An observation is only evidence if someone else could reproduce it. Log a fixed set of fields for every capture: URL, UTC timestamp, intended country, the exit IP you verified, product used, browser profile, language setting and a screenshot or saved HTML. Verify the exit before each batch with the IP address checker, which shows what a server sees from your connection; open it through the same proxy. If the address is your own, the proxy was bypassed and the whole batch is invalid. Registry data from the IP information tool then confirms who holds the address, though it is not a precise geolocation.
Repeat everything. A single capture can reflect an A/B test, a regional promotion that ends at midnight or a temporary outage. Capture each cell on at least two different days and at different times, and keep results only where the observations agree; mark the rest as unstable rather than averaging them. Compare against a control taken without a proxy from your own network, so you know how much of the difference comes from the vantage point. Treat anything seen once as a lead to investigate, not a finding to present.
Keep the research itself clean. Observe public pages only, do not register accounts under invented identities, and do not collect personal data from pages you view. Respect each site’s terms and keep request rates modest; research visits should look like a small number of ordinary visitors, not a crawl. If a study requires logged-in views, use accounts your organization legitimately holds or ask the site owner for access. A proxy changes where you appear to connect from; it does not change what you are allowed to do on the site.