We Pointed 120 Free Proxies at 12 Travel Sites. 31 Reached Any of Them

The travel industry runs on scraped prices, and a lot of people try to do it for free. We run a free proxy list and a verification engine, so instead of guessing how far a free proxy gets against a real travel site, we pointed 120 of them at twelve of the biggest and counted. The short answer: almost nowhere. Only 31 of the 120 reached even one of the twelve sites, and against every major online travel agency and hotel chain in the test, not a single free proxy returned real content.

This is the measured version, method and caveats included, and it is the travel companion to our broader free proxy data study. If you want the working setup instead, it is in how to scrape flight prices and how to scrape hotel prices.

What we did

On 24 August 2026 we took 120 HTTP proxies straight from our own free proxy list, each verified alive minutes earlier, and sent each one a single request to a real search or booking page (not just a homepage) on twelve travel sites: four online travel agencies (Booking.com, Expedia, Agoda, Hotels.com), four metasearch engines (Kayak, Skyscanner, Trivago, Google Flights), two airlines (Delta, United) and two hotel chains (Marriott, Hilton). Each request had a 22-second timeout. That is 1,440 proxied requests, no logins, no sustained scraping.

We classified each response into one of three outcomes:

• Reached: the proxy connected and the site returned a substantial page (over 1,500 bytes) with no visible bot-challenge marker.
• Blocked: the site answered with a challenge or rejection, an HTTP 403, 429, 503 or 202, or a page carrying a CAPTCHA, Akamai, AWS WAF, PerimeterX or similar marker.
• Proxy failed: the proxy never completed a connection at all.

One honest caveat up front, because it makes the numbers worse rather than better: "reached" is generous. It means the proxy got a real page back, not that the fares or hotel listings rendered on it. Many travel pages ship as a near-empty JavaScript shell, so a plain request that "reached" a site may still carry no usable prices. The real rate of getting usable data is lower than the reach rate below.

The headline: 31 of 120 reached anything

Across all twelve sites, only 31 of the 120 proxies reached even one of them. Forty-eight reached none at all, failing every single site in the test. The rest landed somewhere in between, reaching one or two of the twelve and failing the other ten.

Reached, by site

The per-site picture is where it gets stark. Six of the twelve sites, every OTA and both hotel chains, returned real content to exactly zero of the 120 proxies. The only sites any free proxy reached were three of the four metasearch engines and one airline, and two of those barely.

Trivago is the outlier, reached by 20 of 120, because its entry page is lighter than an online travel agency's search results and lets more traffic through before any real defense engages. Delta let 14 through to a real page. Kayak let exactly one. Everything else, including the four OTAs that most hotel scraping targets, returned real content to none of the 120 free proxies.

Why they failed: mostly death, then blocking

Splitting the failures by cause tells the free-proxy story precisely. Across the four online travel agencies, 401 of the 480 requests were the proxy simply never connecting, and the other 79 were the site returning a challenge. Not one of the 480 reached a real OTA page.

That pattern, death first and blocking second, held across the whole test. The dominant outcome everywhere was the proxy dying before it got anywhere, which is the true cost of free: not that travel sites refuse you, but that the tool disappears underneath you between the check and the use. Where a site did answer, it usually answered with a challenge, because the working free pool is overwhelmingly shared datacenter space and travel bot management distrusts that on sight. Of the 120 proxies we sampled, 40 were on datacenter networks outright, and the pool skewed heavily toward a handful of countries rather than the markets a travel scraper would actually want.

The airlines and the odd United result

The two airlines split. Delta returned a real page to 14 of 120, the second-best result in the test, and challenged none, which fits an entry page that is more permissive than an OTA's until the fare search itself runs. United was the opposite extreme: not one of the 120 proxies completed a request to it at all. We report that as measured rather than over-interpret it; a site can refuse proxied connections at the network edge in a way that looks identical to a dead proxy, and the honest statement is that no free proxy in our sample got a United page, not that we know exactly why.

What it means

The takeaway is narrow and firm. Free proxies are the wrong tool for travel data, and the number is 31 of 120 reaching anything and zero reaching a real OTA or hotel-chain page. They are fine for what free proxies are always fine for: a throwaway check of whether a page loads from another country, learning how a proxy plugs into a browser, or testing a parser against a page you already saved. The moment the job is collecting real prices from Booking.com, Expedia, Marriott or Google Flights, on a schedule, free proxies do not do it.

The reason they fail is the reason residential proxies work. Every one of these sites prices by market and defends against automated traffic, so reading them needs an address that reads as an ordinary traveler, in the right country, that stays alive through the request, which is exactly what a free datacenter proxy is not and a residential proxy is. The working setup is in how to scrape flight prices and how to scrape hotel prices, and our residential proxies are $0.44/GB pay-as-you-go, no KYC, so a test run costs cents. For the throwaway checks free is genuinely good at, our free proxy list and proxy checker remain free.

Method notes and honest limits

• One request per proxy per site. A single request is a snapshot, not a sustained scrape; a proxy that failed once might succeed on a retry, and one that reached once might be challenged at volume. The direction of the result is clear, but the exact counts would move on a re-run.
• "Reached" overstates usable data. It counts a real page returned, not rendered fares or listings, so the usable-data rate is lower than the reach rate.
• The pool is what a free list is. These were live, verified free proxies, overwhelmingly datacenter and skewed by country, which is the honest composition of any free pool, not a worst case we selected.
• Sites and defenses change. This is a 24 August 2026 snapshot; a site that blocked today may behave differently next month, which is why we keep the test rig and re-run it.

Sources

• HProxy travel atlas, measured 24 August 2026: 120 HTTP proxies taken live from our own free proxy list, one request each to a search or booking page on 12 travel sites, classified as reached, blocked or proxy-failed. 31 of 120 reached at least one site; 0 reached a real OTA or hotel-chain page.
• HProxy, free proxy data study, for the pool size, composition and lifetimes behind these results.