6.2 Research under surveillance

Curiosity without clustering

Research leaves trails. Some are obvious, like a search engine log tied to a device or account. Others are quieter: a pattern of sites visited in quick succession, or a set of queries that share distinctive wording.

"Clustering" is how those trails get grouped, whether by a company building a profile or a service trying to spot automated behaviour.

You are not trying to erase every trace (that is rarely realistic). You are trying to avoid tight, distinctive clusters that tie a person, a topic and a moment in time together.

Passive reading versus active searching

• Passive reading is consuming what is already in front of you: following a link from a newsletter, reading a printed article, browsing a library's public catalogue without logging in.
• Active searching is explicitly telling a service what you want, usually through a search engine or a site's own search box.

Active searching sends a much clearer signal. A query like "local council surveillance cameras procurement" reveals far more than arriving at a council procurement page from a general news site.

Quick comparison:

• Reading a mainstream newspaper piece about facial recognition: passive.
• Searching for "facial recognition supplier contract Manchester 2024": active, with a precise intent that is easy to classify.

Myth: passive reading is anonymous by default. Reality: IP addresses, device fingerprints and referral headers can still link the visit to a wider profile. The advantage is that the signal is weaker and less specific, so it blends into ordinary behaviour more easily.

In practice:

• Do background reading passively wherever you can.
• Keep a running list of things to search later.
• Use generic sources early on: books, printed reports, general-interest coverage.
• When you do need to search, treat it as a deliberate act with its own hygiene, not something done on impulse halfway through a browsing session.

Query shaping

Query shaping is adjusting how you search so what you reveal is less unique. Not lying, not typing nonsense: just avoiding overly specific queries that act like a fingerprint.

In the UK, searching for a local authority report by full title, year and committee reference can be a single, unmistakable signature. You can often reach the same document by searching more broadly and narrowing with a site's filters or category structure.

Two simple techniques:

• Search in stages. Start generic ("parking enforcement contract report"), then refine using the site's navigation or document list.
• Use less distinctive language. "council contract report parking enforcement", rather than the full tender reference.

That makes it less likely a query can be linked across services, and still gets you there. The trade-off is time: more reading, more irrelevant results. That is the price of a less distinctive trail.

Do not overdo it. Avoid precise searches entirely and you may miss key documents or misunderstand a topic. Precision has value, so choose where it is worth the signal:

• Low stakes (learning the basics, checking a definition): go broad.
• Accuracy matters: be precise, but limit how many highly specific queries you make in a short period.

Timing discipline

Timing is part of the signal. A burst of related searches at 2 a.m. followed by PDF downloads makes a strong cluster, even if each query is modest. Nothing sinister about it; it is just a pattern.

A more disciplined approach:

• Space it out. Read one or two items, stop, come back later.
• If you use several devices, do not run every search from the same device in the same session.

It helps with practical security too. Many services react to unusual activity in a short window by asking for verification, which can tie the activity to an account or phone number. Slower research is less likely to trigger that.

Not a guarantee: large platforms still log and correlate over longer timeframes. But it makes a single intense burst less distinctive.

The cost is inconvenience. Spreading work out takes longer and can break concentration, and there is no technical fix for that. It is a behavioural trade-off. For many people the right balance is a planned window for focused, precise searching, then back to passive reading and note-taking afterwards.

Archives and mirrors

• An archive is a stored copy of a page or document, often kept after the original changes or disappears.
• A mirror is a separate host holding the same material.

Both reduce how often you touch the original site, and help you avoid direct contact with sites likely to log or profile visitors.

For example:

• use a public web archive to read a policy announcement that is known to be updated often;
• use a university repository mirror rather than a vendor's marketing site;
• in the UK, many public bodies publish in several places. A council report may sit on the council website, in a shared document store and in a library catalogue. Picking the least intrusive source is a reasonable, low-friction step.

Myth: archives are neutral and always trustworthy. Reality: they can be incomplete or out of date, and mirrors can be tampered with.

When accuracy matters, cross-check: compare the archive with the current version or another source, and verify critical data or dates against a second copy. Simple, but it takes discipline. Treat archives as convenience and resilience, not an unquestioned source of truth.

Some sites block archiving or forbid redistribution, so archives may be missing key pages. Then direct access is the only route. If you must visit the original, narrow what you reveal:

• do not log in to an account;
• read in a private browsing session;
• keep the visit limited to its purpose.

That does not prevent logging, but it keeps the footprint narrower and avoids binding the visit to a wider identity profile.

Stronger separation and identity gates

The habits above make your trail less distinctive. Sometimes, though, the right answer is stronger separation between the research and your everyday identity.

• Tor Browser is the most robust everyday tool for this. It routes your traffic through several relays so no single point sees both who you are and what you are looking at, and it resets identifying state between sessions. For genuinely sensitive reading (where you would not want the visit linked to your home connection at all), it is a meaningful step beyond private browsing, which only clears local traces.
• A reputable VPN, with the capabilities and limits set out in 5.2, is lighter-weight. It changes where your connection appears to come from, without the stronger guarantees Tor provides.

Neither makes you anonymous if you then log in to an account that identifies you. That is the most common way people undo their own separation.

A newer complication: some research now hits identity gates, not just logging. As the age-assurance rules in 17.2 spread, more lawful material sits behind a check asking you to prove your age, and often your identity, before you can read it at all.

That turns a quiet, deniable visit into a documented one, which is exactly what a careful researcher wants to avoid. When you hit a gate, the data-minimisation guidance in 17.2 applies:

• prefer the least-revealing verification method;
• do not tie the check to accounts that already identify you.

Accessing the same material from a non-UK connection, which is lawful for an adult, will often mean the gate is not shown, for the reasons explained in 17.4.

The wider point: research hygiene is no longer only about blending into logs. Increasingly, it means deciding how much identity you are willing to hand over just to read something.

Putting it into daily practice

These habits are for ordinary research too, not only sensitive topics. Comparing broadband providers, reading about a medical condition, learning about a local political issue: the same patterns apply.

Passive reading first, shaped searches when needed, and deliberate timing keep curiosity from turning into a tightly defined profile.

The trade-offs are real. You may take longer to find things, and sometimes a precise search is the only practical path. That is a rational choice. The point is to notice when a query will be unusually identifying, and decide whether the accuracy is worth the extra signal. In a digitally monitored world, that judgement matters as much as any technical tool.