A quick note on scope. This post is about the consumer-facing face-search engines, the ones anyone can open in a browser and pay for. Government and law-enforcement systems are a different market with different buyers, and we have covered them separately in how law enforcement actually uses face search. Here we are interested in the ordinary commercial question: who hands over a card, and for what.
A search engine with a paywall
The commercial engines share a business model that is easy to describe. Anyone can upload a face and see that results exist, often as blurred thumbnails or a count. To see the results clearly, open the source pages, run more than a handful of searches, or set up ongoing alerts, you pay. Pricing is tiered the way software subscriptions usually are: a monthly plan for casual use, larger plans for heavy use, and business tiers with more searches and more seats.
Notice what is being sold. A conventional search engine is free to use and sells advertising around the results. A face-search engine sells the results themselves. That single difference explains most of what follows, because it means the engine’s revenue rises with the number of faces it can return and the number of people willing to pay to see them. Your face, indexed, is inventory.
Customer 1: the curious
The largest group of paying users is almost certainly ordinary people running searches on other ordinary people. A date before meeting, which we wrote about in reverse-face-searched on a first date. A new neighbour. A contractor who quoted a price. A person from a group chat whose story did not add up. A face from a party, a face from a screenshot, a face from the news.
Most of this is not malicious. It is the same impulse that made people search names, extended to faces because faces are what people actually have. But it is worth sitting with the fact that the engines exist commercially because a large number of people will pay a modest monthly fee for the ability to identify strangers. That demand is the foundation everything else stands on.
Customer 2: the professionals
The engines market heavier tiers to people whose job involves finding out about others: private investigators, skip tracers, insurance fraud teams, journalists and researchers, security consultants, background-check operations and, increasingly, recruiters and HR departments. We covered the hiring angle in recruiters are face-searching you before the interview.
Professional use is where the higher-priced plans earn their keep, because a professional runs many searches and needs the source pages, not just the thumbnails. It is also where the engines’ public positioning tends to live: the marketing speaks of verification, due diligence and fraud prevention, and some of that is real. But a professional plan is a plan, not a licence. Nothing about buying one restricts what the buyer searches for.
Customer 3: businesses
Some engines sell to companies rather than individuals: retail loss-prevention teams, venue security, hospitality groups screening guests, platforms verifying that a profile photo belongs to the person behind it. In these cases the business is paying to compare faces it encounters against the engine’s index of the public web, which is a different thing from a business running its own cameras against its own watchlist.
The distinction matters to you because it means your face can be checked by an organisation you never interacted with online, in a place you physically walked into, against photos that have nothing to do with that place. The index built from the public web becomes a reference library for the physical world.
Customer 4: the ones the paywall does not stop
A paywall is a payment step, not a screening step. Anyone with a card who agrees to a terms-of-service checkbox can buy the same plan as the investigator or the recruiter. That group includes stalkers, harassers, people who want to unmask an anonymous account, and people who saw a face once and want to know where it lives. We have covered what that looks like from the target’s side in your face is a search term, and stalkers know it.
Engines generally prohibit this use in their terms, and some have taken steps against it. But the structural point stands: the business is built to make faces findable to paying strangers, and it cannot know which paying stranger is which. The price is a filter for willingness to pay, and nothing else.
Customer 5: you
Here is the part of the business model that surprises people. Several engines also sell a product to the people they index: a subscription that monitors the web for new photos of your face and alerts you, sometimes bundled with tools to request takedowns from the sites hosting them. The pitch is protection. Buy the plan and you will know when your face appears somewhere new.
Read that carefully. The same company that made your face searchable is offering to sell you notifications about the searchability it created. Both sides of the transaction pay the same business. It is not a scam, exactly; the alerts may genuinely be useful. But it is a business that profits whether you are the searcher or the searched, and that has consequences for how much it wants your face gone.
What the money buys
Subscription revenue pays for the three expensive things a face-search engine has to do continuously. Crawling: fetching an enormous number of public web pages over and over, because new photos appear every second and old ones move. Processing: running every face found through the detection and measurement steps we described in what a face-search engine sees when it looks at you, which takes serious computing power at that scale. Storage and search: keeping every resulting measurement, with its source address, in an index that can answer a query in seconds.
None of these costs go down when your face is removed. All of them go up when the index grows. An engine that is paying to crawl the web anyway has every reason to keep everything it finds and very little reason to spend engineering effort making sure specific faces stay out. We looked at the cost side in more depth in what it actually costs to run a billion-face surveillance engine.
The incentive problem
Put the pieces together. Revenue comes from people paying to see faces. The more faces in the index, the more results, the more valuable the subscription. A face that has been removed is a result the engine can no longer sell. So an opt-out is, from the business’s point of view, a small reduction in inventory that also costs money to process.
This is why removal is rarely a one-time event. An opt-out takes your existing measurements out of the results. It does not stop the crawler, and the next public photo of you can produce a fresh measurement that the opt-out never covered. Some engines say they prevent re-indexing; others are explicit that they do not, a difference we went through engine by engine in face search engines compared.
What you can do
Understanding the business gives you a realistic picture of the task. You are not asking a neutral utility to correct a record. You are asking a subscription company to withdraw a piece of inventory, and then asking again when it restocks. That is doable, but it is a standing chore rather than a form you fill in once.
That standing chore is what FacePrivacy is for. We file opt-out and removal requests with the facial-recognition engines we support, on your behalf, and we re-file them as necessary, because the crawlers do not stop and a face dropped in one month can be picked up from a new photo the next. We cannot guarantee any engine’s response and we do not remove photos from the sites that host them. What we do is keep asking, on a schedule, so that the business built on your face has to keep letting it go.
Your face is their inventory. Keep asking for it back.
FacePrivacy files opt-out and removal requests with the major face-search engines on your behalf and re-files them monthly, so a withdrawn face stays withdrawn.
Get started →