Reverse face search is most useful when the product treats a match as evidence to review, not an automatic identity claim. Sherlock search follows a simple guest workflow: one photo, one hosted payment, server-side confirmation, and a private report containing possible matches with public source links.
This post breaks down the workflow and the decisions that make it safer and easier to understand.
Why the workflow matters
A face-search interface has two separate jobs:
- compare the submitted image with an available public-source index;
- help the user interpret the returned leads without overstating certainty.
The second job is easy to overlook. A visually similar thumbnail can be copied, mislabeled, cropped, or shown outside its original context. A useful report therefore needs source provenance, clear status updates, and language that distinguishes resemblance from identity.
The search state model
A transparent guest journey follows this sequence:
Select photo → validate locally → complete hosted checkout → verify payment on the server → upload securely → compare public-source images → assemble a source-linked report.
Each transition should be explicit. The browser should not upload the search image before checkout is confirmed, and the application should not trust a success URL without verifying the Stripe Checkout Session on the server.
Sherlock Search uses a $2.99 USD one-time base price. It does not require an account or subscription, and eligible promotion codes can be entered during Stripe Checkout.
Input checks before payment
The best time to catch a weak image is before checkout. A practical input gate checks:
- one visible, front-facing face;
- JPG, PNG, or WebP format;
- a file size within the stated limit;
- enough resolution to inspect stable facial details;
- no heavy filter, sticker, or obstruction over the face.
These checks cannot guarantee a match, but they prevent avoidable failures and make the payment boundary easier to understand.
What belongs in the report
A responsible result card should answer four questions:
- What was returned? A possible visual match.
- How should it be prioritized? By a similarity signal, not certainty.
- Where did it come from? A clickable public source page.
- What should the user inspect? Account name, caption, date, domain, facial details, and contradictions.
The original page matters more than the thumbnail. Several result cards may also point back to the same copied image, so a larger result count does not automatically mean stronger corroboration.
Failure states should be designed, not hidden
Long-running comparisons need named stages and a realistic time expectation. If the provider fails or times out, the interface should say what happened and what happens to the payment.
Other important states include:
- the paid user returns in a different browser and must reattach the photo;
- the comparison completes with zero possible matches;
- search media reaches the retention boundary;
- a technical failure triggers an automatic refund of the amount paid.
A zero-match search is still a completed comparison. It does not prove that the face is absent online; relevant images may be private, deleted, unindexed, or visually different.
A review checklist for users
For each possible match, record:
- the public URL and domain;
- the date reviewed and any visible publication date;
- stable facial similarities;
- visible contradictions;
- whether another independent source supports the same conclusion.
This turns a quick visual impression into a repeatable evidence review.
Privacy and responsible use
Search media is retained for up to 24 hours. Users should keep the private recovery link until they have finished reviewing the report.
Reverse face search should be used only for lawful, non-harmful purposes. It should never be the sole basis for employment, housing, credit, insurance, law-enforcement, or other decisions that materially affect a person.
The strongest product promise is not “we know who this is.” It is: “here are public-source leads, the context you need to review them, and clear limits on what the result means.”
