Find where your home connection slows down
A faster plan will not fix every Wi-Fi problem. This worksheet helps you compare the same device in different conditions and decide which next check is justified. It is a troubleshooting exercise, not a diagnosis of your provider.
By A. Ash · Published October 5, 2026 · Measurement method
Before you start
- Write down your subscribed download and upload tiers separately. Use your account or contract, rather than a generic advertisement.
- Use the same capable device and browser for comparisons. Connect it to power and keep its position consistent for repeated Wi-Fi readings.
- Pause downloads, cloud backups, streaming and other competing traffic. If a VPN is enabled, record that fact. Only turn it off if your work or security requirements allow it.
- Allow for the test’s data use, particularly on metered mobile or satellite plans. Check the current estimate beside the test button.
- If your connection is completely unavailable, start with the provider outage links. A browser test cannot measure an offline connection.
Run three comparison sessions
1. Establish a wired baseline
Connect by Ethernet to the router if your device supports it. Avoid comparing a wired computer with a different wireless phone: that changes both the device and connection type. Record the cable, adapter and port capability if known. A 100 Mbps Ethernet link can be a bottleneck on a faster plan.
Run the AI Speed test twice a few minutes apart without changing conditions. Keep both results. If they differ considerably, add a third reading and look for an explanation before treating one result as representative.
2. Test Wi-Fi beside the router and in the affected room
Disconnect Ethernet and test near the router, then in the room where calls or streaming fail. Record the Wi-Fi band if your device exposes it; otherwise write “unknown.” Do not assume a specific band from the network name. Use the same device, test service and workload conditions.
If you cannot run a wired test, this comparison can still reveal a location effect. It cannot separate all router, radio and provider problems as clearly as a wired baseline.
3. Repeat at the time the problem usually happens
Repeat your wired baseline and affected-room test during the problematic period. A quiet morning and a busy evening are different conditions; label them clearly. Keep other household traffic similar if you can. Otherwise record who was streaming or uploading so the comparison has context.
Compare your own readings
Enter results from the same device and test service. This record stays in the open page and clears on reload; download it before leaving. It is not sent to AI Speed.
Your connection comparison
| Connection | Download | Upload | HTTP ping |
|---|---|---|---|
| Ethernet | No readings | No readings | No readings |
| Wi-Fi near router | No readings | No readings | No readings |
| Wi-Fi affected room | No readings | No readings | No readings |
Choose the next check
- If repeated Wi-Fi download or upload readings fall below Ethernet under the same conditions, compare Wi-Fi near the router with the affected room before changing your plan.
- If upload drops while download looks normal, pause cloud backups and other uploads, then compare again. A higher download tier may not solve an upload bottleneck.
- If latency changes while throughput stays similar, check the application’s own latency indicator. HTTP ping here measures a different route from a game or call.
- If all connections slow down, repeat the wired test with a second testing service and at a different time before contacting your provider.
Two or more readings per connection are more useful than a single sample. A difference suggests a next investigation; it does not prove a provider fault. Check timestamps and conditions before comparing medians. This worksheet does not apply a pass/fail threshold.
Copy this record into your notes
Use one row per test. Copy the table or make the same columns in a spreadsheet. Keep your account number and exact address out of anything you share publicly.
| Date & time | Connection / room | Download Mbps | Upload Mbps | Ping / jitter ms | VPN / other traffic / device |
|---|---|---|---|---|---|
| Record time | Wired, quiet period | — | — | — / — | Record conditions |
| Record time | Wi-Fi near router | — | — | — / — | Record conditions |
| Record time | Wi-Fi affected room | — | — | — / — | Record conditions |
| Record time | Wired, problem period | — | — | — / — | Record conditions |
| Record time | Wi-Fi, problem period | — | — | — / — | Record conditions |
Use the pattern to choose your next check
Wired looks consistent, but the affected room is slower
This points toward the local Wi-Fi path as a useful next investigation. Move the router into a clearer location, repeat the near-router reading, and compare another device in the affected room. A mesh node or access point may help, but test placement before buying equipment or upgrading the plan.
Both wired and Wi-Fi readings are repeatedly low
Check the subscribed tiers, Ethernet link speed, device capability and background traffic first. Compare another reputable test using similar conditions. Different test routes can disagree; agreement across repeated wired readings is more useful support evidence than one unusually low number.
Download looks fine, but calls fail during uploads
Pause cloud backups and large uploads, then compare again. Upload capacity and delays under load may matter more than headline download speed. This test does not directly measure packet loss or all forms of loaded latency, so do not label the issue “bufferbloat” from one idle ping result.
The evening pattern is worse on every connection type
Record whether household use also increased. If repeated wired results stay poor with local traffic paused, send the dated evidence to your provider. The pattern suggests when to investigate; it does not prove network congestion or identify its location.
Only one app or game has problems
Check that service’s status and its own connection indicators. AI Speed’s HTTP latency measures its test route, not that app’s server route. A healthy general test does not rule out a service-specific routing or software problem.
Prepare a useful support request
Give your provider the subscribed tiers, affected times, device and connection type, the repeated readings, and the checks you already performed. Ask them to check the line or equipment and explain any next step. Share account details only through the provider’s authenticated support channel.
My plan lists [download] / [upload] Mbps. On [dates and times], this device connected by Ethernet measured [results] with downloads and VPN [conditions]. Wi-Fi in [room] measured [results]. I also checked [outage status, cable or other device]. Please help identify the next line or equipment check.
Before switching plans, use Compare providers to check address eligibility, upload tiers, equipment and total cost. For more checks, open the connection diagnostic or slow-internet guide.
Limits of this exercise
There is no single percentage below an advertised tier that diagnoses every connection. Test endpoint capacity, device limits and the route affect throughput. Ping here is an HTTP round trip; jitter is variation between HTTP samples. This worksheet provides observations for troubleshooting, not a service guarantee, contractual conclusion or measurement of every network condition.