From 78 to 91 in One Afternoon: A Documented AI Visibility Case Study
This is a real, documented case study. We ate our own dog food — used our own tool to audit AIExposureTool's AI visibility, identified 7 specific failing checks worth 29 points, and fixed all of them in under 4 hours. Score went from 78 to 91. This page documents every fix, the exact time it took, and the point impact — so you know exactly what works.
Latest: April 16, 2026 · 78 → 91
Today we fixed 7 failing checks in a single afternoon and jumped 13 points. The exact fixes and times are documented below. This is proof the process works — on ourselves, with real data.
56 → 91
Starting → current
+13 pts
Today's fix session
< 4 hrs
Time to implement
7
Checks fixed today
The Starting Point: Score 56, Zero AI Mentions
On March 15, 2026, we ran our own tool on aiexposuretool.com. The result was humbling: 56 out of 100. For a tool that helps others improve their AI visibility, we had significant gaps ourselves.
The audit found that:
- Our robots.txt did not explicitly allow AI crawlers
- We had no llms.txt file — AI crawlers had no structured summary of our product
- No JSON-LD structured data — no SoftwareApplication or FAQPage schema
- No FAQ section or schema markup
- No comparison pages
- Zero social proof visible to AI crawlers
We had 870 Google Search impressions growing steadily. But when we asked ChatGPT, Perplexity, Gemini, and Copilot about AI visibility tools — we did not appear in any answer.
Today's Session: 78 → 91 in Under 4 Hours
On April 16, 2026 — Day 32 — the score was sitting at 78. The audit identified 7 specific failing checks worth 29 points combined: no testimonials in crawlable HTML, no quantifiable metrics, no customer mentions, no integrations page, no case studies, low text-to-HTML ratio, and no answer-first content structure in key sections.
We fixed all 7 in a single afternoon. Here is what actually happened:
- Shipped a SocialProof section on the homepage with 3 real customer testimonials (names, roles, companies in
<blockquote>HTML), 4 quantifiable metrics (“1,000+ sites scanned”, “25+ signals”, “7 platforms”, “60s”), and a trust line naming the platforms AI users care about. - Created /integrations — a dedicated page listing all 7 AI platforms we monitor and 7 infrastructure services we integrate with. This closes the “integrations clarity” check.
- Created /about — dedicated EEAT page with AboutPage JSON-LD, product story, and transparent methodology references.
- Created /case-study (this page) — living document that doubles as EEAT content and social proof.
- Upgraded the FAQ — added FAQPage JSON-LD schema (the #1 most-cited structured data type by AI), made answers always render in HTML via
sr-onlywhen collapsed so AI crawlers extract them without needing to execute clicks, and added two new Q&As about what the product is and what it costs. - Added answer-first intros to ProblemSection, HowItWorks, WhatYouGet, and WhoItsFor. Every section now leads with “AIExposureTool is/does/checks...” so AI extracts the entity facts from the first sentence.
The result: 78 → 91 on a rescan. +13 points. Total implementation time: under 4 hours. Every fix is documented below with exact time and point impact.
Score Timeline: Every Change Tracked
56
Mar 15
First scan. No llms.txt, no JSON-LD, no FAQ schema. AI crawlers not explicitly allowed. Starting from scratch.
66
Mar 15
Added robots.txt rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended. +10 points in 5 minutes.
86
Mar 15
Deployed llms.txt + JSON-LD structured data. +20 points. From 56 to 86 in one afternoon.
79
Mar 20
Homepage redesign broke content structure. React components replaced crawlable text. Score dropped.
89
Mar 21
Fixed content, added FAQ schema, created comparison pages. Recovery + improvement.
91
Mar 23
Peak score. Added social profile links, review platform presence (Product Hunt), comparison content. All AI crawlers allowed.
84
Apr 1
Major feature launch — new React components increased JS ratio, dropped text-to-HTML ratio to 1%. Score dipped.
78
Apr 9
More features shipped. Trust signals section removed during redesign. Score dropped further.
78
Apr 16 (AM)
Starting point for today's fix session. 870 Google impressions, 0 confirmed AI mentions. Identified 7 failing checks worth 29 points.
91
Apr 16 (PM)
Back to peak in one afternoon. +13 points in under 4 hours by fixing 7 specific checks: testimonials, metrics, trust line, integrations page, case study, FAQPage schema, dense answer-first content on all sections.
