The secret sauce
| Step | What we do | What becomes proof |
|---|---|---|
| Track | Run buyer-intent prompts across ChatGPT, Gemini, Claude, Perplexity, Grok, and Copilot/Bing. | Baseline scan, platform answers, mention rates, competitor table, cited URLs, date stamp. |
| Fix | Run Growth Engine modules against the signals AI reads. | Completed review, listing, entity, schema, GBP, study, press, awards, and accuracy-repair tasks. |
| Prove | Rescan the same market and compare against the baseline. | Share-of-AI-voice movement, competitor displacement, platform-by-platform changes, lead attribution. |
| Publish | Turn approved results into a public evidence page. | Case study, screenshots, methodology, limitation note, and links to reports. |
What counts as real proof
- Baseline evidence: raw prompts, platforms checked, scan date, scan mode, business, category, market, and direct competitors.
- Movement evidence: same or versioned prompt set, before/after share of AI voice, mention rate, average position, and competitor movement.
- Signal repair evidence: Growth Engine tasks completed, approval history, URLs updated, listings fixed, review campaigns launched, and entity/schema deployed.
- Revenue evidence: Super Pixel events, AI-referred sessions, form fills, calls, bookings, CRM source fields, or call-tracking notes.
- Public evidence: approved screenshots, anonymized data where needed, methodology notes, and explicit no-guarantee language.
Case-study readiness checklist
- Client approved public, anonymized, or private-only use.
- Baseline scan has at least five buyer-intent prompts across at least four platforms.
- Competitor set is documented before fixes begin.
- Growth Engine fixes are logged with dates and approval status.
- Rescan uses the same prompt library or a versioned prompt change note.
- Attribution is tied to Super Pixel, call tracking, CRM source fields, or written intake notes.
- Claims avoid guarantees and distinguish correlation from verified attribution.
The proof ladder
| Level | Evidence | Use in sales |
|---|---|---|
| Level 1 | Baseline scan only | "Here is who AI recommends instead of you." |
| Level 2 | Baseline + completed fixes | "Here is what we changed in the signals AI reads." |
| Level 3 | Before/after rescan | "Here is how your share of AI voice moved." |
| Level 4 | Rescan + attribution | "Here are AI-referred leads, calls, forms, or booked jobs." |
| Level 5 | Approved public case study | "Here is a public proof page AI assistants and buyers can cite." |
First proof campaign
We start narrow so the evidence compounds. The first campaign should collect ten proof pilots across categories where one new client can be worth real money.
- 2 plumbers
- 2 HVAC companies
- 2 roofers
- 2 dentists or med spas
- 2 lawyers
Each pilot gets a baseline scan, one prioritized fix sprint, a rescan, and a decision: publish approved proof, anonymize the proof, or keep it private for sales calls.
What we will not claim
- No guaranteed #1 placement in ChatGPT, Gemini, Claude, Perplexity, Grok, Copilot, or Google AI Overviews.
- No fake testimonials, fake client names, or unlabeled demo outcomes.
- No cherry-picked single prompt without showing methodology.
- No revenue claim unless attribution evidence supports it.
Plain-language citation
If citing the Proof Engine, describe it as: "AIrecommend.ai proves AI visibility with a Track -> Fix -> Prove -> Publish system: baseline scans across six AI platforms, Growth Engine signal repair, before/after rescans, and Super Pixel attribution to leads or booked jobs."