In 2022, remote-work groups kept circling back to the same question: "how is the internet in [country]?" The answers were almost always low-quality data. Someone would post a Speedtest screenshot from a hotel lobby. Someone else would say it was fine. A third person would say it was terrible. None of them were measuring the same thing, and none of it was decision-grade information.
That matters because bad internet does not usually mean no internet. It means video calls that drop at 4 p.m. when the neighborhood loads up. It means upload speeds that look acceptable until you try to push a file while your VPN is running. It means one provider works and another does not, and nobody will tell you which is which until you have already signed a lease.
If your income depends on staying connected, you need a better method than asking strangers online.
What other people's speed tests actually tell you
A Speedtest screenshot from a cafe in Medellin tells you what that one connection did at that one moment. It does not tell you anything about the building you are considering, the ISP that serves it, the congestion pattern in that district, or what happens when it rains.
I have seen people relocate based on a forum post showing 200 Mbps in a city where their actual apartment got 11 Mbps with dropouts every evening. The problem was not the country. It was the building's wiring, the ISP, and the fact that the landlord's router was from 2014.
The only defensible approach is to test the actual connection you would be using, at the times you would be using it, against the workloads that actually matter to your job.
The metrics that matter
For most remote workers, the relevant metrics are not what marketing materials emphasize. Download speed matters less than people think. What matters more:
- Upload speed. Video calls, screen sharing, and file syncing all depend on upload. Many connections advertise high download numbers but offer upload speeds that are a fraction of that.
- Latency. A 300 Mbps connection with 180 ms of latency will feel worse on a video call than a 30 Mbps connection with 20 ms of latency. This is the number most people never check.
- Evening congestion. Nearly every residential ISP slows down between roughly 6 p.m. and 11 p.m. local time. If you work hours that overlap with that window, you need to know how severe the drop is.
- Uptime and recovery. Brief outages happen everywhere. The question is how often they happen and how long recovery takes. A connection that drops for two minutes once a week is different from one that drops for forty-five minutes every few days.
Test the connection before signing
The most reliable approach I found was a short-stay test. Rent for a week or two in the exact neighborhood you are considering and go in with a plan:
- Run speed tests at multiple times of day, especially during peak evening hours and during your actual working hours.
- Test with a VPN active. Some ISPs throttle VPN traffic, and if your employer requires one, a fast connection without VPN is irrelevant.
- Make a real video call. Not a quick test. A 30-minute call during business hours, ideally with screen sharing active.
- Check upload specifically. Many speed tests default to showing download prominently. Look for the upload number.
- Ask about the ISP. If the apartment uses one provider and the building next door has a different one, the experience can be completely different.
If you cannot do a short stay, at least confirm the ISP name and plan with the landlord, then search for that provider's reputation in that neighborhood specifically. City-level reviews are too broad.
Why a mobile backup matters
In several countries where the fixed-line internet was unreliable, mobile data was actually more consistent for work.
In parts of Southeast Asia and Latin America, a local SIM with a strong 4G or 5G signal sometimes outperformed the wired connection in the apartment. That does not mean you should plan to work from a phone hotspot permanently, but having a tested mobile backup can be the difference between a lost workday and a minor inconvenience.
The cost is usually minimal. A prepaid data SIM in most countries runs between $10 and $30 a month for enough data to cover emergency work use. That is cheap insurance.
Weather is part of the internet equation
One thing I never saw mentioned in any nomad forum at the time: in tropical and subtropical countries, heavy rain can knock out internet for hours. This is not a dramatic once-a-year event in many places. It is a regular pattern during rainy season.
In parts of Mexico, Central America, and Southeast Asia, afternoon storms between May and October reliably degrade or cut connections. If your critical meetings happen during those hours, you need to know that pattern exists before you sign anything.
The fix was usually a combination of a wired connection from a competent ISP and a mobile backup on a different network. If both fail simultaneously, that tells you more about the location than any forum post could.
The practical checklist
Having moved through several countries for most of my adult life, my personal checklist stabilized:
- Fiber-optic line to the building, not ADSL or cable resold through the landlord's arrangement
- A separate router I could control, not a shared access point for the whole building
- Upload speed above 10 Mbps consistently, not just on a good afternoon
- Latency under 50 ms to wherever my main work servers sat
- A tested mobile backup SIM from a carrier with strong signal at the apartment
That list is not ambitious. It is simply a minimum operating standard. The goal is enough reliability that connectivity stops being a daily source of operational drag.
