Research

Startup website security: 4,160 sites scanned (2026)

We scanned 4,160 startup websites. 65% send no Content-Security-Policy, 54% let any other site load them in a frame, and 59% load trackers before the visitor makes a consent choice. Only 27% send all four basic security headers, and 29% of the domains publish no DMARC record.

Updated ยท Scans from September 14, 2026 to November 14, 2026

Key figures

65.3%

No Content-Security-Policy header

54%

Can be framed by another site

59.1%

Trackers load before the visitor consents

27%

Send all four basic security headers

CSP, frame protection, nosniff and HSTS.

Security headers

What the homepage response tells a browser to enforce.

Security headers, share of sites

  • No CSP header65.3%
  • Frameable54%
  • No nosniff46.8%
  • No HSTS34.6%
  • None of the four22.3%
  • No HTTPS redirect6%
Show the data table
Security headers: share of sites failing each check
FindingSitesOut ofShare
No Content-Security-Policy header2,7154,16065.3%
Can be framed by another site2,2484,16054%
No X-Content-Type-Options header1,9464,16046.8%
No HSTS header1,4414,16034.6%
None of those four headers9264,16022.3%
HTTP does not redirect to HTTPS2494,1606%

65.3%

2,715 of 4,160 sites

No Content-Security-Policy header

The homepage response sets no Content-Security-Policy.

Counted: No Content-Security-Policy header on the homepage response (sec-csp-missing).

Check yours with the Security headers checker

54%

2,248 of 4,160 sites

Can be framed by another site

Another site can load the homepage inside a frame, the setup clickjacking needs.

Counted: Neither X-Frame-Options (DENY or SAMEORIGIN) nor a CSP frame-ancestors directive on the homepage response (sec-clickjacking-frame-ancestors).

Check yours with the Security headers checker

46.8%

1,946 of 4,160 sites

No X-Content-Type-Options header

Browsers may guess a file's type instead of trusting the one the server sent.

Counted: X-Content-Type-Options is missing or not exactly nosniff on the homepage response (sec-x-content-type-options).

Check yours with the Security headers checker

34.6%

1,441 of 4,160 sites

No HSTS header

Nothing tells browsers to use HTTPS for the site every time.

Counted: No Strict-Transport-Security header with max-age of at least six months on the homepage response (sec-hsts-missing).

Check yours with the Security headers checker

22.3%

926 of 4,160 sites

None of those four headers

The homepage sends none of CSP, frame protection, nosniff or HSTS.

Counted: All four of sec-csp-missing, sec-clickjacking-frame-ancestors, sec-x-content-type-options and sec-hsts-missing failed on the same scan.

Check yours with the Security headers checker

6%

249 of 4,160 sites

HTTP does not redirect to HTTPS

A visitor who types the plain http:// address is not sent to https:// on the same host.

Counted: The first hop from http:// is not a redirect to https:// on the same host (sec-http-https-redirect).

Check yours with the SSL checker
Read the startup website security checklist

Privacy and consent

What runs before the visitor says yes, and what they can read about it.

Privacy and consent, share of sites

  • Trackers before consent59.1%
  • No privacy policy10%
  • Banner blocks page9.1%
Show the data table
Privacy and consent: share of sites failing each check
FindingSitesOut ofShare
Trackers load before the visitor consents2,4604,16059.1%
No privacy policy we could reach4184,16010%
Cookie banner blocks the page3804,1609.1%

59.1%

2,460 of 4,160 sites

Trackers load before the visitor consents

Analytics or ad identifiers are stored or read before the visitor makes any consent choice.

Counted: Non-essential identifiers are written or read on the visitor's device before any consent action, in a fresh browser session (comp-trackers-before-consent).

Check yours with the free scan

10%

418 of 4,160 sites

No privacy policy we could reach

We found no reachable privacy policy linked from the site.

Counted: No reachable privacy policy linked from the homepage (comp-privacy-policy-presence).

Check yours with the free scan

9.1%

380 of 4,160 sites

Cookie banner blocks the page

The consent banner covers the page and stops interaction until the visitor answers it.

Counted: A consent overlay traps interaction with the primary content (work-cookie-banner-blocks-interaction).

Check yours with the free scan

Email

Whether the domain tells mail servers which messages really come from it.

28.9%

1,201 of 4,160 sites

No DMARC policy

The domain publishes no DMARC record, so nothing stops mail spoofed in its name.

Counted: No v=DMARC1 TXT record at _dmarc.<domain> (sec-dmarc-policy fail). A record at p=none counts as published here.

Check yours with the Email authentication checker
Read the guide to SPF, DKIM and DMARC records

Things that break

Faults a visitor can hit on the first page they open.

