How to Find Where Your Face Appears Online

TL;DR

Several consumer services let you upload a photo and return every public web page that contains a face match. PimEyes and FaceCheck.ID are face-specific; Google Lens and Yandex Images are general reverse-image tools that include face matching in some configurations; TinEye is whole-image matching and does not do faces. A 30-minute sweep across these services maps most of your public exposure, and PimEyes offers a free opt-out that blocks future matches against your submitted photo. The catch is that none of them can see social-media platforms, none of them can see private or login-walled pages, and a face match is not a positive identification. Anything that looks like you could be you.

What "Reverse Face Search" Actually Does

Reverse face search is a two-step pipeline. You upload a photo, the service detects the face in it, then it compares that face against an index of faces it has previously crawled from the open web and returns a list of pages where similar faces appear.

The difference between a face search engine and a regular reverse-image search is what gets compared. TinEye, the longest-running commercial reverse-image service, works on whole-image fingerprints: upload a sunset photo and it finds other copies of that exact photo, even after cropping, resizing, or color shifts.^[1] TinEye does not try to identify the face inside the image. A face search engine extracts a face embedding (a numerical description of facial geometry) and matches that embedding against others, so it can find the same face across photos where the lighting, angle, or background are different.

The catch is that all of these services see only what their crawlers have already indexed. PimEyes' own marketing states it searches "the open web, excluding social media and video platforms."^[2] A photo of you that lives behind a login, in a private Facebook album, or in a friend's Instagram story is invisible to them, no matter how easy it is for a person with access to find. Treat any "reverse face search audit" as a sweep of your public-web exposure, not a sweep of your actual exposure.

The Tools That Actually Run on Faces

Five services matter for a consumer workflow. They behave differently and they cover different slices of the web.

PimEyes

PimEyes is the highest-profile face-specific consumer service. Its homepage advertises "10+ billion images" and a "99.5% accuracy" figure, with three subscription tiers (Open Plus, PROtect, and Advanced) and a free tier that lets you upload a photo and see the matching pages before you subscribe.^[2] Results draw on crawled blogs, news sites, wedding galleries, and pornographic sites. Like most face search engines, PimEyes' own page says it does not search social-media or video platforms.

The current owner, Giorgi Gobronidze, told netzpolitik.org that the company's database stores "around 2 billion unique faces" as numerical hashes rather than photos, and described opt-out as processing "within a single working day."^[3] Biometric Update reported that PROtect plans (which include takedown assistance for sites hosting the images) ran between $90 and $300 a month in 2022, and quoted Token CEO John Gunn calling PimEyes' approach "utterly disingenuous" because the company could verify identity through providers like Mitek or Jumio but chose not to.^[4]

FaceCheck.ID

FaceCheck.ID is a face-specific engine marketed as "Find People Online by Photo." Its front page lists match-confidence buckets (90 to 100 "Certain Match," 83 to 89 "Confident Match," 70 to 82 "Uncertain Match," 50 to 69 "Weak Match") and a removal-request form.^[5] The site explicitly indexes "public, readily available web pages only" and the same page warns that "many unrelated people look alike" and "never rely solely on a face search alone."

FaceCheck.ID also publishes a hard use-restriction: "You may not use this website to make decisions about consumer credit, employment, insurance, or tenant screening."^[5] The site credits system (no per-search price visible on the front page) is the paywall.

Google Lens

Google Lens is the consumer face-search tool most readers already have, but it is a general visual search product first and a face matcher second. Google Lens' own landing page describes it as a tool to "Search, Shop, Translate, and Identify what you see" using a camera or image, and lists plants, animals, products, furniture, clothing, and text as identifiable categories.^[6] Google Lens does include some face-matching capability in practice, but Google's own published documentation does not advertise person identification as a headline feature, and the product has periodically changed which surfaces return face matches.

What this means in practice: Google Lens is worth running as part of a sweep because it returns visually similar images and visually similar pages from Google's own index, and Google's index is by far the largest. Treat its "people who look like this" surface as a hint rather than a finding.

Yandex Images

Yandex Images is the Russian search engine's reverse-image tool. Yandex does not publish a face-recognition feature on its main images help pages, but in practice it has been independently observed to return stronger face matches across different photos of the same person than Google does, and several reverse-image comparison reviews have ranked it that way. Run Yandex Images for any face search you care about, because it covers a different surface of the web (especially Russian-language pages) than Google, Bing, or PimEyes.

TinEye

TinEye is on this list because it is the tool people reach for first. It is not a face matcher. It is a perceptual image-hash service that finds other copies of the same image, and it is excellent at that job.^[1] Use TinEye when you have a specific photo you want to track (a stolen author headshot, a logo lifted without permission, a screenshot that is being passed around) and skip it for "find every photo of my face," because it cannot match the same face across different photos.

The Workflow: A 30-Minute Reverse Face Search

Do this once a year, or after any event that adds new public photos of you (a wedding, a job change, a news appearance).

1. Pick a clean, well-lit, forward-facing photo of yourself. A driver's-license-style image, a cropped headshot, or a recent photo where your face is unobstructed and centered. The better the photo, the better the matches. Avoid group photos where your face is small, sunglasses, hats, or extreme angles. Crop to just your face before uploading.

2. Run TinEye first, against the exact image. This catches unmodified reposts of that exact photo on blogs, in scam profiles, on image-hosting sites, in scraped article mirrors. TinEye shows you where the same image is, not where other photos of your face are, so its results are easy to triage: if a domain is one you recognize, fine; if not, look at it.

3. Run PimEyes, free tier, no signup. Upload the same photo. PimEyes' free tier shows you the matching pages without requiring a paid plan. Read the matching URLs and decide which ones are public, which are you, and which are false positives (a look-alike). Bookmark or screenshot the results.

