How to Test Your Face Against Facial Recognition

A practical, source-by-source walkthrough of the four face matching systems a person can actually run on themselves, and what each result does and does not prove.

State of Surveillance

TL;DR: Four face matching systems can be run on a person today without a court order or a friend in law enforcement. PimEyes and FaceCheck.ID search public web pages for faces that look like yours. Google Photos groups faces it finds in your own photo library, and Apple Photos groups faces on your device. Meta shut its face recognition setting down and says the face templates will soon be deleted. Each system is built differently and each requires a different removal workflow if you find something you do not want surfaced. Here is what each one actually does, what to expect when you run it on yourself, and how to pull yourself out.

The Four Buckets of Face Matching You Can Run on Yourself

When someone searches "facial recognition test" they end up in one of four very different products. Two are paid reverse face search engines that crawl public web pages: PimEyes, an online face search engine that describes itself as available "for everyone," and FaceCheck.ID, a face recognition search engine that claims to "answer a simple question: Who is that?"[1][2] Two are face grouping features in apps you already use: Google Photos face groups, which Google labels a biometric feature in some jurisdictions, and Apple Photos, which uses on-device face analysis to power its People and Pets album.[3][4] Meta, a fifth product, had a face recognition setting on Facebook that it shut down, and Meta's help page says the face template will soon be deleted for users who had the feature on.[5]

These four products do not test the same thing. A PimEyes result tells you whether your face appears on public web pages the service has indexed. A Google Photos face group result tells you whether Google has clustered enough photos of your face to assign a name to it. An Apple Photos result is local to your device. A Meta result is, as of the deprecation notice, a template on its way out rather than a live search.

The honest read before you start: running these tests against yourself is not a privacy vulnerability check. The privacy question is what these services already have, and the test is what you do with the answer.

PimEyes: What It Does, What It Costs, and How to Pull Out

PimEyes' homepage says it lets you "discover where your images appear online, track reuse, and safeguard your privacy," and describes the product as one that "leverages advanced face recognition technology to help you monitor your images, protect your privacy, and audit copyright usage across the entire internet."[1] The workflow is upload a clear photo, get results in seconds.[1] PimEyes says uploaded images are "never stored, never shared, and never used to train models," and that searches are "encrypted end-to-end and discarded after results are returned."[1]

The free entry point on the homepage runs the face search and returns a report of where your images appear online. Three paid tiers unlock more depth. The published tier text says Open Plus unlocks "up to 75 searches daily" and "up to 10 PimEyes Alerts." PROtect, the tier PimEyes labels "BEST DEAL," unlocks "up to 125 searches daily" and "up to 20 PimEyes Alerts," and "Optional: Takedown Agent Support" is an add-on. Advanced unlocks "Unlimited daily searches," "500 active PimEyes Alerts," and "Deep Search access."[1] An Alert is a saved query PimEyes re-runs and notifies on, which is what turns a one-off test into ongoing surveillance of your own face.

What PimEyes says it retains is in the privacy policy, not the homepage. The policy says PimEyes will "retain User's search results for 30 calendar days after User performs the search," and that facial images of users without an account are "securely stored for 48 hours after the User... sends the photo in order to process the search query."[6] The same policy says PimEyes "will never sell your photos to third parties."[6] The opt-out page lets a person "exercise your right to have any and all of your data removed from our system," requires confirmation of age 18 or older, an anonymized ID or passport scan, an email address, and "more than one photo to include different face angles."[7] After opt-out, PimEyes says the personal data is "permanently removed from our system."[6]

What that means in plain terms. The free search is enough to learn whether PimEyes has indexed your face on a public web page. The paid Alerts are a different product: ongoing automated monitoring of new pages that match your face. If your goal is to find out what is already out there, a free search plus a single opt-out is the honest run. If your goal is to be notified every time a new photo surfaces, you are paying for a service that is, by design, watching the open web on your behalf.

FaceCheck.ID: What It Searches, How the Removal Process Works

FaceCheck.ID's homepage describes itself as a "face recognition search engine," and the workflow is the same upload pattern as PimEyes.[2] The differences are in the index and the result types. The homepage says FaceCheck.ID lets you "Upload a face of a person of interest and discover their social media profiles, appearances in blogs, video, and news websites," and says it will "check a person's photo against millions of faces from mugshot, sex offender websites, and suspects that appeared in the news."[2] The search category tags on the homepage list "Social Media," "Scammers," "News & Blogs," "Sex Offenders," "Mugshots," and "Videos."[2]

The pricing model is credits, not subscription. The homepage says credits let you "See More Images" and "View Up to 3x More per Search."[2] Credit prices sit on a separate Buy Credits page, and this article does not quote them.

