Case Study 08 / GTM Strategy
Finding out why AI assistants couldn't see an industry specialist's website
An audit of how ChatGPT, Perplexity and Claude see the company's website found that every one of its 28 pages served AI crawlers the same 3.5 KB shell: one headline and no body text. The site scored 32/100 for AI-search readiness. The audit traced it to one root cause and delivered a ready-to-apply fix pack for the in-house developer. An earlier GA4 audit decoded 29 live tracking requests across 15 simulated user journeys to show exactly what analytics was and wasn't recording.
Founder track record: this work was led by Clarafold's founder before founding Clarafold, while working as marketing manager at a maritime compliance software company. The client's name is withheld until it approves publication; every figure comes from its own systems. Clarafold client engagements are added as they complete.
- AI-search readiness baseline
- 32/100
- pages audited, all empty to AI crawlers
- 28
- GA4 requests decoded across 15 scenarios
- 29
- root cause behind most issues
- 1
01 · Client
Maritime compliance software company
Maritime compliance software and services company. Name withheld until the client approves publication.
The challenge
The company had genuinely strong material (seafarer founders, industry accreditations, technically precise articles), yet AI assistants answering 'which mooring software helps with MEG4?' or 'who does VDR audits?' named competitors instead. Separately, leadership could not trust what the analytics said about which campaigns produced demand.
02 · System
What existed before.
- A React website that rendered entirely in the browser
- Every URL, including broken ones, returning a 200 'OK' page
- No structured data, no sitemap, no llms.txt; one meta description on all 28 pages
- Analytics events that had never been verified end to end
03 · Intervention
What was built and run.
Looked at the site the way a crawler does
Fetched every page with AI-crawler user agents and compared the result with a full headless-browser render.
Traced symptoms to causes
The duplicate meta descriptions turned out to be a static tag overriding the page-specific ones developers had already written: a one-line fix, not 28 rewrites.
Checked the brand off-site
Searched how third-party sources and AI answers describe the company and its products, and where competitors are cited instead.
Packaged the fix for the developer
A prerender script, nginx configuration, structured data, sitemap, robots.txt, llms.txt and corrected meta descriptions, each with a one-line verification command.
Audited analytics with a real browser
Simulated 15 user journeys (page views, scrolls, form fills, CTA clicks, downloads) and decoded every GA4 request to map what was tracked against what mattered.
04 · Outcome
What changed.
A 32/100 baseline with every finding prioritised by severity and effort, and a one-page brief for the CEO with a 30-day plan designed to run in-house, without an agency.
A developer handover pack covering the critical fixes, estimated at one to two days of developer time rather than a rebuild.
Surfaced issues no one had spotted, including unfinished product pages reachable by search engines and an analytics container that had never been configured.
A GA4 event-gap table and attribution design that defined what 'a lead from campaign X' would actually mean in the data.
05 · Lessons
What we would tell you.
- AI search is a rendering problem before it is a content problem. The best article in the world does nothing if the crawler receives an empty page.
- Audit the root cause, not the symptom list. Most of 25+ findings came from one architectural decision.
- Hand developers fixes, not opinions. Ready files and a verification command for each move faster than a slide of recommendations.
Put It To Work