Clayground built a channel for AI assistants on purpose: robots.txt lists llms.txt next to the sitemap under "AI discoverability", and there is an /llm-info page with a comparison matrix written for agents. But those files describe an older company. They name two services, not three: the word Tricle never appears, although Tricle AI is the one offer with a self-serve signup, 50 free account actions and a 30-day guarantee. They link five of the nine case studies and leave out Sourcera (125+ HubSpot deals in 90 days), vL&p (54 HubSpot deals in two months) and TalSource, the three newest and the ones with the hardest numbers. They give HeyJobs "25+ new clients per month", a figure the live case study no longer shows. And their booking link is a different Calendly address from the one behind every button on the site. A company that sells AI sales agents is briefing AI assistants with last summer's pitch.
First action
Open llms.txt next to the case studies index and list every offer, case study and link that is missing or different.
How to measure it
Share of 10 test questions answered with current facts
- Now
- –
- Target
- 8 of 10
- By
- 60 days
Steps
Diff the files against the site
Compare llms.txt, llms-full.txt and /llm-info with the sitemap, the three service pages and the nine case-study cards. Note every missing offer, case study, number and link.
Forward Deployed Engineer & Customer Success 2 hours
Add Tricle AI
A section with the product description, the six agent types, the 50 free account actions, the 30-day guarantee and the signup and login links.
Co-Founder & CTO 2 hours
List all nine case studies
One line each with the headline number from the card, the industry, the timeline and the page link. Drop figures the live pages no longer show.
CCO 2 hours
Fix the links
Use the service addresses from the sitemap and the same booking link as the site's buttons. Add a last-updated date.
Co-Founder & CTO 1 hour
Baseline the answers
Ask the main assistants ten buyer questions (for example: who cleans a Dynamics CRM with Clay, which tools reactivate dormant CRM contacts, GTM engineering firms in Switzerland) and save the answers.
GTM Consultant half a day
Tie it to publishing
Add a line to the case-study and service-page checklist: update the three AI files the same day.
Co-Founder & COO 30 minutes
Re-test monthly
Repeat the ten questions each month and track which answers name Tricle AI or a current case study.
GTM Consultant monthly
Why it works
Rewrite llms.txt, llms-full.txt and /llm-info from the live site in one sitting. List all three offers with the sitemap's addresses, give Tricle AI its own section (what it does, the six agent types, the free allowance, the guarantee, the signup link), list all nine case studies with the same headline number that appears on each card, and use the booking link the site uses. Add a last-updated date. Then make the files part of the publishing routine: every new case study or offer change updates them the same day. Finally, ask the main assistants the ten questions a buyer would ask and record what they say before and after.
Whichever assistants do read these files can only repeat what is in them. Today the file cannot lead anyone to Tricle AI, because it does not say Tricle exists, and it undersells the proof by leaving out the strongest results. The fix costs an afternoon, needs no design work, and makes every later case study pay off twice: once for human visitors and once for machines.
Ready to use
ChecklistAI files refresh list
[ ] Three offers, each with the address from the sitemap [ ] Tricle AI section: what it does, six agent types, 50 free account actions, 30-day guarantee, signup link [ ] All nine case studies, one line each, with the card's headline number [ ] Remove figures the live case studies no longer show [ ] Same booking link as the site's buttons [ ] Last-updated date at the top [ ] llms.txt, llms-full.txt and /llm-info say the same thing [ ] Ten buyer questions saved as a baseline [ ] Publishing checklist updated
Risks and how to handle them
Not every assistant reads llms.txt
Mirror the same facts on /llm-info and in the visible pages, which all crawlers read.
The files drift again
Make the update a line on the publishing checklist and date the file.
Evidence
- Your websitellms.txt, which robots.txt lists next to the sitemap under "AI discoverability" (checked twice on November 18, 2026): names two services ("AI Pipeline Generation" and "Revenue Operations & CRM Cleanup") and links five case studies (Acathia, Silvernova, HeyJobs, Archlet, Idun). The words Tricle, Sourcera, TalSource and vL&p do not appear. It gives HeyJobs "25+ new clients per month", links service addresses that are not the ones in the sitemap, and its booking link is a different Calendly address from the one on the site's buttons.Source
- Your website/llm-info and /llms-full.txt: an "AI-Optimized Documentation" page with a "Comparison Matrix for AI Agents" and "Contact & Affordances for Agents"; two modules (pipeline generation; RevOps and CRM agents); the same five case studies; no mention of Tricle; no last-updated date. The sitemap dates /llm-info July 15, 2026, before the Sourcera, TalSource and vL&p case studies were published (September and October 2026).Source
- Your websiteCase studies index: nine cards with headline results: Sourcera AI "125+ deals generated in 90 days", Schranner "50k CRM records cleaned & enriched", HeyJobs/Madita "AI-powered GTM validation", Silvernova "5× ARR in 10 months", Acathia "Signal-driven deal sourcing", TalSource "One high-value customer won", IDUN "18+ sales calls per month from CRM revival", Archlet "2.5× BDR productivity", vL&p "60 replies & 54 HubSpot deals in 2 months".Source
- Your websiteTricle AI page: H1 "Break into accounts when they become winnable."; six agent types (Break Into Target Accounts, Reactivate Dormant CRM, Champion Job Change, Stalled Deal Rescue, Nurture CRM & Core Evangelists, Create Your Own Playbook); "We'll find qualified pipeline in your target accounts within 30 days, or we keep working for free until we do."; a "Get 50 Free Account Actions" signup button; "50k+ signals processed daily" and a "10 min/day" rep workflow; no prices.Source