We Scored 6 of the Web's Most Authoritative Sites — GEO Averaged Just 41/100

It is easy to assume a low score means your site is unfinished. So I ran some of the best-known technical sites on the web through the same tool — and they all came back low in exactly one place.

This post shares what happened when I scored six authoritative sites with IndexReady's own analyzers. Every number below is a real measurement I ran for this article, not an estimate.

How the test was run

The six sites are ones most developers have visited at least once. I scored a single page from each — the homepage in most cases, and the "Search engine optimization" article for Wikipedia — across IndexReady's 18 SEO checks and 14 GEO checks.

Scores include live PageSpeed Insights data and are out of 100 for SEO and 100 for GEO. Everything reflects the pages as they were on June 1, 2026.

The result: high SEO, low GEO

Site	SEO	GEO
Digital Agency (Japan)	85	39
MDN Web Docs	81	37
Next.js	77	28
web.dev	66	44
Wikipedia (JA)	66	55
GitHub	63	45
Average	73.0	41.3

SEO averaged 73. The spread runs from 63 to 85, but these are passing grades — fair enough for sites that have spent years building search traffic.

GEO is the problem. The average was 41. Even the top scorer, Wikipedia, only reached 55, and Next.js landed at 28. The gap between SEO and GEO was over 20 points on every single site. web.dev and Japan's Digital Agency — both run by organizations that live and breathe the web — still scored around 40 on GEO.

This is not the tool being harsh. The same analyzers gave these sites a 73 on SEO, so the scale is not dragging everything down. SEO has had attention; GEO has not. That is the simplest reading.

Finding 1: almost nobody ships llms.txt — not even the leaders

The single biggest GEO drain was llms.txt. None of the six scored well on it. MDN, web.dev, Wikipedia and the Digital Agency scored zero. Next.js and GitHub had no llms.txt either and only earned partial credit from AI-crawler rules in robots.txt.

llms.txt is a proposed file for telling AI systems about your site, but Google has stated plainly that it does not currently treat it as a ranking or citation signal. This measurement quietly confirms that from the other side: if the most respected technical sites on the web have not adopted it, there is no urgency to chase it — and certainly no basis in this data for making it your top priority.

Finding 2: the GEO gap was really a structured-data gap

What separated the GEO winners from the losers was structured data (JSON-LD).

The top scorer, Wikipedia (GEO 55), earned 5 of 10 on structured data and 6 of 8 on schema completeness. As an encyclopedia, it leads with clear definitions, a deep section structure, and citation links — exactly the signals AI engines look for.

The Next.js homepage, by contrast, scored zero on structured data and zero on schema completeness. It is a design-forward landing page with almost no JSON-LD, and its last-place GEO of 28 came mostly from that. MDN and the Digital Agency also scored zero on structured data — rich documentation that simply was not declared in a machine-readable form.

AI search reads structure, not appearance. A page that is perfect for humans is still hard for an AI to classify without Article or BreadcrumbList markup. These six sites mapped that divide cleanly.

Finding 3: citability and breadcrumbs were near-zero everywhere

Two of the newer checks showed an even starker pattern.

• Passage citability (whether the page has self-contained passages of roughly 134-167 words that an AI can quote) scored 2 out of 6 on all six sites. Homepages tend to be short taglines and bullet lists, with few standalone "answer paragraphs."
• Breadcrumb schema also scored 1 out of 4 on every site. Almost none of the homepages declared an explicit BreadcrumbList.

Some of that is the nature of a homepage. But the fact that most of these sites offer no self-contained passage an AI can lift as a cited answer is telling.

Finding 4: even the giants have basic SEO holes

GEO stole the spotlight, but SEO had shared gaps too.

Image optimization missed full marks almost everywhere — missing width/height, no WebP/AVIF — with even the Digital Agency and web.dev scoring 1 of 5. Hreflang was rarely declared, even on multilingual sites, landing at 2-3 points. GitHub scored zero on sitemap.xml; Next.js scored zero on canonical. Surprising basics, missing on huge sites.

The flip side: none of these leaders score full marks either. A score is not a game to max out — it is a map of which holes your own site has.

The specific holes, site by site

Averages hide the details, so here are the individual ones. Each is a check that actually lost points in this run.

• Digital Agency (SEO 85, top score) is polished, but scored 5 of 6 on HTTPS. HTTPS itself is fine; some security headers such as HSTS were missing. Worth tightening for a government site.
• MDN Web Docs (SEO 81) scored zero on OGP tags for the homepage I tested. The og:title / og:image tags that control how a shared link looks were simply not on that page. Even a world-class reference can miss basic tags on some pages.
• Next.js (GEO 28, last place) scored zero on both canonical and robots.txt references. It is a demo-forward landing page, and combined with no structured data (also zero), it reads to AI search as "a page whose content is hard to extract."
• GitHub (SEO 63, lowest) had no detectable sitemap.xml from the homepage (zero) and scored 2 of 5 on heading structure. As an app landing page, function comes first and document structure is thin.
• Wikipedia JA (GEO 55, top) surprisingly scored zero on meta description. It still led on GEO because its structured data, schema completeness, and citation links carry the page. A hole in one check can be more than offset by a strength elsewhere.

No authoritative site nails every check. Each one puts effort where its goals demand and skips the rest — and a score makes that allocation visible.

How to close the GEO gap

For nearly every site, the data pointed at the same two things: structured data and citability. Both are fixable today, without special tooling.

Start with structured data. On an article page, simply declaring Article already helps AI understand who wrote what, when, and about what:

<script type="application/ld+json">
  {
    "@context": "https://schema.org",
    "@type": "Article",
    "headline": "Article title",
    "author": { "@type": "Person", "name": "Author name" },
    "datePublished": "2026-06-01",
    "image": "https://example.com/og.png"
  }
</script>

Adding a BreadcrumbList to show the page's place in the hierarchy also fills the breadcrumb check that almost every site here missed.

Then citability. Open each section with a conclusion that stands on its own, without surrounding context: define the thing in one sentence ("X is..."), then back it up. That self-contained passage of roughly 134-167 words is the unit AI quotes in an answer. A short homepage tagline cannot fill that role.

The upside: these are exactly the areas most sites are deferring. Once your SEO fundamentals are in place, investing in structured data and citability lets you stand out while the field is still thin.

What this data says

Three things stand out from the six measurements:

1. SEO is mature; GEO is untouched. Even sites with years of search success have clearly deprioritized AI-search optimization.
2. The GEO gap is a structured-data gap. Whether content is declared in a machine-readable form mattered more than how rich it was.
3. llms.txt can wait. The leaders have not adopted it and Google does not weight it.

Turn that around and it is an opportunity: invest in structured data and citability now, and you compete in a space most sites have not touched yet.

Caveats

In fairness, the limits of these scores:

• I scored one page per site, not the whole site. Homepages skew low on structured data; article pages would score differently.
• The scores are a reference based on IndexReady's own criteria, not a guarantee of rankings or AI citations.
• Every site has a different goal. A demo-first landing page skipping JSON-LD is not necessarily "wrong."

Even so, line them up with one ruler and the pattern appears. Where would your site land on the same ruler? Drop in a single URL and see how wide your own SEO-to-GEO gap really is.