Where your answers go on a purity test

When you tick 100 boxes about your sex life on a website, three things can happen to those ticks: nothing, they get counted, or they get sent to companies you have never heard of. The difference is invisible from the page — every purity test looks like the same list of checkboxes — but it is visible in the page's source, and you can check any of them yourself in about a minute.

Here is what to look for, what we found when we looked, and what this site does.

A checkbox quiz has four places your answers can leak

1. The form submission. If ticking boxes ends with a request to a server that carries the boxes, your answers left your browser. This is not automatically bad — it is how any site that shows per-question percentages must work — but it is a fact you should know rather than assume.

2. Analytics events. Some sites fire an event per interaction, not per page. A quiz instrumented that way sends a stream like question_17_checked to an analytics vendor, which reconstructs your answer sheet without the site ever posting a form.

3. Session recording. Tools in this category record mouse movement, scrolling and clicks, then replay them as video for the site owner. On a page where the clicks are the answers, a session recording is your answer sheet, timestamped, with your hesitation included.

4. Advertising cookies. An ad network on the page sets identifiers that follow the browser elsewhere. The network does not need your answers to profit from knowing which page you were on — and the page you were on is a purity test.

What we found on other purity test sites

Checked on 3 September 2026 by fetching each homepage and reading the scripts it loads. This is a snapshot: sites change, and a script present today may be gone next week.

Clarity, the session-recording script in that first row, is the one worth pausing on. Its ordinary use is watching where visitors get confused on a checkout page. On a hundred-item sexual-history checklist, the same tool captures which items a person ticked, in what order, and where they stopped to think.

One correction to our own row, found by measuring instead of assuming. This site loads no analytics of its own beyond the first-party counter — but Cloudflare, which serves the site, injects its own 30 KB measurement script into pages requested by a browser. It was not switched on deliberately, it does not receive answers or scores, and it is being turned off; until it is, the honest version of our row is the one above. The lesson is the one this whole article is about: check the page a browser gets, not the page you think you published.

None of this is exotic or hidden — it is the default way a content site gets built and paid for. It is simply not compatible with the sentence "your answers stay on your device", which several sites in this niche print above the test.

How to check any quiz site in a minute

You do not have to trust anyone's claim, including ours.

1. Open the page source. Right-click, View Page Source, then search for googlesyndication, googletagmanager, mediavine, hotjar, clarity.ms, facebook.net. Each hit is a third party receiving something.
2. Watch the network while you take it. Open developer tools, go to the Network tab, tick a few boxes and submit. Requests that leave for another domain are the answer to "where did my answers go" — and clicking one shows you its payload.
3. Read what it stores against what it promises. A site that shows per-question percentages must be receiving answers from somebody. If it also claims answers never leave your browser, one of the two statements is false.

What this site does, in the same detail

Taking the test sends nothing while you tick. The list is rendered once, the score is computed in your browser, and no request happens per answer.

Submitting sends one small request. By default it contains the test key and the score, and the row we write holds three things: which test, what score, when. No IP address, no user agent, no session identifier, no cookie.

Two things are optional and off unless you turn them on:

• Sharing which statements you checked. Ticking that box sends the list of item numbers, which go straight into a counter per question — question 12 gained one tick, and so did the submission count for that question. No row keeps a set of answers together, so there is nothing to join back to a person. Those counters are what the per-question pages publish, and each one waits until 30 people have opted in before showing a share, because a percentage over twelve volunteers describes one person.
• An age band. Six coarse bands, never a birthdate, stored beside the score so the by-age averages can exist. Leaving it blank changes nothing about your result.

Rate limiting is the one place an IP address is touched at all: the count of recent requests per address lives in memory for a minute and is never written down. The privacy policy states all of this in the same words, and the about page explains why the answer-sharing is opt-in rather than default — a smaller number of volunteers who agreed beats a complete set collected without asking.

What honesty costs, and why it is worth naming

An opt-in dataset grows slowly. Our statistics page has published nothing for most of its existence because it refuses to publish a figure that cannot carry weight, and the per-question shares appear one question at a time as each clears 30 volunteers.

A site with analytics on every checkbox would have that data on day one, and never have to ask. That is the trade, stated plainly so you can weigh it: slower numbers, and a claim about your answers that survives someone opening the network tab.