Browsing listings: reading is an action the server performs
Reading feels like watching and it is not. Each page you look at was requested from a machine, one request at a time, in an order.
The market server · Whoever runs the platform · Anyone who later obtains the server
People instinctively treat browsing as observation and posting as disclosure. On a website the distinction does not hold. Nothing appears on your screen that was not fetched, and every fetch is a message from you saying which thing you wanted. The market server is on the receiving end of every one of those messages, in sequence, with the gaps between them.
What the requests contain
A request names a page. On a market that usually means a category, a search phrase, a vendor profile or an individual listing. Search terms in particular are conspicuous, because a search phrase is something you composed rather than something you clicked, and it is often phrased the way you would say it out loud.
A run of requests also has an order and a tempo. Twelve seconds on one listing and eleven minutes on another is a difference the timings describe without anyone needing to interpret them. Returning to the same page across separate sessions is likewise a fact about the sequence, not an interpretation of it.
What a browse trail suggests
A trail supports statements about attention. Which categories were opened. Which vendor profiles were returned to. Whether the reading was broad and shallow or narrow and repeated. Whether a session looks like someone comparing options or someone going straight to a page they already knew about.
It also supports comparison between accounts without saying anything about either of them individually. Two accounts that visit an unusual combination of pages in the same order within a short window are doing something similar. That is a statement about similarity, and it is the kind of statement that gets over-read constantly.
What it does not prove
A page view is not an intention. Curiosity, price comparison, reading reviews, following a link from a message, checking whether a listing still exists, and idle scrolling all produce identical records. There is no field distinguishing them, and no amount of trail resolves the ambiguity, because the ambiguity is in the person rather than in the data.
A trail is also not an identity. It describes what an account looked at. Attaching that to a person requires something else: a name reused elsewhere, covered in registering an account, an address typed into a form, covered in typing a delivery address, or a device that kept a copy, covered in what lands on the disk. Those are the exposures that actually carry identity. A list of pages does not.
Time spent on a page is not attention either. A tab left open while you make tea looks exactly like eleven minutes of careful reading. This is worth stating plainly because inference from dwell time is one of the places people most readily assume more precision than exists.
Signed out versus signed in
Browsing while signed out produces requests that are not attached to an account. The server still receives them, and requests arriving over one connection are still recognisable as belonging to one visit, because that is how a web session works at all. What is missing is the label. The trail is anonymous within the visit and does not join up with what the account did last week.
Browsing while signed in attaches every request to the account row, through the session token described in signing in. The trail joins up across sessions, across devices and across months. That is not a hidden mechanism, it is the entire purpose of signing in: the server is meant to know it is you, so it can show you your orders and your messages.
The practical middle ground is that most markets restrict browsing to signed-in users, so the choice may not exist. Where it does exist, the difference is real but narrower than it appears, because the moment you sign in during the same visit, the earlier part of that visit is joined to the account. Reading first and signing in afterwards inside one session does not separate the two.
| What the request supports | Signed out | Signed in |
|---|---|---|
| Which pages were fetched | Yes | Yes |
| The order and timing | Yes | Yes |
| Joined to an account name | No | Yes |
| Joined to what that account did previously | No | Yes |
| Joined to a real identity | No | No |
What changes the answer
4 things change how much this action gives away. None of them takes it to zero, and none of them is a promise.
- Separating reading from doingAnything read before signing in is not attached to the account, in a separate visit. Within one visit, signing in joins the lot. This changes what is joined, not what is received.
- Fewer distinctive search phrasesA phrase you composed is more particular than a category you clicked. Navigating by category produces a less individual record. It does not make the record disappear, and it is worse for finding what you were after.
- Not treating the trail as private thinkingThe most useful adjustment is expectational rather than technical. If you would not want the sequence of pages read back to you, that sequence is the thing to reconsider, because nothing downstream removes it.
- A new circuit for a new visitRequesting a fresh circuit changes the path your traffic takes to the site, which matters for your network and for the relays. It does nothing about the account label, and while you are signed in the label is what joins the trail together. See asking for a new circuit.
What this card is not
This card does not claim that browsing history is stored anywhere. It states that the requests are necessarily received, and that whether they are recorded is a choice you cannot see from the page.
Questions that come up
Does the market know how long I looked at a listing?
It knows when the page was requested and when the next request arrived. The gap between them is not the same as attention, since a tab can sit open for reasons unrelated to reading.