AI Mirror — Business Memo

1. Executive summary

AI Mirror turns an invisible acquisition risk into visible evidence: what AI systems say when buyers ask who to choose.

“LG found Linkedist through ChatGPT” is not a cute anecdote. It is a market signal: AI answers are becoming a discovery layer.

The immediate business is a productized audit. The audit asks real buyer prompts across AI systems and returns screenshots, visibility scores, competitor comparisons, incorrect claims, and fixes. The first customer should not buy software. They should buy certainty.

The discipline: Do not build a SaaS platform until customers have paid for repeated manual audits and asked for monitoring.

2. Core thesis

Search behavior is splitting. Buyers still use Google, but increasingly ask AI systems for shortlists, comparisons, alternatives, and recommendations.

When a buyer asks ChatGPT, “Which agency should we hire?”, the answer may become a hidden lead-routing event. If the AI recommends a competitor, the target company never sees the lost demand. No impression. No click. No attribution. Just absence.

Old world

Company competes for Google rankings, ads, referrals, review sites, and direct brand recall.

New layer

Company also competes inside generated AI answers, where the buyer may accept the shortlist without visiting ten websites.

First-principles breakdown

Buyers need trusted shortcuts.
AI systems provide compressed recommendations.
Recommendations are shaped by public information, citations, brand clarity, category language, and source consistency.
If a company is absent or misdescribed, it loses opportunities before the sales funnel begins.
Companies will pay to see and reduce that invisible loss if the proof is concrete enough.

3. Market timing and signal

The category is early, messy, and full of bad language: GEO, AEO, AI SEO, LLM optimization. That is good for a wedge if the offer avoids jargon.

Strong signal: A public Lithuanian agency case claims a major enterprise client discovered them through ChatGPT and signed a large multi-year contract. Even if treated conservatively, it proves the story is believable and commercially interesting.

The market does not need another “AI SEO platform” pitch. It needs a simple fear/curiosity product:

See what ChatGPT says about your company.

Why now?

Buyers are experimenting with AI for vendor discovery.
Companies do not know whether they appear in AI answers.
Founders and marketers fear competitors getting there first.
Most existing SEO agencies will rebrand slowly and vaguely.
A manual, receipt-based audit can move faster than tooling-heavy competitors.

4. Customer, pain, and best first niches

The buyer is not “everyone with a company.” The best buyer has enough margin, competitor pressure, and brand/search sensitivity to care.

Segment	Fit	Why it may buy	Concern
B2B SaaS	Strong	Buyers ask for comparisons, alternatives, best tools, use-case recommendations.	Competitive category may already be noisy.
Agencies	Strong	They understand lead-gen and can resell the audit to clients.	May copy the offer if shown too much process.
High-ticket local services	Good	Clinics, legal, real estate, consultants can monetize one lead.	AI buyer behavior varies by category.
Hotels / restaurants	Selective	Tourists may ask AI for recommendations.	Low willingness to pay unless premium or group-owned.
Small local businesses	Weak	Curiosity exists.	Budget and sophistication are too low.

Bad ICP filter: If a company cannot clearly name its competitors, target buyer, category, and acquisition value, it is too early for this audit.

5. Product offer

The offer should be concrete, one-time, and receipt-driven. No dashboard, no login, no subscription at the start.

Offer name: AI Mirror Audit — “See what ChatGPT says about your company.”

Included in the first paid audit

50 buyer-intent prompts across category, alternatives, comparison, location, budget, and use-case.
4 AI systems tested: ChatGPT, Claude, Gemini, Perplexity.
Competitor visibility map.
Prompt-by-prompt screenshots or exported receipts.
Wrong, stale, missing, or confusing claims.
Top public-source gaps likely causing weak AI representation.
10 prioritized fixes the company can execute.
Optional 30-minute walkthrough for higher pricing tier.

Positioning

Say this

“We show where AI recommends your competitors, ignores you, or gets you wrong.”

Do not say this

“We optimize your AI search visibility through GEO/AEO workflows.” Too abstract. Too agency-slop.

6. Delivery workflow

The first version should be manually delivered with enough automation to save time, but not enough to hide from customer reality.

Intake: company URL, target buyer, geography, core services, top 3–5 competitors, current positioning.
Prompt map: generate 50 buyer prompts across discovery, comparison, alternative, objection, local, and use-case queries.
AI runs: execute prompts across selected AI systems using controlled accounts/sessions where possible.
Capture: save answers, screenshots, citations, timestamps, model names, and visible sources.
Score: measure mention frequency, ranking order, sentiment/description quality, wrong claims, and competitor dominance.
Diagnose: inspect public pages, profiles, content, schema, reviews, case studies, and third-party mentions causing the AI picture.
Recommend: deliver 10 fixes ranked by likely impact and effort.
Retest: optional paid follow-up after fixes are implemented.

Important limitation: AI answers are probabilistic. The audit must say “snapshot under these prompts/models/date,” not pretend to be perfect truth.

7. Client acquisition: how it actually works

The acquisition engine is not ads. It is founder-led outbound using painful personalized proof.

Acquisition mechanism

Choose one niche for 7 days.
List 50 companies with visible competitors and meaningful customer value.
Run 5 mini-prompts per company.
Find the ugly result: competitor appears, company absent, wrong description, stale claim.
Send one short message with a screenshot or concrete stat.
Offer the full 50-prompt audit for €299.