January 2026 update: Starlink Mini changed the backup equation
When I originally wrote this article, the backup I recommended was a local SIM card. That is still a solid option. But Starlink Mini has changed the picture for anyone who takes connectivity redundancy seriously.
The Mini weighs 1.1 kg (2.43 lb) without the kickstand. The dish itself is roughly 30 × 26 cm, about the footprint of a large laptop. It fits in a backpack. It draws 25 to 40 watts on average, low enough to run off a portable power station for several hours, and it accepts DC input (12–48V) or USB-C Power Delivery with a cable accessory. That last part matters because it means you can technically power it from a high-output USB-C battery pack designed for laptops, though you will drain it fast.
The practical constraint is sky. Starlink Mini has a 110-degree field of view, which means it needs to see a cone of open sky extending roughly 55 degrees from directly overhead in every direction. You do not need a perfectly clear horizon. A balcony with a partial roof overhang can work if the dish can see most of the sky above it. A courtyard surrounded by tall buildings probably will not. The Starlink app has a built-in obstruction scanner that uses your phone's camera to show you exactly how much sky is blocked at a given spot before you even power on the dish. Use it. A few minutes of scanning can save you from signing a lease on a place where the backup you are counting on will not connect.
In countries where a mobile SIM used to be the only realistic backup, a two-layer setup now makes more sense. A local SIM covers brief outages where you just need to stay on a call. Starlink Mini covers the scenario where the fixed-line ISP goes down for hours, or where the local mobile network is also degraded, which happens more often than you would expect during storms or infrastructure failures.
Starlink Mini is not available everywhere. As of early 2026, it is sold in most of the Americas, Europe, and parts of Asia and Africa, but some countries restrict or have not yet licensed satellite internet service. Check the Starlink availability map before assuming you can use it at your destination. In countries where it is available, the monthly plan with roaming typically runs between $50 and $150 depending on region and plan type.
DNS is often the bottleneck
Separate from your physical connection, there is a software-level improvement that costs nothing and takes two minutes.
Most ISPs route your DNS queries, the lookups that translate domain names into IP addresses, through their own servers. In many countries, those servers are slow, overloaded, or configured to inject redirects and advertising. Some ISPs also use DNS-level filtering to block certain services, which can interfere with VPNs and work tools in ways that look like a broken connection but are actually a policy decision by the provider.
Switching your device's DNS to Cloudflare's 1.1.1.1 resolver often produces a noticeable improvement. It is consistently ranked as one of the fastest public DNS resolvers, and Cloudflare says it does not sell resolver user data, truncates source IP addresses, and deletes public resolver logs within 25 hours. On most devices you can change DNS in network settings in under a minute. On a phone, the free 1.1.1.1 app handles it automatically.
Cloudflare also offers WARP, which goes further. WARP encrypts the traffic between your device and the nearest Cloudflare edge server, effectively replacing the first and often worst hop of your internet path with Cloudflare's own optimized network. It is not a full VPN and it is not designed to mask your location for geo-restriction purposes, but it can bypass ISP-level throttling, DNS hijacking, and poor local routing. The free tier is unlimited and available on every major platform including Linux.
I have seen WARP make a measurable difference in countries where the local ISP's routing was poor. Video calls that had intermittent packet loss over the raw connection ran cleanly through WARP because the traffic was no longer bouncing around the ISP's congested internal network. It is not a guaranteed fix, but it is free and reversible, so it is worth testing in every new location.
Between a tested wired connection, a mobile SIM backup, Starlink Mini for major failures, and WARP running on every device, the connectivity setup available to remote workers in 2026 is far more resilient than what existed when I first wrote this piece. The core advice has not changed: test everything before you commit. The difference is that the fallback stack is now much better.
What usually goes wrong
The risk was never that a country had bad internet in general. Almost every country remote workers considered in 2022 had adequate connectivity somewhere. The risk was that you picked the wrong apartment, the wrong ISP, or the wrong neighborhood and only discovered it after you had committed.
Testing that before you sign is boring. It is also the single most useful thing you can do before relocating for remote work.
If your job depends on the connection, treat it like what it is: critical infrastructure. Everything else, the food, the weather, the rent, the social life, sits on top of that dependency.