Removal is the second half of the product. FaceCheck.ID's removal page says image removal is "always free," and that the requester begins the process by "Using your own photo, search and find your face in the search results."[8] There are two identity confirmation paths. The faster path is a selfie. The page says "Most selfie requests are approved instantly, and your photos are deleted right away."[8] The slower path is "an anonymized ID document instead," which "requires manual review" and yields "a tracking link."[8] Manual review, according to the same page, takes "1 to 5 business days."[8] Photos are "IMMEDIATELY HIDDEN" from new searches during review.[8]

FaceCheck.ID also reserves the right not to remove photos of "child sex offenders, convicted rapists, and other violent criminals who could pose physical harm to innocent people."[8] If your goal is to surface mugshots or news appearances of yourself that you want suppressed, that line is the ceiling of the free removal offer.

Google Photos Face Groups: The Biometric Toggle Inside the App

Google Photos runs face recognition through a feature called Face Groups, and the help page notes that the feature "isn't available in all geographic regions, all domains, or all account types." Google's help page describes it as a way "to easily search and manage your photos, you can apply a label to people or pets that appear in photos grouped by Google Photos."[3] The same help page says Google's face model "may be considered biometric data in some jurisdictions," which is the explicit acknowledgment that what looks like a photo organization feature is a biometric pipeline.[3] Retention is "until you either delete them or your Google Photos account is inactive for more than two years."[3]

Turning it off removes the data. The help page says: "If you turn off 'Face Groups,' you delete: Your face groups; Face models used to create those face groups; Face labels you created."[3] That is the honest set: not just the labels, but the face model itself.

The honest read on what this test tells you. Running Google Photos face groups against yourself does not crawl the open internet. It tells you whether Google's algorithm has clustered enough photos of your face in your library to recognize you, and it tells you how Google's clustering handles edge cases. A reader who wants to see what the public web knows should be running PimEyes or FaceCheck.ID, not Google Photos.

Apple Photos: On-Device Face Analysis, and Why There Is No Master Off Switch

Apple's approach to face matching in Photos is built differently from Google's. Apple's legal Photos privacy page says the feature "uses on-device analysis to recognize the faces of people in photos, and groups them together in the People & Pets Album."[4] That is the substantive privacy claim: the face model is built on the device, not on Apple servers. Apple's support page says "the Photos app scans your photos to help you quickly recognize the people, scenes, and objects within the photos," and that with iCloud Photos turned on, the people you favorite or name "stay up to date across your Apple devices," which is the synchronization layer that moves those face groups between your iPhone, iPad, and Mac.[9]

The Apple setup does not expose a single master toggle that turns face recognition off in the same way Google does. Apple's privacy page describes turning off Enhanced Visual Search, Apple's on-device scene-and-object classifier that runs over the photo library to match landmarks, text, and objects, at "Settings > Apps > Photos," but it does not describe a separate toggle that disables the face grouping feature in People and Pets.[4] The support page describes how to suppress an individual person: tap a face in the album and choose "Don't Feature [Name]," with options to "Feature This Person Less" or "Never Feature This Person."[9]

What that means in practice. Apple Photos is the one face matching system in this article where your photo library is the only thing being matched against. The face model lives on your device. There is no public web search implied. The privacy work you can do is per-person suppression inside the album, not a global kill switch.

Meta: Face Recognition Is Shut Down and Templates Are Being Deleted

Meta deprecated its Facebook face recognition feature. The Facebook help page on face recognition currently states that "Die Gesichtserkennung ist nicht mehr verfügbar und die damit verbundenen Funktionen wurden deaktiviert," which translates to "Face recognition is no longer available and the associated functions have been deactivated."[5] For users who had the feature on, the same page says Meta can no longer recognize them, and that "deine Vorlage wird demnächst gelöscht," which translates to "your template will be deleted shortly."[5]

This is not a face search you can run on yourself. It is a record that the largest public face template collection in the consumer internet no longer accepts new templates. The privacy question for Facebook becomes tag suggestions, manual tagging, and any new identity verification flows Meta ships. As of the current help page, those flows are selfie-based age or identity verification, not face matching against other users.[10]

How to Read What You Find

Different tests answer different questions. PimEyes and FaceCheck.ID answer the same question, whether your face is indexed on public web pages, with different indexes. Google Photos face groups answer whether your own photo library recognizes your face consistently. Apple Photos answers whether the on-device grouping feature recognizes your face across your own photos. Meta face recognition, as of the deprecation notice, is no longer a system that runs.

