Original research · July 2026

The State of AI Readiness 2026

Somewhere right now, someone is asking an assistant a question your site should be the answer to. Whether you turn up in that answer comes down to one thing: can the machine reach your site, understand it, and quote it with confidence? We built a metric for that, the AI Pulse Score, and ran it across 13,838 live websites. This page is the method and the open data behind it, with nothing held back.

The full methodology — every signal, weight, calibration constant, validation check, and limitation — is specified in the accompanying white paper, The AI Pulse Score: A Reproducible, Open Metric for Website AI-Readiness (v1.0), written so that any party can recompute any number in this study. The paper is permanently archived on Zenodo: doi:10.5281/zenodo.21615311.

What the AI Pulse Score measures

Four things determine whether an AI answer engine can use your site. Each is scored 0–100; together they make the Pulse. The gap between the high pair (access, durability) and the low pair (structure, substance) is the whole story.

Nearly every site is on HTTPS and lets AI crawlers in — access is a solved problem. Where the web collapses is structure (schema markup, clean metadata) and substance (answerable, citable content). Those are exactly the signals a CMS migration drops on the floor if nobody is measuring — which is why the durability pillar exists.

Grades: 46.8% of the web is D or F

AI readiness by CMS

Every platform with at least 15 measurable sites, by median AI Pulse. The pattern is clear: commerce and site-builder platforms lead (they ship product and structured-data schema by default), while legacy open-source CMSs lag. WordPress tops the list at 80.9; Joomla sits lowest at 64.8.

AI readiness by industry

Every industry with a reportable sample, by median AI Pulse. Content- and commerce-driven sectors, which live on being found, lead; institutional sectors lag.

Method and limitations

• Each site was fetched once at the homepage with a single identified request; no crawling. 13,838 were measurable; 790 were bot-walled or unreachable and are excluded, not counted as failures.
• All four pillars are objective response checks, so every site is scored identically. A separate LLM-judged content-density read (median 32.4 answerable units per 1k tokens, n=5,847) is published in the dataset but deliberately not folded into the headline score.
• Homepage-only, so content-depth signals are a floor (interior pages typically score higher). CMS is fingerprint-based; unfingerprinted sites are grouped as “Unknown”.

Sources and related work

The research, standards and vendor documentation the AI Pulse Score builds on: the same reference list as the white paper, linked where a stable public URL exists.

1. Pew Research Center. 34% of U.S. adults have used ChatGPT, about double the share in 2023, June 2025. Why AI answer engines are becoming a first-stop channel.
2. Gartner, Inc.. Gartner Predicts Search Engine Volume Will Drop 25% by 2026, Due to AI Chatbots and Other Virtual Agents, press release, February 2024. (gartner.com newsroom.)
3. Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., Deshpande, A.. GEO: Generative Engine Optimization, KDD ’24 / arXiv:2311.09735. Peer-reviewed evidence that page structure changes what generative engines cite.
4. Lewis, P., et al.. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks, NeurIPS 2020 / arXiv:2005.11401. The retrieval architecture that makes machine-readable pages citable at answer time.
5. Knowles, M., et al. (Vercel and MERJ). The Rise of the AI Crawler, 2024–2025. Measured AI-crawler behavior at CDN scale — most AI crawlers do not execute JavaScript.
6. OpenAI. Overview of OpenAI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User).
7. Google. Google’s crawlers and fetchers (including Google-Extended).
8. Perplexity. Perplexity crawlers.
9. Howard, J. (Answer.AI). /llms.txt — a proposal to provide information to help LLMs use websites, 2024.
10. Koster, M., Illyes, G., Zeller, H., Sassman, L.. Robots Exclusion Protocol (RFC 9309), IETF, 2022.
11. Schema.org. Schema.org — shared structured-data vocabulary.
12. Le Pochat, V., et al.. Tranco: A Research-Oriented Top Sites Ranking Hardened Against Manipulation, NDSS 2019. The sampling frame for this study.

Cite this

Replatform Radar, The State of AI Readiness 2026, July 2026. replatformradar.com/research/ai-readiness-2026. Licensed CC BY 4.0 — the dataset and every number on this page may be reused with attribution. Archived on Zenodo: doi:10.5281/zenodo.21615305. Ready-to-paste formats:

Replatform Radar, "The State of AI Readiness 2026," July 2026, https://replatformradar.com/research/ai-readiness-2026. doi:10.5281/zenodo.21615305. Open dataset, CC BY 4.0.

{{cite web |author=Replatform Radar |title=The State of AI Readiness 2026 |date=2026-07-24 |url=https://replatformradar.com/research/ai-readiness-2026 |publisher=Replatform Radar |doi=10.5281/zenodo.21615305 |access-date=}}

@misc{replatformradar2026aipulse,
  author       = {{Replatform Radar}},
  title        = {The State of AI Readiness 2026},
  year         = {2026},
  doi          = {10.5281/zenodo.21615305},
  howpublished = {\url{https://replatformradar.com/research/ai-readiness-2026}},
  note         = {Open dataset, licensed CC BY 4.0}
}