Proxy for Qwant: Country View, Session State and Logging Routine

There are two distinct reasons to use Qwant with a proxy, and the two should not be confused. The first is privacy: you want to separate your exit address from your provider's address. The second is measurement: you want to check what a user in another country sees. The setup is the same in both cases; the expectation is different.

On the engine side, the key point is that the results page is not a single block. The web listing, image and news tabs, sponsored units and shopping blocks are fed by different rules. A page that looks rich in one country may consist of nothing but the organic listing in another.

Below we first separate these components, then cover the measurement routine, exit selection and verification steps. For the comparative behaviour of other engines, the search engine guides section is the right place.

What exactly does a proxy change on the Qwant side?

The only thing that changes is which address the request reaches the engine from. The query text, the browser's language list, your saved preferences and the hints derived from your screen dimensions all stay the same. That is why the sentence "I'm hidden behind a proxy" is incomplete; the identity at the network layer changes, while the traces on the client side stay put.

The second limit is on the content side. Over an HTTPS connection, a proxy only opens a tunnel and carries encrypted bytes; it cannot read your query. On the other hand, which domain you connect to can be visible on the proxy server. For someone setting up a proxy for privacy reasons, the real question is this: what does the provider log and how long do they keep it (logging and privacy)?

The third is scope. A proxy only routes the layer you define it for; an exit defined in a browser profile does not cover the other applications on the same machine. If you want to take all of a device's traffic into a single tunnel, the tool you are looking for is probably not a proxy; the difference in scope is decisive here, because a proxy is defined at the application or profile level, not at the device level.

How do vertical modules diverge across EU markets?

Qwant is an engine hosted in Europe whose product decisions centre on EU markets. The most visible expression of this on the results page is the vertical modules that sit alongside the web listing: image and news tabs, map-like shortcuts, summary blocks that open depending on the query. These modules are fed by sources separate from the web index — the news side relies on publisher feeds, the image side on a separate index. The result is this: for the same query, one of the modules may appear on the page in one market and never open in another, and that does not mean the index does not recognise the query.

You see the divergence best when you place neighbouring EU markets side by side. A French-language query may fill the news module with local publishers on an FR exit; the same query on a BE exit produces a bilingual set of publishers, because the market itself is bilingual; on a DE exit the module may close entirely because the query language and the market language do not match. In all three rounds the first five lines of the organic listing may largely overlap. What diverges is not the ranking but which modules the page is built with — its skeleton.

The effect on the logging routine is direct: in a Qwant record, a "rank" column alone is not enough. For each round, also write down which vertical modules opened and where they sat on the page. When a module opens, the organic lines move down; the "rank dropped" observation between two countries is often a difference of skeleton, not of ranking. Keeping the module list as short labels is enough; what matters is that the column exists in every round — a column added later cannot be filled in for past rounds.

In measurement practice, treat the query language and the exit country as two independent variables and vary only one at a time. Asking French-language queries from a France exit and asking the same query from a neighbouring market produce different module breakdowns; if you combine the two in the same chart, what you hold is not the engine's behaviour but the trace of your own setup. If you are going to compare, open a separate series for each market and name the series by country code.

Separating cookies, local preferences and network signals from one another

The state that affects the results page accumulates in three places, and these change at different speeds. Cookies and preferences in local storage sit in the browser; the language list comes from operating system and browser settings; the exit address is on the network side. Changing all three at once in a measurement makes the difference impossible to interpret.

The most common mistake is treating a private window as a clean environment. A private window keeps no local record, but it does not change your language list or exit address; moreover, in some configurations your extensions keep running. A clean environment means a separate browser profile with its own cookie store and its own proxy definition.

The second common mistake is reading a cookie cleanup as meaning the results have been reset. The network signal that determines the country view stays in place; if the region preference in the interface has been saved, that comes back too. To separate things correctly, vary one variable at a time and note the result.

If you work as a team, this separation turns into a process. Keep in a single table which profile is bound to which exit, which language list it carries and who uses it. This is the information most often lost in a handover.

What should stay fixed in a series that spans weeks?

Long-running tracking consists of rounds stacked on top of one another. Rounds can only be compared if the conditions are the same. There are four layers to fix, and each carries a different responsibility: the query set, the client profile, the exit and the logging format.

The query set is the layer that breaks most easily, because over time the pressure to "add just one more" appears. Add it, but version the set and do not combine the old version and the new version in the same chart. The client profile is the second layer: the browser version, language list, screen dimensions and active filters must not change during the run.

On the exit side, the rule is one address per market. ISP exit or if you are using a static address, this is satisfied automatically; if you work from a pool, set the sticky session longer than the round duration. The fourth layer is logging: the raw HTML, the query, the exit label, the timestamp and the time zone are stored together.