What a "match" means depends on the product. A PimEyes match is a public web page where the algorithm thinks your face appears. A Google Photos match is a cluster of your photos that share a face model. An Apple Photos match is a local album. A Meta match is, currently, a historical question.

What a "no match" means also depends on the product. PimEyes "no match" means PimEyes has not indexed a public web page with your face, which is not the same as no public web page existing with your face. FaceCheck.ID "no match" means the same thing for FaceCheck.ID's index. A negative result from a single product is not a privacy finding.

The honest read: the four products overlap less than they look. A reader who wants to know what the public web knows should run PimEyes first because it indexes the open web at large, then FaceCheck.ID if mugshots or news appearances are a specific concern. A reader who wants to know whether their phone is silently building a face model of them should check Google Photos and Apple Photos. A reader worried about Facebook should know that Meta shut down the feature and is deleting the template.

If You Find Something: How to Pull Out

The four products have four different opt-out paths. PimEyes' opt-out form requires an adult submission with an anonymized ID or passport scan, an email address, and at least two photos showing different angles of the face.[7] PimEyes says the personal data is "permanently removed from our system" after the request.[6] FaceCheck.ID's removal page offers a selfie path for instant deletion and a document path for "1 to 5 business days" of manual review, with photos hidden from new searches during review.[8] Google Photos face groups are deleted by toggling Face Groups off; Google's help page says the toggle removes "Your face groups; Face models used to create those face groups; Face labels you created."[3] Apple Photos does not expose a single master toggle; per-person suppression is via the "Don't Feature [Name]" control in the People and Pets album.[9]

None of those removals delete the original photo from where it lives on the open web. PimEyes says it removes data "from our system," not from the source pages that PimEyes indexed.[6][7] FaceCheck.ID says "All images are indexed from public, readily available web pages only," meaning the upstream page still exists.[8] Removing yourself from a face search engine hides you from that engine; it does not unpublish the photo.

The Honest Takeaway

A self-test is not a vulnerability check. The privacy cost of face matching is paid by the people who have their face enrolled, and running the search does not change that. What running the test does is move the question from "I wonder if my face is indexed" to "yes, here are the seventeen." Knowing which pages have your face is the precondition for asking the page owner to take it down, asking PimEyes or FaceCheck.ID to suppress the indexed copies, and tightening the settings in the apps you use.

The two real questions are upstream. First, are the photos you care about actually public? If your face appears on a work roster, a court filing, a school directory, or a conference bio, it is on the public web by design and a face search engine has already indexed it. Second, does the platform you trust run face recognition without asking you? Google's face groups feature is not available in all regions or account types, and Google itself flags it as biometric data in some jurisdictions.[3] Apple's on-device analysis runs inside Photos, with no master toggle to switch it off.[4] Meta's face recognition feature is off, with the template slated for deletion.[5]

The product to test yourself against is the one whose answer you cannot predict. If you know whether your face is on the public web, you do not need PimEyes. If you have never opened Google Photos, you do not need Google Photos. The honest runs are the ones that can return a result you did not already expect.

Sources

  1. PimEyes, homepage (product description, tier text, search workflow)
  2. FaceCheck.ID, homepage (product description, search categories, removal link, credits text)
  3. Google, Photos Help: Manage Face Groups (what the feature does, biometric status, retention, what deletion removes)
  4. Apple, Legal Photos & Privacy (on-device face analysis, Enhanced Visual Search off toggle, iCloud encryption)
  5. Meta, Facebook Help: face recognition setting deprecation notice (face recognition unavailable, template deletion)
  6. PimEyes, Privacy Policy (search-result retention 30 days, non-account facial image retention 48 hours, opt-out deletion)
  7. PimEyes, Opt-Out form (age confirmation, anonymized ID or passport scan, multiple face angles, email)
  8. FaceCheck.ID, Remove My Photos (free removal, selfie path, document path, 1 to 5 business days, exclusion categories)
  9. Apple Support, Find People and Pets in Photos on your iPhone or iPad (People and Pets album, naming, per-person suppression controls)
  10. Meta, Facebook Help: How video selfie age verification works on Facebook (current selfie-based verification flow)

Published: October 16, 2026