// Quarterly Report · Published 2026-05-28 · By Grow Research
State of the Agent-Readable WebQ2 2026
An open quarterly measurement of how reliably the AI industry's own marketing sites can be cited by ChatGPT, Perplexity, Claude, and Google AI Overviews. Across 390 tracked AI companies, only 13% clear the threshold above which AI engines cite by name.
Six headline findings
- 63% of 390 AI companies are missing or under-serving llms.txt.
- 67% ship insufficient JSON-LD for reliable AI citation.
- 13% clear the agent-native bar (score ≥ 85).
- 27% score below 55 — effectively opaque to ChatGPT, Perplexity, and Claude.
- 66% fail the first-byte speed threshold AI crawlers timeout against.
- Dev Tools leads at 72/100 average; Agents trails at 57/100.
By category
Dev Tools leads agent-readability at 72/100; Agents trails at 57/100 — a 15-point gap inside one industry.
// Infra
69/100
avg across 82 sites
// Models
65/100
avg across 87 sites
// Agents
57/100
avg across 103 sites
// Dev Tools
72/100
avg across 118 sites
Top 5 and bottom 5
// Agent-native
- Framer AI98/100
- JetBrains AI95/100
- Anthropic94/100
- Fly.io94/100
- Grafana94/100
// Effectively opaque
- Tennr33/100
- Abridge36/100
- Covariant37/100
- Genesis Therapeutics37/100
- Monica37/100
What changed since Q1
- • llms.txt adoption remains the single largest gap — 63% of the dataset still ships none or one too thin to serve as inference context.
- • JSON-LD coverage (67% weak) is the second-biggest leak; most failing sites ship Organization and stop there.
- • Speed failures concentrate in JS-rendered marketing pages with TTFB above the 800ms wall AI crawlers won't wait past — 66% of the dataset.
- • Citability — the front-loaded answer pattern AI engines extract from — is still rare even among the top 5.
Implications for buyers
When 27% of the AI industry's own marketing sites are below the threshold AI engines will cite, the upside is asymmetric. The first competitor in a category to ship clean HTML, valid JSON-LD, and a real llms.txt becomes the default answer to category queries for the entire quarter — until others catch up.
None of the failing rows have a content problem. They have a reachability problem. Fixable in a single sprint.
Cite this report
APA
grow.contact (2026). State of the Agent-Readable Web — Q2 2026. Retrieved from https://grow.contact/report/q2-2026
BibTeX
@techreport{grow_aiweb_q2_2026,
author = {{grow.contact}},
title = {State of the Agent-Readable Web --- Q2 2026},
institution = {grow.contact},
year = {2026},
month = {May},
url = {https://grow.contact/report/q2-2026},
note = {CC BY 4.0}
}Pull quote
"63% of 390 top AI companies ship no usable llms.txt — and 27% score below the threshold AI engines will cite by name." — grow.contact, Q2 2026 report