Adversarial Plan Review: Two Independent Frontier Models

Plan reviewed: C:\Users\annet\ClaudeProjects\SEOBeliever\PLAN.md (90-day execution plan, dated 2026-06-01)

Review run: 2026-06-01 via OpenRouter (OpenAI-compatible API)

Method: Identical hostile-reviewer prompt sent to two different model families. Responses preserved verbatim below.

Model IDs that actually answered

• Grok (xAI): answered as x-ai/grok-4.3. The originally requested IDs all 404ed: x-ai/grok-4 and x-ai/grok-3 (both deprecated, xAI redirects callers to Grok 4.3), and x-ai/grok-2-1212 (no endpoints available). Fell forward to the current x-ai/grok-4.3.
• Frontier reasoning model: answered as openai/gpt-5 on the first attempt. The prepared fallbacks (openai/o3, openai/gpt-4o, anthropic/claude-opus-4.7) were not needed.

---

Review 1: Grok (xAI), model x-ai/grok-4.3

1. Sequencing error: Phase B2 (St. Louis HRT anchor) is scheduled for weeks 3-7 while A1-A3 remain incomplete and RankGridEngine is still under build. The national openness scan in B1 cannot run until DataForSEO is funded and the engine is coded, yet the plan treats the warm-intro landing as parallel. This creates a hard dependency chain that collapses if any prior step slips. It matters because 15 hrs/week gives zero slack for overlap; one delay pushes the entire 90-day revenue signal past the runway. Fix: move all new-anchor outreach to week 5 at earliest and require completed RankGridEngine plus 30-day Rebuilding data before any pitch occurs.

1. Unrealistic bandwidth: Annette is assigned on-camera YouTube, sales calls, owner relationships, quality bar enforcement, and strategic oversight simultaneously within a hard 15-hour cap. Offshore and Claude are listed for volume, but both still require her training, review, and exception handling. This breaks in week 4 when the first client deliverable needs review while a new sales call and training session collide. It matters because lifestyle caps are non-negotiable per the plan, so the work simply does not ship. Fix: remove YouTube entirely until month four and cap live client relationships at one until the first offshore hire is already billing revenue.

1. Single point of failure on warm intro: The entire national HRT vertical and St. Louis anchor rest on one daughter-provided introduction. No backup cities or outreach channels are specified if the contact declines or the clinic is not open. It matters because the plan already dropped every other vertical with similar access friction; losing this single path leaves zero qualified targets inside 90 days. Fix: run the openness scan on three additional metros in parallel during week 4 and require at least two documented inbound leads before any pilot is offered.

1. Vanity metrics masquerading as success: The 90-day list includes AI citations, GBP 3-pack appearances, DR growth past 15, and published before/after pages. None are tied to booked revenue, pilot-to-paid conversion, or cash collected. It matters because the runway is only $3,500/month and every week without new paying clients burns the buffer. Fix: replace every metric with “minimum one paid pilot at $X/month signed by day 60” and “second anchor under contract by day 75,” with public scoreboard tracking dollars, not rankings.

1. Riskiest unstated assumption: The plan assumes the Rebuilding Seminars case study will produce a clean, publishable before/after with measurable Maps and AI wins inside 60 days, thereby unlocking all subsequent credibility and sales. No allowance is made for client delays on implementation, tracking setup failures, or results that are modest or negative. It matters because the entire brand promise (“measure, don’t guess”) and future pitches collapse without that first named proof point. Fix: require an alternative case study source (even a prior Bone Voyage-style project) to be fully written and published by day 45 regardless of Rebuilding outcomes.

1. Underestimated sales-cycle and delegation friction: Offshore hire timing is “once a closed pilot pipeline justifies it,” yet no estimate exists for how many calls or how many weeks of relationship-building are required to reach that pipeline. GEO work is explicitly kept close because it is not yet proceduralized. It matters because 15 hours cannot simultaneously run discovery calls, train staff, and maintain quality on two verticals. Fix: pre-write the full repeatable GEO checklist and citation SOP before any hire is authorized, and budget four weeks of pure sales activity with zero concurrent case work.

