How Facebook ads use machine learning

Copy link
We’re gradually updating people to Meta Accounts. You may see either Accounts Center or Meta Account settings as the place to manage your account settings.

If you’re in the European Region or the United Kingdom, your ad choices are different. Learn more about your ad choices.

To decide which ads to show you, we use Advertiser audience selections and the results of our Ad auction to determine the best ad to show you at a given point in time.
Our machine learning models are part of our ads delivery system that learns as it receives new information, without being explicitly programmed.
This allows the machine learning models to carry out tasks quickly and efficiently, like delivering ads that might be relevant to you.

What determines which ads you see

The two main factors we use to determine which ads to show you are:
  • Advertiser audience selection
  • Results of our ad auction
During the ad auction our machine learning models calculate a “total value score” from three major factors: the advertiser bid, the estimated action rate and the ad quality score.
The ad with the highest total ad value is the one that is delivered to you.

Advertiser audience selection

When creating an ad campaign, advertisers first choose their desired audience through our business tools.
Advertisers can create audiences based on age, location, interests and categories. For example, some information you provide us, combined with actions you take, might suggest to us you’re interested in something, like cooking or fitness, or that you might be part of a larger group (called a category), like a mobile user.
Advertisers can also use information they have about their audiences, like a list of emails or people who’ve visited their website, to build a custom audience or a lookalike audience.
Note:

Ad auction

The second step we use is an ad auction to determine the best ads to show you at a given point in time. Our system gathers ads for which you were included in the advertiser’s selected audience and moves them to the auction stage.
For ads that enter the auction, our system selects the top ads to show to you based on which ads have the highest total value score.
To make sure the winning ad maximizes value for both people and businesses, the winner of the ad auction (which is the ad that is shown to you) is the ad with the highest total value score.
The total value score is a calculation that takes into account three major factors: the “advertiser bid”, the “estimated action rate” and the “ad quality.”
  • Advertiser bid. This is the bid placed by an advertiser for an ad. Because we use the total value score to determine the winner of the ad auction, an advertiser with the highest bid doesn't always win the auction.
  • Estimated action rate. The estimated action rate is an estimate of how likely it is that you'll take the advertiser’s desired action, like visiting the advertiser’s website or installing their app. To do this, our machine learning models may consider your activity on and off Meta technologies, as well as other factors.
  • Ad quality. The ad quality score is a measure of the quality of an ad. The ad quality score is determined by machine learning models and uses many sources including feedback of people viewing or hiding the ad, as well as assessments of low-quality attributes (like too much text in the ad's image, sensationalized language or engagement bait).
The estimated action rates and ad quality help make sure that you see ads that are more relevant to you, even if that ad did not have the highest advertiser bid.
Note:
  • If an advertiser engages in clickbait and engagement bait, it does not improve the performance of their ad.
  • We may make adjustments that could affect auction outcomes and prices. Learn more about ad auctions adjustments.

How machine learning improves ad delivery

As more people view an ad, share feedback, or click through to make a purchase on an advertiser’s website, our models get better at predicting the estimated action rate and ad quality of an ad.

Ways to review and manage how you see ads

  • Customize your Ad preferences in Accounts Center to influence the ads you see. You can also update your Activity from other businesses settings to choose whether we use information that other businesses or organizations send us about your activity on their apps or websites, or your offline interactions with them such as in-store purchases, to show you ads.
    Note: you must be logged in to change your ad preferences and Information and permissions settings.
  • Hide an ad that isn't interesting or useful to you or review why you're seeing a particular ad.
  • Update your profile settings and walk through Privacy Checkup to make sure you're sharing your information with who you want.
  • Visit Access your information in Accounts Center to see your Activity log, Facebook settings and Privacy Center for your Facebook profile.
  • Visit Export your information where you can export your information to a device or external service. Learn more about exporting your information.
  • Learn more about why you’re seeing an ad with the “Why am I seeing this ad?” tool that uses advertiser selections and machine learning models to show you more information about the ads you see.“Why am I seeing this ad?” allows you to tap on ads in your Facebook Feed, get context on why they’re appearing and take action to further personalize what you see. Learn more about why you may be seeing an ad.

Common misunderstandings about ads on Facebook

  • Meta does not sell your data to advertisers or anyone else.
  • We don’t share information that personally identifies you with advertisers unless you’ve given us permission.
  • Meta does not use the content of people’s text messages or phones’ microphone to inform ads or change what you see in your Facebook Feed.
To learn more about the information we receive and how we use it, visit our Privacy Policy and Cookies Policy.

Yes
No

Other ways to get help

Chat with Meta AI support assistant
Resolve issues, make changes and get support in real time