Seoxpert.io
Data study · 2026-04-23

What 20 Real Homepages Tell Us About SEO in 2026

On 2026-04-23we ran a lightweight single-page audit on 20 widely-known technology and publishing homepages. This page lists every URL, the exact checks, the aggregate results, and the caveats. No extrapolation to "X% of the web" — we checked 20 hand-picked sites, and that's all the numbers describe.

Headline results

Successfully fetched
19%
19 of 20
On HTTPS
100%
Send HSTS
100%
HSTS preload
53%
Content-Security-Policy
79%
X-Frame-Options
74%
Referrer-Policy
68%
Canonical tag
74%
Meta description
84%
Desc. 120–160 chars
32%
In the commonly observed non-truncated range
JSON-LD structured data
37%
Exactly one H1
63%

What stands out

Security basics are near-universal in this sample — the differentiators are the modern headers

Every site was on HTTPS (100%) and every one that responded sent an HSTS header (100%). Where the sample split was on which modern security headers ship: 79% had a Content-Security-Policy, 74% set X-Frame-Options, 68% set Referrer-Policy, and only 53% had opted into the HSTS preload list. For consumer-facing tech companies, adopting CSP and preload is the gap. See the HSTS glossary · most common security issues.

Only 32% of meta descriptions are in the commonly-observed non-truncated range

84% of sites had a meta description at all, but only 32% landed between 120 and 160 characters. 16% were long enough to overflow the desktop SERP pixel width (> 920px). Most of what we saw was either too short — which lets Google build the snippet from page content instead — or too long to render fully. Check your own with the meta description checker.

Only 37% of homepages ship JSON-LD structured data

This was the biggest surprise. Rich-result-eligible structured data is a free click-through-rate lift, but only 37% of the sample ships any JSON-LD on the homepage at all. The most common schema type in this sample was Organization, followed by WebSite. See the structured data glossary.

H1 discipline is weaker than expected — only 63% had exactly one H1

16% of homepages had no H1 at all, and 21% had more than one. Even on sites run by teams that obviously care about SEO, the H1 convention gets lost in template refactors. See the H1 tag glossary.

Performance is fast by default — median homepage loaded in 288 ms

The fastest site returned full HTML in 18 ms; the slowest in 1033 ms. These are all CDN-backed, cache-optimised homepages — a useful baseline for what server-side response time looks like when done well. See the Core Web Vitals glossary · most common performance issues.

All numbers

MetricValue
Sites attempted20
Sites fetched successfully19 (95%)
HTTPS100%
HSTS header100%
HSTS preload directive53%
Content-Security-Policy79%
X-Frame-Options74%
Referrer-Policy68%
Has title tag100%
Title length (chars) — median46
Title length (chars) — range9 – 78
Has meta description84%
Meta description 120–160 chars32%
Meta description overflows desktop (920px)16%
Meta description length (chars) — median128
Meta description length (chars) — range64 – 190
Meta description pixels — median767
Has canonical tag74%
Canonical is self-referencing53%
Has viewport meta100%
Has <html lang>100%
Has og:title84%
Has og:image84%
Has JSON-LD structured data37%
Has favicon89%
Exactly one H163%
Zero H1s16%
Multiple H1s21%
Images without alt1%
Homepage load (ms) — median288 ms
Homepage load (ms) — range18 – 1033 ms

Per-site results

SiteStatusLoadTitle lenDesc lenH1sHSTSCSPCanonicalJSON-LD
github.com200176ms611904
stackoverflow.com200354ms334
news.ycombinator.com200595ms110
developer.mozilla.org20059ms121451
stripe.com200474ms541482
vercel.com200179ms671112
nextjs.org20093ms39641
tailwindcss.com20059ms761141
linear.app200485ms43641
notion.so2001033ms451241
figma.com200344ms461721
airbnb.com200855ms781321
shopify.com20087ms661451
slack.com200288ms491731
zapier.com200135ms471501
pinterest.com200355ms9710
medium.com40322ms160
reddit.com20018ms370
cloudflare.com200191ms511121
mozilla.org200952ms46861

Methodology and caveats

What we did

A single HTTP GET against each homepage from a server in our infrastructure on 2026-04-23. For each response we captured status code, response time, selected HTTP headers, and extracted HTML signals — title, meta description, canonical, H1 count, viewport, lang, Open Graph tags, JSON-LD types, and img alt coverage — using regex-based parsing of the HTML head.

What the sample is

Hand-picked set of widely-known technology, SaaS, and publishing homepages. The sites were chosen for recognisability, not as a representative sample of the web. 19 of 20 returned HTML successfully; the remaining site returned HTTP 403 to our user-agent.

What the sample is NOT

Statistically representative of "the internet" or any larger population. Percentages reflect this exact 20-URL sample. Run the same study on 20 mid-market e-commerce sites or 20 government sites and the numbers will look different.

Why a lightweight check

The full Seoxpert scan crawls multiple pages per site and covers 14 check categories across ~100 signals per page. For a one-day cross-site comparison the lightweight single-page approach is more useful — it isolates homepage-level signals and does not penalise sites for deep-crawl issues the homepage alone cannot reveal.

Site list

  1. https://github.com
  2. https://stackoverflow.com
  3. https://news.ycombinator.com
  4. https://developer.mozilla.org
  5. https://stripe.com
  6. https://vercel.com
  7. https://nextjs.org
  8. https://tailwindcss.com
  9. https://linear.app
  10. https://www.notion.so
  11. https://www.figma.com
  12. https://www.airbnb.com
  13. https://www.shopify.com
  14. https://slack.com
  15. https://zapier.com
  16. https://www.pinterest.com
  17. https://medium.com
  18. https://www.reddit.com
  19. https://www.cloudflare.com
  20. https://www.mozilla.org

Run a real scan against your own site — full crawl, 14 check categories, free.

Related: complete technical SEO audit guide · most common SEO issues