Two public profiles · no score or winner
mindbox.ru and pnas.org
Measured takeaway
mindbox.ru allows model-training crawlers; pnas.org blocks model-training crawlers.
mindbox.ru allows GPTBot; pnas.org blocks GPTBot.
Measured in Botfy's 2026-09-07 publication.At a glance
A difference appears only when both sites have comparable evidence. Unavailable or unresolved observations remain incomplete.
| Evidence | mindbox.ru | pnas.org |
|---|---|---|
| robots.txt | Served and parsed | Served and parsed |
| Google crawl posture | Open | Open |
| Unnamed crawler fallback | Partly Restricted they inherit the wildcard (*) group's rules | Partly Restricted they inherit the wildcard (*) group's rules |
| Base /llms.txt | Published | Not probed |
Treatment by purpose
Only resolved blocked, allowed, or mixed states count as a policy agreement or difference. Unnamed agents still inherit each site's wildcard rules.
| Declared purpose | mindbox.ru | pnas.org |
|---|---|---|
| Model training | Allowed | Blocked |
| AI search | Allowed | Not explicitly addressed |
| Live assistant fetch | Allowed | Not explicitly addressed |
| Opt-out directives | Allowed | Not explicitly addressed |
| Ambiguous / undocumented | Allowed | Not explicitly addressed |
Named agents
Measured differences appear first. “Not explicitly addressed” is not the same as allowed; the site's wildcard fallback governs that crawler.
| Agent | mindbox.ru | pnas.org |
|---|---|---|
| GPTBot OpenAI | Allowed | Blocked |
| ClaudeBot Anthropic | Allowed | Not explicitly addressed |
| CCBot Common Crawl | Allowed | Not explicitly addressed |
| Bytespider ByteDance | Allowed | Not explicitly addressed |
| Meta-ExternalAgent Meta | Allowed | Not explicitly addressed |
| FacebookBot Meta | Allowed | Not explicitly addressed |
| cohere-ai Cohere | Allowed | Not explicitly addressed |
| cohere-training-data-crawler Cohere | Allowed | Not explicitly addressed |
| AI2Bot Allen Institute | Allowed | Not explicitly addressed |
| Kangaroo Bot Kangaroo LLM | Allowed | Not explicitly addressed |
| OAI-SearchBot OpenAI | Allowed | Not explicitly addressed |
| Claude-SearchBot Anthropic | Allowed | Not explicitly addressed |
14 additional tracked agents
| Agent | mindbox.ru | pnas.org |
|---|---|---|
| Applebot | Allowed | Not explicitly addressed |
| PerplexityBot | Allowed | Not explicitly addressed |
| YouBot | Allowed | Not explicitly addressed |
| ChatGPT-User | Allowed | Not explicitly addressed |
| Claude-User | Allowed | Not explicitly addressed |
| Claude-Web | Allowed | Not explicitly addressed |
| Perplexity-User | Allowed | Not explicitly addressed |
| Amazonbot | Allowed | Not explicitly addressed |
| Meta-ExternalFetcher | Allowed | Not explicitly addressed |
| MistralAI-User | Allowed | Not explicitly addressed |
| DuckAssistBot | Allowed | Not explicitly addressed |
| Google-Extended | Allowed | Not explicitly addressed |
| Applebot-Extended | Allowed | Not explicitly addressed |
| anthropic-ai | Allowed | Not explicitly addressed |
How each site compares with its own rank peers
The sites may belong to different rank bands. Each percentage keeps its own population and denominator; Botfy does not subtract rates from unlike cohorts.
| Trait | mindbox.ru | pnas.org |
|---|---|---|
| Wildcard group | Yes 94.1% Ranks 1,001–10,000 · n=5,437 | Yes 94.1% Ranks 1,001–10,000 · n=5,437 |
| Effectively open file | No 16.4% Ranks 1,001–10,000 · n=5,437 | No 16.4% Ranks 1,001–10,000 · n=5,437 |
| Sitemap declared | Yes 65.6% Ranks 1,001–10,000 · n=5,437 | Yes 65.6% Ranks 1,001–10,000 · n=5,437 |
| robots.txt served | Yes 60.7% Ranks 1,001–10,000 · n=9,037 | Yes 60.7% Ranks 1,001–10,000 · n=9,037 |
| robots.txt returned 404 | No 5.4% Ranks 1,001–10,000 · n=9,037 | No 5.4% Ranks 1,001–10,000 · n=9,037 |
| Over the RFC parse limit | No 0.2% Ranks 1,001–10,000 · n=5,437 | No 0.2% Ranks 1,001–10,000 · n=5,437 |
| Blanket disallow | No 3.7% Ranks 1,001–10,000 · n=5,437 | No 3.7% Ranks 1,001–10,000 · n=5,437 |
How this comparison is measured
Both sites use Botfy's latest published observations. A missing fetch, unresolved agent verdict, and measured absence are different states. Peer percentages use each site's exclusive Tranco rank band and print their own denominator. Populations, sources, and limits →