4. Run FaceCheck.ID with the same photo. FaceCheck.ID returns confidence buckets, which is useful for triage. Anything in the 90-100 "Certain Match" bucket on a domain you do not control deserves attention; the 50-69 "Weak Match" bucket is mostly noise.

5. Run Google Lens on the same photo. Use lens.google.com or the Google app. Google Lens does not market itself as a face matcher, but it does surface visually similar pages from Google's web index, which is the largest single source of public pages. Treat the results as a hint, not a finding.

6. Run Yandex Images with the same photo. yandex.com/images, upload, look at the similar images and similar pages tabs. Yandex covers a different slice of the web than the other four services and frequently surfaces matches the others miss, especially on Russian-language pages.

7. Triage the results. For every URL that surfaces in the 90-100 confidence band, decide whether the use is one you control (your own website, your own social profile), one you tolerate (a news article), or one you do not (a scam profile, an abuse gallery, a doxxing thread). Triage is the part the search engines cannot do for you.

How to Read the Results Without Overreacting

Face search engines are probabilistic. A match is "this face looks like the face in your photo," not "this is you." FaceCheck.ID's own page says "many unrelated people look alike" and instructs users to "always cross reference multiple sources."^[5] PimEyes' homepage advertises a "99.5% accuracy" figure, but the figure is paired with "10+ billion images" and an upload pipeline that discards images after results are returned, not with a published methodology showing what accuracy means in adversarial conditions (look-alike relatives, heavy makeup, a five-year age gap, a weight change).

Concrete rules of thumb:

90+ confidence buckets on a domain you do not control: treat as real until you prove otherwise.
Weak-match bucket on a domain you do not control: open the page, eyeball the photo, and only act if the photo is actually of you.
Match on a domain you do control: nothing to do.
Match on a face that looks like you but is younger, older, or differently built: probably not you. The EFF's Street Level Surveillance guide notes that face recognition systems are "particularly bad at identifying Black, Brown, Asian, and non-gender conforming individuals," which means the systems are also bad at distinguishing a person from a same-ethnicity look-alike.^[7]

If you want a real number on how often a face search engine mistakes someone for you, the published evidence does not give one. Treat the figure on the marketing page as aspirational.

Removing Yourself From PimEyes

PimEyes is the only one of the five services above that publishes a free opt-out workflow. The company's own opt-out page walks you through it in three steps: upload a clear photo of your face, upload an anonymized ID or passport scan (the photo and name visible, the rest redacted), and provide an email address.^[8] The page states PimEyes stores only "face fingerprints of photos found on publicly available websites" and that "we do not have any information about you in our system (name, email, etc.)."

What opt-out actually does, in PimEyes' own words: "the uploaded photo will not be included in future search results."^[8] What it does not do: take down the underlying photo from the website where it lives. PimEyes is explicit that "we are not liable for the origin of the photo, nor able to take it down from the website." For a photo you cannot reach directly (a revenge-porn upload, a stolen identity profile), you need to go to the host site, not to PimEyes.

Two caveats worth taking seriously. The PimEyes CEO told netzpolitik.org that opt-out is processed "within a single working day," but an independent test by netzpolitik.org in 2020 found that a confirmation email arrived and "nothing was deleted."^[3] The Biometric Update piece quotes the New York Times finding that a user who received an opt-out confirmation still surfaced in roughly 100 search results weeks later, including from the exploitation scene the user was trying to scrub.^[4] Plan to re-check PimEyes against your submitted photo a week after you opt out, and again a month later, and resubmit if your matches are back.

For FaceCheck.ID, the homepage offers a "DMCA Takedown Request" link under "Remove my Photos from FaceCheck Search Engine."^[5] For Google, Bing, and Yandex, the levers are the site's standard search-removal tools (Google's "Results about you" tool, the EU's right-to-be-forgotten process where applicable) rather than face-specific. TinEye has no face data to remove, because it does not store face embeddings.

What This Guide Does Not Do

It does not find photos of you on Facebook, Instagram, TikTok, Snapchat, LinkedIn, or X. Those platforms block the crawlers that feed face search engines, and the engines themselves explicitly exclude social media from their index.^[2] To find your exposure on those platforms you need a separate workflow: download your data archive from each one and search the archive by date, by tagged location, and by face.

It does not find photos of you behind a login. Private accounts, family group chats, Discord servers, paywalled archives, anything inside a corporate intranet: all invisible.

It does not tell you whether someone has run a face search against your photo. None of the five services above publishes a "who has searched for this face" log to the subject of the search, and the engines treat search history as their own business.

It does not prevent a stalker, an abuser, or an investigator from running a face search. The same engines that let you audit yourself let anyone else with a photo do the same. The CEO of PimEyes framed this directly to netzpolitik.org: "The user is the stalker, not the search engine."^[3] Operational privacy means treating your public photos as if anyone who has a copy of them can run them through these engines, not as if you can opt out of the engines being there.

And it does not give you a clean "I am not in any face database" answer. The EFF's Street Level Surveillance guide states that "more than half of all adults in the United States have their likeness in at least one face recognition database."^[7] A clean PimEyes result means your public-web photos are not surfaced there today, not that you are not in PimEyes' index, and not that you are not in someone else's index. The most useful thing this workflow gives you is a list of specific pages to deal with, not a clean bill of health.

The Bottom Line

Run TinEye, PimEyes, FaceCheck.ID, Google Lens, and Yandex Images against a clean headshot once a year. Triage the results. Opt out of PimEyes with a clean photo and an anonymized ID scan, then re-check a week later. Treat anything you cannot reach through opt-out as a host-site problem. And remember that the absence of a face-search result is not the presence of privacy. It is just one slice of your public exposure on a single day.