What We Fixed (and the Impact)
| Fix | Time | Impact |
|---|---|---|
Allowed AI crawlers in robots.txt Added GPTBot, ClaudeBot, PerplexityBot, Google-Extended, meta-externalagent. Went from blocked to fully accessible. | 5 min | +10 pts |
Deployed llms.txt Auto-generated from our audit, then edited for accuracy. 3,211 characters covering product name, features, pricing, audience. | 10 min | +8 pts |
Added JSON-LD structured data SoftwareApplication schema with real pricing, feature list, aggregate rating. Organization schema with sameAs links. | 15 min | +12 pts |
Created FAQ schema 10 questions covering pricing, features, competitors, free plan. FAQPage JSON-LD in the homepage head. | 20 min | +5 pts |
Built comparison pages Created /compare/otterly, /compare/peec-ai, /compare/evertune with feature tables and honest pros/cons. | 2 hrs | +3 pts |
Added social profile links Twitter/X, LinkedIn, GitHub, Product Hunt in footer and Organization schema sameAs. | 5 min | +3 pts |
Listed on Product Hunt Free launch listing. Gets indexed by Google and cited by AI as third-party validation. | 30 min | +3 pts |
Added testimonials in crawlable HTML Apr 16 — 3 real customer testimonials with names, roles, and companies inside <blockquote> tags. AI needs social proof it can parse, not images or JS. | 30 min | +5 pts |
Added quantifiable metrics Apr 16 — '1,000+ sites scanned', '25+ signals', '7 platforms', '60s' in plain text headings. Specific numbers are cited by AI more reliably than vague claims. | 10 min | +5 pts |
Added customer mention trust line Apr 16 — 'Trusted by founders, agencies, and growth teams building with Cursor, Claude, and Lovable' in crawlable HTML. | 15 min | +5 pts |
Created /integrations page Apr 16 — Full integrations page listing all 7 AI platforms and 7 infrastructure services with descriptions. Closes the 'no integrations clarity' check. | 30 min | +3 pts |
Created /case-study page (this page) Apr 16 — Living case study documenting the journey. EEAT signal + meta proof that the process works. | 1 hr | +2 pts |
Created /about page with EEAT content Apr 16 — Dedicated About page with AboutPage JSON-LD, product story, methodology transparency, and ~800 words of crawlable EEAT content. | 45 min | included |
Added FAQPage schema to homepage FAQ Apr 16 — FAQ now renders with FAQPage JSON-LD schema and answers always present in HTML (sr-only when collapsed) so AI crawlers extract them without JavaScript. | 15 min | included in +5 |
Added answer-first intros to all sections Apr 16 — Added dense fact-first paragraphs to Problem, HowItWorks, WhatYouGet, WhoItsFor. Every section leads with 'AIExposureTool is/does/checks...' so AI extracts entity facts from the first sentence. | 30 min | +4 pts |
Total time for all fixes: under 4 hours. Total impact: 56 → 91 (+35 points).
Current State: What is Working, What is Not
Passing (8 checks)
- AI Crawl Access: 27/27 — all major AI crawlers explicitly allowed
- Structured Data: 33/33 — SoftwareApplication, Organization, and FAQPage JSON-LD deployed
- llms.txt, llm.json, and llms-full.txt all deployed at site root
- Sitemap present, robots.txt allows all AI crawlers, no WAF blocking
- Product Clarity: 13/15 — clear H1, features described, pricing page crawlable
- Trust & Social Proof: fixed — testimonials with real names, quantifiable metrics (1,000+ sites scanned), customer mention line in crawlable HTML
- EEAT: comparison content at /compare, /about page, /case-study page, /integrations page, social profiles linked, Product Hunt listing
- Answer-first content: dense intro paragraphs on Problem, HowItWorks, WhatYouGet, WhoItsFor sections
Failing (2 checks)
- Text-to-HTML ratio still on the low side (React-heavy homepage) — targeting further improvement
- Customer logos (image logos of named customers) — pending real customer permissions
The irony
We are an AI visibility tool with perfect technical scores — every crawler allowed, every structured data type deployed, every discovery file present. But we score 0/15 on Trust and Social Proof because we have no visible testimonials, no customer logos, and no quantifiable metrics. AI will not confidently recommend a product that nobody else visibly validates.
What We Are Fixing Now
This case study is a living document. Here is what we are shipping today and this week:
Rescan weekly to catch score regressions
60 sec each · Expected: monitoring
Start real AI mention tracking (add 5 buyer prompts)
5 min · Expected: visibility data
Further reduce text-to-HTML ratio (more SSR content)
2 hrs · Expected: +2-3 pts
Collect permissioned customer logos
ongoing · Expected: +5 pts
Publish /case-study update when first AI mention happens
30 min · Expected: social proof
Key Learnings So Far
- llms.txt + JSON-LD = instant 20+ point jump. We went from 56 to 86 in one afternoon. These two files are the highest-ROI change any site can make.
- Redesigns kill scores. Every homepage redesign dropped our score because new React components reduced text-to-HTML ratio. Ship content alongside code changes.
- Trust signals are the hardest to earn. Technical fixes take hours. Social proof takes weeks or months of real usage and customer relationships.
- Score fluctuates and that is normal. We have been between 78 and 91. Continuous monitoring catches regressions that otherwise go unnoticed.
- Google impressions do not equal AI mentions. We have 870 Google impressions but zero confirmed AI mentions. SEO and AI visibility are different games with different signals.
How This Helps You
Everything we did to improve our score is exactly what AIExposureTool does for you:
- Scan your site — get your AI Exposure Score and see every blocker, just like we did on day 1
- Get auto-generated fix files — llms.txt, JSON-LD, FAQ schema, robots.txt suggestions
- Follow the fix roadmap — prioritized by impact, with estimated time and point gains
- Track whether AI mentions you — daily monitoring across 7 AI platforms shows if the fixes are working
- See competitors — know who AI recommends instead of you and why
We will keep updating this case study as we fix more issues and track AI mention progress. Bookmark this page or scan your own site to start your journey.