Sharp edgestested 2026-05-24 · public-tough-v1 · proxy: none

SeleniumBase

Beats Playwright on Amazon, fails murkier everywhere else.

SeleniumBase wraps Selenium with a large convenience and stealth toolbox. We ran 4.49.2 through its Driver API against the 21-site corpus, no proxy, headless.

It passed 4 of 21 — Amazon, Target, Rightmove, Wikipedia. Amazon is notable: both Playwright and Patchright drew 503s there the same day. Whatever SeleniumBase's driver does differently, one hard target noticed the difference.

The concerning pattern is how it failed. Seven targets — eBay, Booking, Expedia, Zillow, Realtor, Trustpilot, Reddit — didn't return clean block signals; they returned pages missing the expected content. Half-loaded results are the worst failure mode for a scraper because they can look like success to naive pipelines. Indeed served a Cloudflare interstitial, Yahoo the consent wall, and the captcha sites their usual checks. Getting trustworthy results out of SeleniumBase on hard sites means per-site waits and content assertions.

Test conditions

21 real public sites (public-tough-v1), 2026-05-24, no proxy, single residential IP, Linux. Deterministic validators: expected status code plus required and forbidden content markers. Pass rates measure the tool and the network position together. Without a proxy, results reflect how each site's protections treat one plain residential connection as much as the tool itself.

What worked

  • +Passed Amazon search where Playwright and Patchright both failed on the same day.
  • +Target, Rightmove, and Wikipedia rendered and passed cleanly.
  • +Mature ecosystem with many built-in evasion and convenience options we haven't yet exercised (UC mode variants, etc.).

What failed

  • Seven of seventeen failures were half-loaded pages with missing content rather than clean refusals — ambiguous results that require your own validation layer.
  • Cloudflare interstitial on Indeed; captcha walls on Walmart, AliExpress, Airbnb, Tripadvisor, Yelp, CoinMarketCap, YouTube.
  • Headless without a proxy, its extra machinery bought only one extra pass over stock Playwright.

When to use it

When its specific tricks (undetected-driver modes, form/flow helpers) match your target, and you're prepared to write content assertions for every page. Not a set-and-forget stealth solution.

Real capabilities, murky failure modes — verify everything it returns.

Run evidence

Every attempt below is backed by saved artifacts: request/response metadata, body content, and (for browsers) screenshots. Raw JSON and artifacts live in the runner data directory.

Run · seleniumbase-common-sites-v1-20260524

4/21 passedversion 4.49.2proxy: none2026-05-24

adapter: seleniumbase.Driver.get

TargetResultDetailLatency
amazon.comamazon_search_laptoppassstatus 2001.9 s
walmart.comwalmart_search_laptopfailblock marker: “Robot or human”996 ms
ebay.comebay_search_laptopfailexpected content missing123 ms
aliexpress.comaliexpress_search_laptopfailblock marker: “Robot”4.6 s
target.comtarget_search_laptoppassstatus 2003.0 s
booking.combooking_search_parisfailexpected content missing380 ms
airbnb.comairbnb_search_parisfailblock marker: “captcha”3.1 s
tripadvisor.comtripadvisor_restaurants_parisfailblock marker: “captcha”639 ms
expedia.comexpedia_hotels_parisfailexpected content missing913 ms
zillow.comzillow_homes_for_salefailexpected content missing847 ms
realtor.comrealtor_new_yorkfailexpected content missing1.1 s
rightmove.co.ukrightmove_londonpassstatus 2003.7 s
indeed.comindeed_python_remotefailblock marker: “Cloudflare”98 ms
yelp.comyelp_sf_restaurantsfailblock marker: “captcha”528 ms
trustpilot.comtrustpilot_amazonfailexpected content missing441 ms
newyork.craigslist.orgcraigslist_nyc_bikefailblock marker: “blocked”412 ms
finance.yahoo.comyahoo_finance_aaplfailredirected to consent.yahoo.com1.1 s
coinmarketcap.comcoinmarketcap_homefailblock marker: “captcha”1.5 s
reddit.comreddit_webscrapingfailexpected content missing443 ms
youtube.comyoutube_search_scrapingfailblock marker: “captcha”1.9 s
en.wikipedia.orgwikipedia_web_scrapingpassstatus 200633 ms