Single biggest reason it fails: The entire revenue engine depends on landing and converting a warm-intro HRT client in St. Louis before any other proof exists, yet no time is allocated for the inevitable discovery calls, objections, and pilot scoping that must still be done personally inside the 15-hour limit. When that single conversation does not close on schedule, every downstream metric, hire, and metric stays at zero while the runway burns.

---

Review 2: Frontier reasoning model, model openai/gpt-5

Brutal truth: this reads like a maker’s playground, not a go-to-market plan. With 15 hours a week, you are spreading scarce attention across tool-building, YouTube, and four verticals while measuring the wrong outcomes. Here are the ranked top failure points and fixes.

1) No revenue gates. Success metrics are vanity

• Why it matters: Being cited in AI, DR > 15, and a public scoreboard do not pay bills. Without booked, collected revenue milestones, you will burn runway building artifacts that never convert into cash.
• Specific fix:
• Add two hard gates: by Day 30 at least 1 paid pilot signed and collected. By Day 60 at least 2 paying anchors with total MRR at or above $3,500. If not met, pause all non-selling work.
• Replace “success metrics” with:
• Number of discovery calls booked per week
• Proposals sent and decisioned per week
• Pilots started
• Cash collected and MRR
• Lead source split
• Make all other metrics secondary.

2) Bandwidth fantasy vs workload

• Why it matters: 15 hours is not enough to sell, deliver, build a custom rank-grid engine, launch a YouTube channel, produce case studies, and train offshore staff. You will context-switch and miss critical windows with prospects.
• Specific fix:
• Hard weekly allocation until Day 60: 6 hours sales and follow-ups, 6 hours delivery for 1 pilot, 2 hours process/SOP, 1 hour admin. Zero hours on YouTube and zero engineering until revenue gate is met.
• Set a WIP cap of 1 live client and 1 active sales vertical until a pilot converts to paying.

3) Tool-building trap: RankGridEngine before revenue

• Why it matters: You are burning weeks on an internal tool that already exists in the market. No client cares that you “built your own” if you cannot show results. Engineering is a single point of failure on your time.
• Specific fix:
• Kill custom build in Phase A and B. Use off-the-shelf: Local Falcon or Local Viking for grids, PlePer for GBP audit, LowFruits/KeywordInsights for clustering. Budget $300 per month max.
• Revisit internal build only after 3 paying clients and at least 60 percent utilization of billable hours.

4) Sequencing error: gating decisions on 30-day SEO outcomes

• Why it matters: Local SEO rarely moves reliably in 30 days. Gating anchor #3 on a 30-day signal from anchor #1 means you will stall for the wrong reason.
• Specific fix:
• Gate on inputs you control, not rank movement. Examples:
• Completed tech audit and fixes by Day 14
• GBP fully optimized with categories, services, photos, Q&A, products by Day 10
• 1 cornerstone cost page shipped by Day 14
• 20 citations claimed by Day 21
• Review velocity target achieved by Day 30
• Open a second anchor once these inputs are completed and client has approved the case study terms, regardless of early rank noise.

5) Risky, weak offer design and free pilots

• Why it matters: Free pilots attract uncommitted clients, extend sales cycles, and destroy perceived value. You will train tire-kickers and publish nothing because approvals stall.
• Specific fix:
• Create a 90-day paid pilot with risk reversal:
• Price: $1,500 setup + $1,250 per month for 3 months. Money-back guarantee if target input milestones are delivered and there are zero measurable lifts in at least two leading indicators (GBP actions, calls, discovery impressions) by Day 60.
• Contract includes explicit permission to publish a case study with anonymized spend and performance, plus logo usage with approval.
• Payment upfront each month. No net 30.

