How the market actually prices this
Three models dominate, and they suit different users. None of them is inherently better value; they fail for different people.
| Model | Typical shape | Suits |
|---|---|---|
| Capped free tier | 10 to 20 checks total, then pay | Nobody for long. It is a trial |
| Credit subscription | Around $5 to $10 a month for a few hundred checks | Regular professional use |
| Per-image API | Fractions of a cent per image, minimum spend | Automated pipelines at volume |
| Unmetered free | Daily allowance, no payment | Occasional and personal use |
The capped free tier is the most common and the most misunderstood. Twenty checks is not a free tier; it is a sample. It exists to run out at the moment you start finding the tool useful.
What a paid tier usually adds
- Volume. More checks per month, which is the actual product in most credit-based plans.
- Saved history. A searchable record of past checks, which matters enormously in claims, moderation and any audited process.
- Branded reports. Your logo on a PDF that goes to a client. Cosmetic, and it is what agencies ask for first.
- API access. The only way to run checks inside your own systems rather than by hand.
- Team seats. Shared access and shared history, usually the point at which pricing jumps.
- Support with a response time. Worth something when a result is holding up a decision.
Notice what is absent from that list. None of it makes a score more accurate. The model answering the question is almost always the same one the free tier uses, and vendors are generally straightforward about this if you ask directly.
What a paid tier buys
- Higher monthly volume
- Saved searchable history
- Branded client reports
- API and team seats
What it rarely buys
- A more accurate model
- Broader generator coverage
- Freedom from uploading
- A different verdict
The question that separates honest pricing
Ask a vendor whether the paid tier uses a different model, and if so, for the benchmark figures for both. There are three possible answers and each tells you something.
- Same model, more volume. The common and honest case. You are buying throughput and workflow.
- Different model, figures provided. Legitimate, and now you can judge whether the accuracy gain justifies the price.
- Different model, figures unavailable. Treat the accuracy claim as marketing until it is measurable.
The cost nobody prices
Almost every paid detector is an API. That means uploading, and uploading is a cost that does not appear on the pricing page.
If the images are claim photographs, identity documents, unpublished journalism or client work under NDA, sending them to a third-party service creates a data processing relationship. That needs a contract, a retention position and a line in a privacy notice, and in a regulated organisation the procurement conversation costs more than the subscription.
When paying is clearly right
There are situations where a free browser tool is the wrong instrument, and being direct about them is more useful than pretending otherwise.
- You need an API. Anything running inside your own pipeline needs a service, and a browser tool cannot provide one.
- You need a shared audit trail. Where several people review and a regulator may ask later, saved shared history is not a convenience.
- You are checking thousands a day. A daily allowance meant for people does not fit an automated workload.
- You need a contractual guarantee. An SLA, an indemnity or a named support contact only comes with a commercial relationship.
- You need branded client deliverables. Agencies charging for verification usually need their own name on the report.
How to evaluate before paying
-
Build a test set from your own work
Twenty images you know the answer for, including the awkward ones: screenshots, night-mode shots, heavily retouched files. Vendor benchmarks never include these.
-
Run it through the free tier first
The detection is usually identical, so free is a fair test of the thing that matters most before any money moves.
-
Count false positives separately
How often it flags your genuine images is the number that decides whether the tool is usable on people. It is rarely the number on the pricing page.
-
Price the upload question
Ask your privacy team what a third-party image processor costs in review time. That is part of the total, and it is often larger than the licence.
-
Then buy the workflow
Pay for history, seats and reporting once you know the detection works on your material.
The hidden cost of a capped free tier
A twenty-check trial has a subtler effect than running out. It changes how you use the tool while you still have checks left.
People ration. They check the image they already suspect and skip the eleven around it, which removes exactly the comparison that makes a batch informative. The single most useful technique in image verification is the one a capped tier discourages.
It also discourages testing the tool on images you know the answer for, which is the only way to calibrate your own trust in it. Spending five of twenty checks on your own photographs feels wasteful, and it is the most valuable thing you could do with them.