a platform is getting better."