6) AI citation focus is unproven and measured poorly

• Why it matters: AI Overviews, ChatGPT, Gemini, and Perplexity outputs are volatile, personalized, and not attributable to revenue. Screenshots are not a reliable metric. You risk optimizing for noise and selling a story you cannot reproduce.
• Specific fix:
• Treat AI citations as an R&D layer only. Do not include in 90-day success criteria.
• Measurement protocol: use a fixed query list, logged-out Chrome profiles, location fixed by GPS override, capture HTML and timestamps, and store in a repo. Track only presence/absence and position where deterministically measurable.
• Primary success criteria: GBP actions, tracked calls, form fills, booked revenue. Use CallRail or similar and UTM discipline.

7) Split-focus across four verticals with thin validation

• Why it matters: You cannot pursue Boulder dentistry, Boulder ortho, Boulder med spas, and national HRT while selling and delivering in 15 hours. Thin “openness” checks are not go-to-market proof.
• Specific fix:
• Pick one beachhead for 60 days. Either Boulder dentistry implant-cost or St. Louis HRT, not both.
• For the chosen vertical, run a 1-week “Open, Willing, Able” validation:
• 15 ICP accounts
• 10 discovery calls booked
• 5 live audits delivered
• 3 proposals sent
• If you cannot hit this in 2 weeks, switch vertical.

8) Single point of failure: you and your “warm intro” pipeline

• Why it matters: One person doing sales, QA, and relationships is fragile and does not scale. Relying on a daughter’s intro for the first anchor is not a pipeline.
• Specific fix:
• Hire a part-time SDR or VA for 10 hours per week immediately on a $600 to $800 budget for 60 days to:
• Build lists of 100 ICPs per week
• Book 5 discovery calls per week to your calendar
• Send a standardized audit-led offer
• Create a one-page offer deck and a 30-minute discovery script. Calendar two fixed sales blocks each week.

9) Compliance and E-E-A-T gaps for medical cost content and HRT

• Why it matters: Dentistry and HRT are YMYL. Cost claims and medical advice require disclaimers, licensed review, sourcing, and on-page proof of expertise. AI-written content without medical oversight risks trust, rankings, and legal trouble.
• Specific fix:
• Retain a licensed clinician per vertical as content reviewer at $300 to $500 per month. Add “Medically reviewed by” bios with NPI links.
• Add clear disclaimers on cost pages, cite state board rules, list what is included or excluded in prices.
• Sign BAAs if you touch PHI. Use HIPAA-compliant call tracking and form handling for clinics.

10) DataForSEO dependency and cost control are under-scoped

• Why it matters: Hitting APIs at scale can burn budget or get throttled. Google SERP products shift. You cannot afford to debug rate limits during sales windows.
• Specific fix:
• Cap total external data spend at $200 per month until revenue gate is hit.
• Prefer existing rank-grid SaaS that absorbs API volatility.
• If you insist on DataForSEO, precompute monthly cost ceilings by keyword and grid count. Enforce an automated hard stop. No “Live” mode until a client is paying.

11) Bone Voyage proof risks credibility

• Why it matters: DR jumps are not proof of business outcomes. Without traffic-to-lead-to-revenue attribution or client attestation, you are over-claiming and inviting skepticism.
• Specific fix:
• Publish a signed case-study attestation or redact the client and focus on measurable business metrics: organic leads per month, booked revenue lift, CAC payback.
• Remove DR as the hero claim. Relegate it to a supporting metric.

12) Offshore plan is hand-wavy and ignores onboarding tax

• Why it matters: Offshore only helps if SOPs, QC, and tool access are tight. Otherwise you create rework and churn precious hours on training.
• Specific fix:
• Before hiring, write three SOPs with checklists: GBP optimization, citation build-out, review outreach workflow. Add acceptance criteria and spot-check sampling.
• Run a paid test task with 3 candidates. Keep 1. Budget $8 to $12 per hour for 20 hours in Month 2 only if you have a paying client.
• Implement a QA checklist you can review in 30 minutes per week.

