WebopsyScan your site

Research · updated November 22, 2026 at 10:14 a.m.

What 22,496 launched sites do before you click “accept”

Sites launched publicly (on Product Hunt, Hacker News, Y Combinator, Reddit, GitHub and AI app builders), each loaded in a real browser with no consent given. 24% ran analytics or advertising trackers before the visitor agreed to anything. 43% had no privacy policy a reader could find.

By where the site came from

Share of sites in each group. n is the number of live sites scanned.

Built with AI app builders
n = 7,652
Product Hunt and Hacker News launches
n = 4,410
Y Combinator companies
n = 1,382
Indie projects (Reddit, GitHub)
n = 9,052
All sites
n = 22,496
Trackers run before consent19%28%40%24%24%
A session recorder runs before consent2%4%2%3%3%
No readable privacy policy57%41%32%34%43%
Policy is missing expected disclosures32%44%50%48%42%
No terms of service64%48%36%44%51%
Missing security headers78%80%84%76%78%
A critical accessibility failure7%7%7%7%7%
Clears the bar (Clean Bill of Health)12%18%12%21%17%

What fires before consent

Number of sites where each service ran on page load before any consent choice: seen sending data or setting its cookie in the browser, or loaded on a page with no consent mechanism at all. A site can run several.

  1. Google Analytics3,574
  2. PostHog1,465
  3. Google Ads1,008
  4. Meta Pixel587
  5. Microsoft Clarity534
  6. LinkedIn Insight252
  7. Reddit Pixel222
  8. HubSpot94
  9. TikTok Pixel84
  10. X Pixel75
  11. Segment57
  12. Amplitude56

What privacy policies leave out

Among the 11,680 sites with a readable English policy, the disclosures most often not found.

  1. The right to complain to a regulator60%
  2. Children's data43%
  3. A legal basis (EEA/UK)40%
  4. International transfers37%
  5. Whether data is sold or used for ads27%
  6. A last-updated date20%
  7. Why it is used18%
  8. Users' rights15%

Policies that do not describe the site

Of 2,516 sites where trackers fired and the policy could be read, 16% run an ad pixel or session recorder their own policy never mentions.

Method.

Sites are discovered from public launch lists (Product Hunt, Hacker News “Show HN”, the Y Combinator directory), from certificate-transparency logs for app-builder domains (lovable.app, bolt.host and others) and from public posts. Each home page is loaded once in headless Chrome with no cookies and no clicks; the browser's network log and cookie jar, plus the tags in the page when there is no consent mechanism at all, are what “ran before consent” means here. A Google Consent Mode ping that reports storage denied is not counted. Policies are fetched (and rendered when they need JavaScript) and checked for fourteen disclosures GDPR, CCPA and PIPEDA expect, as text heuristics. Sites that could not be reached, sat behind a bot wall, or served a platform placeholder are excluded. Numbers move as the scanner re-checks sites.

A finding here is what is visible from outside on the day of the check, not a legal conclusion. The scanner is open about what it can and cannot see: coverage. Run it on any site, free: webopsy.tech.