Fenrock sells AI to banks, but its robots.txt disallows the AI agents it lists under "Common AI / LLM crawlers". That list includes the agents that fetch a page live when a person asks an assistant a question, not only training crawlers. Even a reader that is allowed in gets almost nothing: without running JavaScript, the homepage returns only its title. The headline, product copy and "Request a Demo" button are all drawn by scripts. When a BSA officer or bank COO asks an assistant which tools help a community bank clear an alert backlog, the assistant can't read Fenrock's own description. It falls back on third-party write-ups, or on rivals like Hummingbird and Bretton that publish full pages. This is not a speed problem: real visitors pass Core Web Vitals and the SEO score is 100. It is a readability problem for machines.
First action
Move the user-triggered assistant agents to Allow in robots.txt and redeploy.
How to measure it
Assistant citations on five buyer questions
- Now
- –
- Target
- 3 of 5
- By
- 90 days
Steps
Set the policy
Decide which agents to allow: user-triggered fetchers yes, training crawlers a separate call.
CEO 1 hour
Edit robots.txt
Move the chosen agents to Allow and keep the search-engine rules as they are.
CTO 30 minutes
Prerender the homepage
Static HTML at build time, with the same markup hydrated afterwards.
Founding Engineer (Full stack) 3 days
Check the raw HTML
Fetch the page without JavaScript and confirm the headline, use cases and demo link are there.
Founding Engineer (Full stack) 1 hour
Add structured data
Organization and SoftwareApplication markup using the same wording as the page.
Founding Engineer (Full stack) 1 day
Baseline five questions
Ask the major assistants five buyer questions and log what they say about Fenrock.
CEO 1 hour
Re-check monthly
Log citations and errors, and fix the page wherever answers are wrong.
CEO monthly
Why it works
Make two changes. First, in robots.txt, move the user-triggered assistant agents (the ones that fetch a page because a person asked) from Disallow to Allow. Whether training crawlers stay blocked can be a separate decision. Second, prerender the homepage at build time on Vercel so the initial HTML holds the headline, the four use cases (compliance, fraud investigations, loan operations, customer operations), "the analyst decides", audit logs, SOC 2 and a working demo link. Add Organization structured data. Then write five questions a bank buyer would ask an assistant, record today's answers, and re-check monthly whether Fenrock is cited and described correctly.
Bank leaders research vendors before taking a call, and a two-person company can't be in every room. Being readable wherever buyers ask questions is cheap reach. The same prerendered HTML also gives proper link previews when a link is forwarded inside a bank by email, Teams or Slack.
Ready to use
ChecklistReadable-site checklist
[ ] robots.txt: user-triggered assistant agents under Allow [ ] Training-crawler policy decided and documented [ ] Homepage prerendered: headline, four use cases, "the analyst decides", audit logs, SOC 2 [ ] Demo link present as a plain HTML link [ ] Organization structured data [ ] Page checked without JavaScript [ ] Five buyer questions baselined, with a monthly re-check
Risks and how to handle them
Concern about content being used for training
Allow only the user-triggered agents and keep training crawlers blocked.
Prerendering breaks the current animations
Hydrate the same markup and test on mobile before shipping.
Evidence
- Your websiterobots.txt (read November 10, 2026) allows Googlebot, Bingbot, DuckDuckBot, Twitterbot and facebookexternalhit, but under "Common AI / LLM crawlers" disallows GPTBot, ChatGPT-User, CCBot, ClaudeBot, AnthropicBot, Applebot-Extended, PerplexityBot, Bytespider, Diffbot and YouBot.Source
- Your websiteA reader that does not run JavaScript (fetched at www.fenrock.ai and fenrock.ai on November 10, 2026) receives only the page title; the headline, product copy and "Request a Demo" button are drawn by scripts, about 1.1 MB of which load from a Google CDN.Source
- PageSpeedGoogle PageSpeed (November 10, 2026): mobile performance 59, desktop 88; in the mobile lab test the first and largest content both appear at 6.8 s, the largest element is the "Request a Demo" button, and 430 KB of JavaScript is unused (about 3.9 s of potential savings). Real Chrome visitors pass Core Web Vitals: LCP 1.5 s, INP 32 ms, CLS 0 at the 75th percentile. SEO score 100; accessibility 96 with one colour-contrast issue.Source
- Competitor researchHummingbird homepage: "AI-powered compliance" for banks, fintechs and crypto firms; screening, transaction monitoring, investigations and SAR/CTR filing; named customer logos and testimonials from BSA leaders; a trust center, security page, webinars, a compliance AI field guide and a "Take a tour" product tour.Source