13) Website spine prioritizes trivia over conversion

• Why it matters: llms.txt and maximal schema will not produce leads in 90 days. Missing are clear offers, booking, and authority assets that convert.
• Specific fix:
• Cut llms.txt and long-tail schema busywork until revenue gate is met.
• Ship immediately:
• Services and pricing page with the 90-day pilot offer
• 2 detailed teardown audits as lead magnets
• Calendar booking
• Case-study template with redacted numbers if approvals lag

14) Measurement noise and lack of causal chain

• Why it matters: “GBP in 3-pack for ≥3 target queries” and “AI citations for SEO consultant Boulder” can fluctuate daily. You cannot attribute cause or use them to make go/no-go decisions.
• Specific fix:
• Build a simple causal dashboard per client:
• Inputs: fixes shipped, citations claimed, reviews added, pages published
• Leading indicators: GBP discovery impressions, calls, direction requests
• Lagging: organic sessions to target pages, tracked calls, form fills, closed revenue
• Weekly review drives next actions. Public scoreboard is optional and private until you have two clean stories.

15) Pricing and scope are undefined

• Why it matters: Without clear deliverables and pricing tiers, sales stalls and scope creeps into your 15 hours.
• Specific fix:
• Define 3 packages now:
• Audit and Quick Wins: $1,500 one-time, 2-week turnaround
• 90-Day Local Pilot: $1,500 setup + $1,250 per month, scope listed
• Ongoing Local Growth: $1,750 to $2,500 per month retainer, defined cadence
• Write SOW templates and a change-order policy.

16) Fragile reliance on specific models and toggling AI features

• Why it matters: Your “Claude Code” multiplier and AI Overviews are outside your control. Model changes or API costs can gut your throughput.
• Specific fix:
• Maintain at least one alternative stack: GPT-4o or open-source LLM locally via LM Studio for drafting. Keep prompts in a repo. Timebox all AI drafting and always human-edit for E-E-A-T pages.

17) Weak risk budgeting given $3,500 per month runway

• Why it matters: Hitting YouTube, engineering, and national scans can easily exceed $1,000 in tools and services per month with no near-term ROI.
• Specific fix:
• Put the monthly budget in writing and enforce:
• Tools and data: ≤ $300
• Marketing content: $0 until revenue gate
• SDR/VA: $600 to $800
• Contingency: $200
• If Month 1 revenue gate is missed, cut SDR and go full sales yourself for 2 weeks.

18) YouTube in Phase C is still a distraction inside 90 days

• Why it matters: YouTube is a compounding asset with a 6 to 12 month payback. Even in Phase C it steals focus from delivery and sales.
• Specific fix:
• Defer YouTube until you have 2 paying anchors and a repeatable delivery rhythm. If you must, record 2 videos in a single 2-hour batch in Week 10 and outsource editing with a $150 cap.

19) Case Study Zero may not map to your ICP

• Why it matters: Rebuilding Seminars sounds unlike dental or HRT local services. Publishing an irrelevant before/after does not help you sell the verticals you want.
• Specific fix:
• Use Rebuilding only for the prediction scoreboard experiment. Do not feature it as your primary case unless it is a local service with GBP, reviews, and maps exposure. Your first hero case must be from the chosen vertical.

20) Legal and platform risks of scraping and publishing

• Why it matters: Automating SERP data and publishing AI engine outputs can violate terms and invite blocks or takedowns. You cannot afford account loss.
• Specific fix:
• Use vendor tools that abstract scraping risk. When you must scrape, throttle and rotate carefully and keep output internal. Get explicit client permission before publishing any SERP screenshots that include competitors.

Single biggest reason this could fail: you are optimizing for artifacts and vanity signals instead of booked, collected revenue, while splitting 15 hours across engineering, content, and four verticals.