NOXEL Research Lab

NOXEL Research Lab publishes original data on SEO, AEO, GEO, and AI search visibility — drawn from real website audits, not recycled advice.

NOXEL Research Lab has published its first preliminary findings — a v1 dataset of 53 publicly accessible websites, analysed for AI crawler accessibility and structured data adoption. See the findings below, or read our full methodology. This is a starting point, not a final word — we'll expand the sample and republish as more data comes in.

What this is

Most SEO content restates the same advice found everywhere else. NOXEL Research Lab is built to do the opposite: publish findings that come directly from real audits run through NOXEL SEO, with a transparent methodology behind every number.

Every statistic we publish here will include a precise figure, a sample size, a methodology, a date, and a source — no invented numbers, no vague claims.

Explore

Available now

Methodology

How NOXEL calculates SEO, AEO, GEO, and Core Web Vitals scores — the real weights and criteria.

Available now

AI Search Readiness — v1

First findings on AI crawler accessibility and structured data adoption, n=53.

Live · expanding

In progress

Benchmarks

Industry and category benchmarks for technical SEO, performance, and AI search readiness.

Building sample size

NOXEL AI Search Readiness — v1 Findings

Sample: 53 publicly accessible websites, curated across marketing/SEO agencies, SaaS tools, and a small institutional control group (government, academic, and reference sites). This is a working v1 dataset, not a random or representative sample of the web — we're publishing it now because a transparent, modest-sized study beats waiting indefinitely for a "perfect" one. Individual domains are never named; every figure below is an aggregate.

AI crawler accessibility

20.8% of sites blocked at least one major AI crawler (GPTBot, PerplexityBot, and similar).
Of those blocks, 91% were attributable to explicit robots.txt rules — the remainder had no clear cause from robots.txt alone and would need server-level investigation.

Structured data adoption

28.3% of sites had no JSON-LD structured data at all.
66.0% had no Organisation or LocalBusiness schema.
94.3% had no BreadcrumbList schema.

In short: our research found that structured data adoption — especially BreadcrumbList schema — remains rare even amongst technically sophisticated sites, and AI crawler blocking usually comes down to robots.txt, not firewalls.

AI crawler accessibility refers to whether crawlers like GPTBot and PerplexityBot can successfully fetch and read a page — a prerequisite for that page ever being cited by an AI answer engine.

How we conducted this research

First, we selected 53 publicly accessible sites across three categories: marketing/SEO agencies, SaaS tools, and institutional reference sites.
Then, we ran each domain through NOXEL SEO's own audit engine — the same technology behind every NOXEL SEO report — checking AI crawler accessibility and structured data presence.
Next, we aggregated the results into the percentages above; no individual domain or score is published.
Finally, we plan to grow this dataset towards a larger, more representative sample and will republish updated findings as that happens.

Our data was scanned in September 2026. Every figure above states its sample size plainly, as promised.

Why original research matters for AI search

Large language models like ChatGPT, Claude, Perplexity, and Gemini don't rank pages the way traditional search engines do — they synthesise answers, and they favour sources that offer something a dozen other pages don't already say. Recycled advice, reworded definitions, and generic "10 tips" listicles rarely get cited, because an AI model has already absorbed that information from thousands of other sources during training. Original data — a real sample, a real methodology, real percentages — is different. It's the kind of content an AI model can't already know, which is exactly what makes it worth quoting.

This is the core idea behind NOXEL Research Lab: instead of writing more generic SEO advice, we run real audits through NOXEL SEO's own engine and publish what we actually find. Every study here follows the same discipline — a disclosed sample size, a described methodology, a publication date, and aggregate-only results with no individual site named.

What's coming next

The v1 dataset above is deliberately modest — 53 sites is enough to publish honestly, not enough to call representative of the wider web. Our next release will grow the sample towards a larger, more stratified dataset (more categories, more sites per category, a clearer sampling methodology), and will add new signals beyond AI crawler accessibility and structured data — things like content freshness, entity clarity, and topical depth across the same sites.

If you'd like to be notified when the next NOXEL Research Lab release goes live, or if you run a site and want it included in a future study (with your explicit consent — see our data policy above), reach out via our contact page.