Skip to content
— STUDY · EPISODE 4

How many Romanian sites publish llms.txt — and does it matter?

Of the 87 Romanian domains that AI engines actually cite when answering Romanian-language questions, 56 publish an llms.txt file (64.4%), measured on 2026-08-09. The file is already the majority practice in this niche. It does not appear to be what earns the citations, though: the most-cited domain in the sample carries a plugin-generated llms.txt with none of the summary line the specification asks for.

· Dan Cristian Alexandrescu, Websem · ORCID

— THE NUMBERS
Every percentage is computed against the 87 domains in the sampling frame, measured on 2026-08-09, in a single run.
Domains in frame87
Publish llms.txt56 (64.4%)
— plugin-generated12 · mean 25,070 B
— hand-written44 · mean 8,803 B
Spec-conformant (H1 + summary)39
With no curated link8
Block answer-engine crawlers2
Block training scrapers only1
— WHAT THE DATA SHOWS

1. llms.txt is already the majority, against the common impression

56 of 87 domains serve one. Public discussion still treats llms.txt as a curiosity; in the niche the engines actually cite, it is the dominant practice.

2. The large files are plugin dumps, not curation

The 12 plugin-generated files average 25,070 bytes against 8,803 for the hand-written ones — nearly three times larger, with no H1 and none of the summary line the specification asks for. The largest in the sample, seo365.ro, runs to 68,722 bytes.

3. A site can cancel its own visibility from robots.txt

The distinction most often confused: blocking CCBot or Bytespider is a rights decision that does not remove you from answers. Blocking GPTBot, ClaudeBot or Google-Extended does, whatever you write in llms.txt.

  • chatbotsite.roGPTBot, anthropic-ai · publishes llms.txt anyway
  • canopy.roClaudeBot, Google-Extended, GPTBot
— METHODOLOGY

The sampling frame. Not domains we picked, but the domains AI engines actually cite when answering Romanian-language questions about marketing, AEO and automation. The list comes from the sources cited in the answers monitored in LLM Pulse (project 3039, 25 prompts, 5 surfaces), filtered to Romanian-market domains. Inclusion rule: the domain appears among the 200 most-cited sources. That yields 87 domains.

The measurement. One GET to /llms.txt and one to /robots.txt per domain, on 2026-08-09, in a single complete run. A file counts as served if it answers 200, has content-type text/plain and exceeds 80 bytes — the threshold removes error pages returned with a 200 status. Conformance to the llmstxt.org specification is checked on two objective elements: the first line is a Markdown H1, and a blockquote carrying the summary appears within the first eight lines.

The two crawler categories. An agent counts as blocked only if robots.txt carries a block dedicated to it with "Disallow: /". Answer-engine crawlers: GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, ClaudeBot, Claude-User, anthropic-ai, Google-Extended. Training or archive scrapers: CCBot, Bytespider, Applebot-Extended, Amazonbot, Meta-ExternalAgent, cohere-ai. The distinction is material: the first category decides whether you appear in answers, the second does not.

What the study does not measure. It does not establish a causal relation between llms.txt and citations. It shows a suggestive negative correlation — the most-cited domain has the least conformant file — but with 87 domains and a single run, the effect of llms.txt cannot be isolated from domain age, authority or content. Nor does it measure whether the bots read the file at all: that would need server logs, which we do not have for third-party domains.

Reproducibility. The scan is a script, and the figures on this page are generated from the raw data rather than typed: scripts/scan-llmstxt.py and scripts/build-llmstxt-study.mjs. The raw data is below, under CC BY 4.0.

— THE DATA, ROW BY ROW
All 87 domains in the frame, sorted by file size. "—" means the domain serves no llms.txt.
DomainBytesSourceSpecLinksBlocked bots
seo365.ro68,722handyes128
smarters.ro61,922pluginno275
marketinglab.ro60,891pluginno180
limitless.ro48,212pluginno164
marketos.ro44,586pluginno123
snsys.ro38,956handyes149
targetweb.ro32,376pluginno83Bytespider
apexdigital.ro31,929handyes64
websem.ro29,064handyes74
vivinet.ro25,418pluginno108
dafe.ro21,312handyes51
nextchapter.ro20,120handyes92
digitalrocket.ro17,799handyes65
dmaster.ro12,345handyes59
ewdigital.ro11,453pluginno44
privatebrands.ro10,414handyes63
mda-digital.com9,779handno0
aifactory.ro8,965handno0
seocherry.ro8,652pluginno38
sanbi.ai8,596handyes51
goai.ro8,259handyes61
automatizaricuai.ro6,197handyes17
growwwise.com5,902handyes39
difrnt.ro4,981handyes29
creativdigital.ro4,756handyes32
digitaliomarketing.ro4,432handyes18
natix.ro4,211handyes21
seozilla.ro4,168handyes0
code24.ro4,018handyes19
aysa.ro3,983handno0
klain.ro3,773handyes40
optimizareai.online3,541handyes24
spacemedia.ro3,478handyes26
marketiu.ro3,385handyes14
groway.ro3,362handyes3
sphinx-it.ro3,360handyes3
themarkers.ro3,283handyes32
nion.ro3,264handyes3
geoflux.ai3,161handyes2
chatbotsite.ro3,126handyes0GPTBot CCBot anthropic-ai Bytespider cohere-ai Meta-ExternalAgent
seo-ai.ro2,910pluginno35
digitalizare.ai2,893handno0
automatizez.ro2,601handyes15
dancovision.ro2,596handyes8
aiautomatizari.ro2,595handyes13
bluedotfusion.ro2,532handyes15
arobsgrup.ro2,472pluginno15
outglow.ro2,241handyes6
digitalcraft.ro2,004handyes16
agentiadeai.ro1,821handyes11
metaflow.ro1,631handyes8
brandscan.ro1,486handyes0
agentiegeo.ro1,316handno0
mahucode.ro1,195pluginno13
nexaiagency.ro956handyes9
fgd.ro749pluginno5
aiengineoptim.ro
upswing.ro
iagency.ro
iqads.ro
blogdigital.ro
aifrontdesk.ro
agentiiseo.ro
agentii-seo.ro
roweb.ro
dwf.ro
adsem.ro
pionmedia.ro
remotemarketing.ro
mindloop.ro
arcsoft.ro
apolloboostmedia.ro
gregoire.ro
ksd.ro
instapress.ro
sostenia.ro
start-seo.ro
seoagency.ro
flowmatic.ro
donamarketing.ro
daredigital.ro
autoads.ro
aientity.ro
cosmicweb.ro
ai4business.ro
canopy.roAmazonbot Applebot-Extended Bytespider CCBot ClaudeBot Google-Extended GPTBot Meta-ExternalAgent
gomag.ro