18% to 76% AI Visibility in 12 Weeks.
Agent Readiness.
How a regional legal-tech SaaS company went from invisible in AI-driven buyer research to being cited by all four major AI engines. Same product. Same market. New source of truth. Built by Usmart Technologies on the three-phase Agent Readiness framework.
Invisible to the Buyer's First Question
A regional legal-tech SaaS company serving solo and small-firm attorneys across a five-state region was watching inbound demo requests decline quarter over quarter. Their sales team knew the pipeline was drying, but the marketing team could not point to a channel that had broken. Google traffic was steady. Paid search was steady. Referral traffic was steady. And yet the actual demo requests were down 34% year-over-year.
The founder ran an informal test after reading a peer's LinkedIn post about "AI is eating the vendor discovery stage." He asked ChatGPT: "What are the three best legal practice management platforms for solo attorneys?" His company was not mentioned. He tried Claude. Same result. He tried Perplexity. Same result. Each engine confidently recommended two well-known incumbents plus one competitor his team had barely heard of, whose product was arguably worse but whose website was much easier for an AI to parse.
When he asked directly: "What does [their company] cost?", ChatGPT quoted a price 2.5x higher than their actual entry tier. Claude said the pricing was not publicly available. Perplexity linked to a five-year-old third-party review with outdated pricing. The buyer research funnel had moved to AI, and their company was either invisible or misrepresented at every checkpoint.
Pain Points
- ● Demo requests down 34% YoY despite steady traffic
- ● Not mentioned in any "best of" AI recommendations
- ● AI quoted their price 2.5x higher than reality
- ● Pricing hidden behind "Book a Demo" wall
- ● Feature comparisons buried in PDF sales decks
- ● Marketing copy heavy on "operational excellence" jargon
Same Prompt. Same AI. Twelve Weeks Apart.
Illustrative representation of the shift observed across the four major AI engines. Real client screenshots on file, anonymized here for confidentiality.
Three-Phase Deployment
- ✓ Phase 1 · Audit: 50 buyer-intent prompts across ChatGPT, Claude, Perplexity, Copilot
- ✓ Phase 2 · Fix: llms.txt at domain root, machine-parsable pricing table, 12 new use-case pages
- ✓ Full JSON-LD schema (Product, Offer, SoftwareApplication, FAQPage) on every commercial page
- ✓ Lightweight MCP endpoint exposing live pricing and integration availability
- ✓ Phase 3 · Retainer: Monthly rerun of the 50 prompts, quarterly source-of-truth refresh
Rebuild the Source of Truth the AI Actually Reads
Usmart ran the full three-phase Agent Readiness engagement. Phase 1 (Audit) produced a 22-slide visibility scorecard: 50 buyer prompts across four engines, with side-by-side screenshots showing every mention (or missing mention), every wrong price, and every competitor being recommended in the client's place. That deck was the moment the founder committed to the fix.
Phase 2 (Fix) took six weeks. We published an llms.txt file at their domain root that gave AI systems a curated summary and a directory of the canonical URLs to crawl. We rebuilt their pricing page as an HTML table with per-tier features (no more PDF sales decks, no more "Book a Demo to see pricing"). We wrote 12 new use-case pages, one per buyer profile (solo attorney, two-partner firm, litigation-only, transactional-only, family law specialist, and so on), each with a title that matched how buyers actually phrase the query.
We layered Product, Offer, SoftwareApplication, and FAQPage schema onto every commercial page. And because their pricing changes quarterly with promotional tiers, we deployed a lightweight Model Context Protocol (MCP) endpoint that lets AI agents query live pricing and integration availability directly, rather than caching stale numbers.
The 12-Week Visibility Curve
18% hit rate
9 of 50 target prompts mentioned the company. All 9 mentions came from Copilot (which pulls Bing). Zero on ChatGPT, Claude, Perplexity.
32% hit rate
llms.txt deployed. Pricing table live. Perplexity picked up changes first (Day 5). Copilot and ChatGPT followed by end of Week 3.
58% hit rate
12 use-case pages indexed. Claude began citing them for buyer-profile prompts. First AI-attributed demo request landed Week 5, Day 3.
76% hit rate
MCP endpoint live. All 4 engines citing consistently. 38 of 50 target prompts producing correct mentions with accurate pricing.
From Invisible to Consistently Recommended.
The company's actual product did not change. Their price did not change. Their sales team did not grow. The AI layer between buyer and vendor got fixed, and pipeline recovered accordingly. Demo request volume in the 90 days following the Week 12 milestone was up 47% versus the trailing 90-day baseline.
The Three Highest-Leverage Changes
Public Pricing Table
Removing the demo gate from pricing was the single biggest change. Within 10 days, all four engines were quoting accurate numbers. Every "book a demo to see pricing" model competitor lost this round.
Use-Case Pages
12 pages, each targeting a specific buyer profile with the language buyers actually use. This is what let AI systems answer "which platform is best for a family-law solo practice?" with the client's name.
MCP for Live Pricing
Their pricing changes quarterly with regional promos. The MCP endpoint means the AI always cites the current price, not a cached number from three quarters ago. This preserved the recovery through the next pricing cycle.
Their Product Was Never the Problem. Their Source of Truth Was.
This engagement did not change the client's product, pricing, positioning, or team. It changed one thing: the machine-readable representation of their business at the moment an AI agent asks about their category.
Twelve weeks earlier, that representation did not exist. AI engines were guessing (badly) from cached blog posts and outdated third-party reviews. Twelve weeks later, that representation lives at three places: their llms.txt file, their canonical use-case pages with schema, and their MCP endpoint for live queries. The AI now has an authoritative source to cite, and cites it.
Every SMB in a competitive category is running this same experiment, whether they know it or not. The only question is whether they are the vendor being recommended or the vendor being skipped.
See what AI is saying about your business.
Run the free chat-based Agent Readiness preview. Five minutes, no signup required. You will see your first visibility gaps in real time and get a personalised 360° report on your SEO / AEO / GEO visibility valued at $800, yours free.
Want the full framework playbook? Read Agent Readiness: How SMBs Get Recommended by ChatGPT, Claude, and AI Purchasing Agents.