Things that break, share of sites

  • Sideways scroll at 320px24.9%
  • Broken internal links5.7%
  • JS error on load4.8%
Show the data table
Things that break: share of sites failing each check
FindingSitesOut ofShare
Page scrolls sideways on a small phone1,0374,16024.9%
Internal links that lead to a 4xx error2374,1605.7%
A JavaScript error on page load2004,1604.8%

24.9%

1,037 of 4,160 sites

Page scrolls sideways on a small phone

At 320 CSS pixels wide, the page scrolls horizontally.

Counted: Horizontal overflow at a 320 CSS px viewport, the WCAG 2.2 reflow width (speed-overflow-320).

Check yours with the free scan

5.7%

237 of 4,160 sites

Internal links that lead to a 4xx error

At least one link on the homepage points to a page on the same site that answers with an error.

Counted: At least one same-site link from the homepage resolves to a 4xx status other than 429 (work-broken-internal-links-4xx).

Check yours with the free scan

4.8%

200 of 4,160 sites

A JavaScript error on page load

The homepage throws an uncaught JavaScript error while it loads.

Counted: An uncaught exception during the initial load of the homepage (work-uncaught-js-error-on-load).

Check yours with the free scan

Even security startups skip CSP

Of the 138 security products in the sample, 61% send no CSP and 29% send all four headers. Across every site it is 65% and 27%. Categories with 120 sites or more, each share taken over that category's own sites.

No CSP header, by category

  • Payments69.4%
  • AI68%
  • Design65.8%
  • Analytics65.6%
  • Developer Tools64.2%
  • Marketing64.1%
  • No-Code64%
  • Content63.5%
  • Productivity63%
  • Security60.9%
Show the data table
Security findings by site category
CategorySitesNo CSPFrameableNo HSTSTrackers before consentNo DMARCNo privacy policyAll four headers
Productivity1,13663%51.6%31%54%31.3%7.5%29.1%
AI67468%57.7%33.8%66.2%28.5%5.8%25.7%
Developer Tools59764.2%52.6%32.2%57.5%27.6%11.2%28.8%
Analytics33765.6%54.9%32.6%61.7%24.9%6.8%27%
Content29963.5%53.8%35.8%64.2%29.4%7%30.1%
Marketing25164.1%53%33.5%61%23.9%5.6%26.3%
No-Code21164%52.1%31.8%60.7%26.5%10%25.1%
Design20265.8%53%42.1%58.9%33.7%8.9%22.8%
Payments18669.4%53.8%32.8%61.8%32.8%4.8%26.9%
Security13860.9%44.9%34.8%47.1%26.8%8.7%29%

How this was measured

  • The sample is every website scanned on LaunchScaler between September 14, 2026 and November 15, 2026 whose scan completed and reached the site: 4,160 sites. Most are products found on public launch directories. The rest are sites someone entered into the free scan.
  • Each site counts once, on its most recent scan in that window. Sites LaunchScaler runs itself, and sites whose owners asked us to remove their data, are left out.
  • A scan fetches the homepage and its response headers, requests the plain http:// address, looks up the domain's DMARC record, follows the same-site links on the homepage, and loads the page in a fresh browser to see what it stores before any consent choice.
  • Each finding names the check it counts, and its card says exactly what was counted. The checks are defined in LaunchScaler's audit catalog, each with the published source behind its bar. A site counts when its latest scan failed the check.
  • The category is the one the scan derived from each site's own pages.
  • The figures are counted inside the database and only totals leave it. No site is named here, and the downloads hold the same totals.

Limits

  • A check that could not be read on a site counts as not failing, so every share here is a floor.
  • The homepage stands in for the site. Inner pages can differ.
  • A DMARC record set to p=none counts as published, so the email figure covers only domains with no record at all.
  • The consent figure is a technical reading of the first page load, not a legal finding.
  • These are sites that reached LaunchScaler, mostly from launch directories, not a random sample of the web.

Download the data

The findings file has one row per finding, plus one for the sites that send all four headers, with the count, the sample, the share, the catalog checks and the definition. The category file has one row per site category, with each count and its share.

How to cite

Suggested citation
LaunchScaler, "Security of startup websites, November 2026," November 15, 2026. Sample: 4,160 websites scanned September 14, 2026 to November 14, 2026. https://www.launchscaler.com/research/startup-security

Updated . Snapshot file startup-security-2026-11.

Companion study

AI crawler readiness of 4,026 new product sites (2026)

The same scanner, read for AI crawlers, search basics and speed across 4,026 new product sites.

Read the study

One address, 156 checks, each with its fix.

No account, just the address.