Subject: ChatGPT recommended your competitor I tested 5 buyer prompts in your category. In 4/5, ChatGPT mentioned [Competitor]. In 0/5, it mentioned [Company]. This may be nothing, or it may be invisible lost demand. I can run the full AI Mirror Audit: 50 buyer prompts, 4 AI systems, competitor visibility map, wrong/stale claims, and 10 fixes. One-time: €299. Want me to send the sample?

Channels ranked

Channel	Use first?	Reason
Founder-led LinkedIn DM	Yes	Best for screenshot-led, personalized proof and fast replies.
Cold email	Yes	Scales better after copy is proven.
Public LinkedIn posts	Yes	Use anonymized “AI recommended competitor 29/50” examples.
SEO agencies partnerships	Soon	Possible reseller channel after delivery is repeatable.
Paid ads	No	Too early. Need proof of conversion first.

8. Technology architecture

The tech should support fast manual delivery first, then automate the boring parts only after the audit sells.

V0: manual + lightweight automation

Input: form or Notion/Airtable intake.
Prompt generation: prompt templates + human review.
AI querying: manual browser sessions and/or API calls where available.
Capture: screenshots, JSON/text exports, timestamped model metadata.
Scoring: spreadsheet or simple script calculating mentions, competitors, wrong claims.
Report: generated HTML/PDF with screenshots, tables, and recommendations.

V1: internal audit engine

Company profile → Prompt set generator → Model runners → Answer store → Entity extraction → Competitor scoring → Claim diagnosis → Report builder

Data layer

Postgres for companies, prompts, answers, model metadata, competitor mentions, findings, reports.

Workers

Queue-based prompt execution, retries, rate limits, screenshot capture, source extraction.

Report layer

Static HTML to PDF, shareable private links, exportable CSV, and client-ready recommendations.

What not to build yet

Client dashboard.
User accounts.
Self-serve prompt playground.
Monthly monitoring system.
Automated “AI optimization” content generator.

Tech risk: If the product starts with infrastructure, it becomes avoidance. The first technical goal is faster delivery of paid reports, not software aesthetics.

9. Pricing and business model

Start cheap enough to remove friction, but high enough to prove willingness to pay.

Tier	Price	Use	Notes
Snapshot	€99	Fast validation / impulse buy	Risk: too cheap, attracts low-quality buyers.
AI Mirror Audit	€299	Best starting offer	50 prompts, 4 systems, fixes. Good pain/price ratio.
Audit + walkthrough	€799	Higher-intent B2B customers	Add call, deeper competitor diagnosis, fix roadmap.
Monthly monitoring	€300–€1,000/mo	Only after repeated demand	Retest prompts, monitor competitor changes, recommend updates.

Recommendation: Start at €299. If customers buy too easily or want a call, create €799 premium tier. Do not lead with monthly subscription until they ask for retesting.

10. Risks and failure modes

This can be real and still fail if executed vaguely. The largest risk is not market timing. It is turning a sharp pain into generic AI-marketing slop.

Risk	Why it matters	Mitigation
Education burden	Customers may not understand GEO/AEO.	Never lead with acronyms. Lead with “ChatGPT recommended your competitor.”
AI answer variability	Results differ by model, time, and prompt wording.	Frame as dated snapshot. Use prompt set, screenshots, timestamps, model names.
Weak causality	Hard to prove fixes caused better AI answers.	Sell audit first, not guaranteed rankings. Retest after changes.
Agency commoditization	SEO agencies may copy positioning.	Move fast, specialize by niche, build proprietary prompt/report process.
Building too early	Dashboard becomes fake progress.	Ten paid audits before platform.
Low-quality clients	Small clients may want miracles for €299.	Filter for serious B2B/high-ticket buyers only.

11. Validation plan

Validation must measure payments, not excitement. “Interesting” is not signal. Money is signal.

7-day sprint

Pick one niche. Recommended: B2B agencies or B2B SaaS.
Create 30 prospect list.
Run 5 mini-prompts per prospect.
Send personalized outbound with a screenshot/stat.
Offer €299 audit with 48h delivery.
Close at least 2 paid audits.
Deliver manually and capture objections, questions, and willingness for monthly retest.

Decision gates

Green light

2+ paid audits from 30 targeted prospects, or 5+ strong calls with clear budget and urgency.

Yellow light

Good replies but no payment. Change niche, proof format, or price. Do not build.

Red light

People like the idea but do not forward, pay, or ask for their own result. Kill or reposition.

Brutal metric: If you cannot sell 2 paid audits manually, software will not save it.

12. Roadmap

Phase 0
Today

Create landing sample, memo, 1-page report template, and first outbound list. No engineering beyond artifacts.

Phase 1
Days 1–7

Manual outbound to 30 prospects. Run mini-prompts. Close 2 paid audits. Deliver manually.

Phase 2
Days 8–21

Standardize prompt library, scoring sheet, report generator, and delivery checklist. Raise price if demand is real.

Phase 3
Month 2

Build internal audit engine: prompt runners, answer store, entity extraction, scoring, HTML/PDF report export.

Phase 4
Only if pulled

Monthly monitoring, agency reseller package, niche-specific benchmarks, and private client portal.

Final recommendation

Pursue it, but only as a focused paid validation sprint.

The correct posture is not “we found the next SaaS.” The correct posture is: “we found a sharp commercial anxiety created by AI answers; now we test whether companies will pay to see their own exposure.”

The next action: Pick one niche, run 30 personalized mini-audits, and sell 2 paid AI Mirror Audits at €299 before writing another line of product code.