A VPN and Tor produce the same visible result. The site you connect to records an address that is not yours, and your internet provider sees an encrypted tunnel instead of a readable list of destinations. That similarity is why the two end up in the same sentence so often, usually with an argument about which one is more private attached.

That argument does not have an answer, because the two are not competing on the same axis. One concentrates trust in a company you choose. The other spreads it thinly across strangers nobody chose. Almost everything else follows from that.

One hop versus three#

A VPN is one relay. Your client encrypts traffic to the provider's server, the server decrypts it and makes the request on your behalf, and the answer returns the same way. One company operates that server, and it is the only machine in the path holding your real address and your destination at the same moment. It holds both for every request you make, for as long as you stay connected.

Tor is a chain. The Tor Project describes it as passing your traffic through at least 3 different servers before sending it on, with a layer of encryption removed at each one by the relay it was addressed to. The first relay knows your address but not your destination. The last relay, the exit, knows the destination but not who asked for it. The relay in the middle knows neither end, only the two relays either side of it.

The same page puts the contrast bluntly: ordinary proxy servers "create a single point of trust and failure," whereas a relay in a Tor path leaves you "still protected from this node figuring out both who you are and where you are going on the Internet."

Tor paths are also temporary. The client will reuse the same circuit for new TCP streams for 10 minutes while it is working, then build a fresh one, and it switches immediately if a circuit fails. A single long-running connection is the exception: it stays on the circuit it started on. A VPN session is normally one tunnel to one server until you disconnect it.

The trust question#

A VPN asks you to trust one company, completely, and to keep trusting it. That provider can see the shape of everything you do, can record it, and can be compelled to hand over what it recorded. A no-logs policy is a promise about behaviour, not a property of the system. You cannot check it from your side of the tunnel, and no amount of encryption between you and the server changes that, because the server is the party doing the decrypting.

Tor asks for something narrower: that not all three relays in your circuit are the same adversary. Relays are run by volunteers who are mostly strangers to each other, frequently in different countries and legal systems. That is a weaker requirement in the way that matters, because no single operator failing, lying, or receiving a court order is enough on its own.

It is a stronger requirement in another direction. An observer positioned to watch traffic entering your first relay and leaving the exit at the same time can line up the timing and volume of both. That is a documented limit of low-latency anonymity networks rather than an oversight, and Tor does not claim to defeat it.

So the trade is one accountable party you pick and pay, against a distributed set nobody controls. "Which is more private" collapses into "which failure would you rather survive."

What Tor will not carry#

A VPN moves IP packets. Whatever your machine emits goes down the tunnel: TCP, UDP, ICMP, some game protocol nobody has heard of. Nothing in the path needs to understand the traffic in order to forward it.

Tor does not work like that. It delivers TCP streams and a narrow subset of DNS requests, and that is the list. Anything that depends on UDP has no route through the network, which puts QUIC and most real-time voice and video outside what a circuit can carry.

The reasons are documented and specific. Carrying arbitrary IP would mean building a secure Intrusion Detection System so exit operators could still enforce exit policies, and those policies are a large part of why anyone is willing to run an exit at all. Raw packets leak host detail too: "IP packets reveal OS characteristics. We would still need to do IP-level packet normalization, to stop things like TCP fingerprinting attacks." And DNS cannot be passed straight through, because Tor "must rewrite DNS requests so they are delivered to an unlinkable DNS server rather than the DNS server at a user's ISP."

That last point is worth holding onto when you compare leak behaviour. A misconfigured VPN can quietly send name lookups to your provider's resolver while everything else goes through the tunnel, which is the exact failure a DNS leak test is built to catch, and which our guide to DNS leak testing walks through. Tor has no separate resolver path to forget about, so that particular mistake is not available to you.

The same limitation explains the standing advice against torrenting over Tor. The Tor Project calls it unsafe and not recommended, both because most torrent clients hand your real address to peers directly, and because the load and the legal exposure land on volunteer exit operators.

Why Tor feels slow#

Three relays instead of one means three layers of encryption instead of one and, more importantly, a path chosen for diversity rather than for distance. Your circuit can cross continents twice on the way to a destination that was two networks away. Relay capacity is donated rather than provisioned, so the slowest hop in the chain sets your ceiling.

None of that is a tuning problem waiting to be solved. It is the price of the property Tor is buying. A commercial VPN can put a well-provisioned server near you precisely because one company decides where the servers go, and that is the same centralisation that makes it a single point of trust.

What the far end sees#

Here the two invert, and it surprises people.

Tor exit relays are public. The Tor Project publishes the current exit list, and any site can download it and compare against it. Our own VPN and proxy check does exactly that: it tests the address against that published list and returns a Tor-exit verdict outright, alongside the AS number and organisation behind the address. Identifying Tor traffic is a lookup, not an inference.

Commercial VPN endpoints are not published anywhere. A site has to work them out from the network the address sits on, the hosting organisation that owns it, and reputation data, which is an educated guess that can be wrong in both directions. We covered that machinery in how VPN detection works, and the related question of why some proxy addresses are effectively invisible in residential vs datacenter proxies.

Put the two facts together and the shape is clear. Tor gives you the stronger anonymity property and is trivially identifiable as Tor, which is why some sites refuse it wholesale. A VPN is harder to identify with confidence, but the identification you actually care about has already happened at the provider, which knows exactly who you are. You can see what any given address looks like from the outside with an IP address lookup.

Running both at once#

People ask about stacking them, and the order changes what you get.

Connecting to a VPN first and running Tor through it hides your Tor use from your internet provider, and the first relay sees the VPN's address instead of yours. What it does not do is improve your position against an observer watching both ends of the circuit. It also reintroduces a company that knows you use Tor and when, which is the knowledge you were trying to distribute in the first place.

The reverse, tunnelling to a VPN through Tor, is harder to configure and usually worse. Reaching the VPN means authenticating to it, authenticating means an account, and an account is a durable identity bolted to the end of an anonymous path.

Neither arrangement is a straight upgrade. Both add a party to the list of parties who have to behave.

Picking one for the job in front of you#

A VPN is the right tool when you want a stable address in a country you choose, when you are on a network you do not control and would rather its operator saw less, or when the tunnel has to carry protocols Tor cannot. It is also the only one of the two that is comfortable to leave on all day.

Tor is the right tool when the thing you are protecting is the link between you and one specific destination, and when you are not willing to hand that link to any single company. Use it through the Tor Browser rather than pointing an ordinary browser at the network, because the browser handles application-layer leaks the network itself was never designed to fix.

Both leave a large category untouched. Neither hides you from a service you log into, and neither erases what your browser volunteers to every site it visits. That is a separate layer with separate tools, which is what a browser fingerprint check and a WebRTC leak test are for. If you are weighing the options from the user's side rather than the architecture's, how to hide your IP address compares them without the marketing.

The short version: a VPN changes who can see you. Tor changes how many people would have to agree before anyone can.

Try it

Check what an address looks like

Test any address against the Tor Project's published exit list, and against datacenter and VPN networks. You get the verdict, the confidence behind it, and the AS number and organisation it was based on, so you can judge the reasoning rather than trust a label.

Run a VPN and proxy check →