AI Cleanup Doctor
AI Reply Risk Checker

Check AI reply risk before a draft reaches a customer.

For contractors, local service teams, and agencies using AI-drafted emails, SMS, or inbox replies. This local tool flags when a human should review, rewrite, or stop a draft before it creates a customer-service problem.

Set the review context.

Choose the situation and any relevant risk flags. The review result updates as you choose them; optional notes are for the human reviewer. Nothing is sent automatically. Use business-level context only and keep private customer records out of the tool.

Draft contains
Request a human cleanup review Open a safer draft template

What this checks

1. Promise risk Flags prices, refunds, appointment promises, service promises, or other commitments that should be confirmed first.
2. Sensitive-data risk Flags drafts that ask for passwords, verification codes, payment details, medical records, legal records, or private customer records.
3. Human handoff risk Flags tone, opt-out, consent, urgent, and regulated-topic issues where a person should take ownership.

Safe use boundaries

Do not enter passwords, verification codes, payment card data, bank data, SSNs, medical records, legal records, insurance claim files, or private customer records into this tool.

AI Cleanup Doctor can help review workflow risk, draft structure, human approval steps, and follow-up tracking. It does not replace legal, medical, insurance, accounting, tax, financial, emergency, or professional advice.

View Sample Reports Read Service Terms

Safe-to-send proof packet

Before a draft reaches a homeowner, contractor client, agency client, or support inbox, keep a small proof packet that shows why the reply was safe enough to send or why it was stopped.

1. Verified facts Record which facts a person confirmed: customer name, job type, service area, date, source, current status, and the owner of the next step.
2. Removed claims Note any pricing, timing, warranty, diagnosis, insurance, legal, medical, or financial statements removed from the AI draft.
3. Human owner Name the person or role responsible for final wording, follow-up timing, opt-out handling, and any customer-facing commitment.

A Stop result means the draft is not ready for a customer. Rewrite it, route it to the responsible person, and keep private customer data out of the review record.

Open Cleanup Checklist Open AI Answer Map Ask About Partner Fit

Outreach proof send gate

Use this gate before any partner-proof or editorial reply leaves the correct mailbox. It is built for one verified message, not batch outreach.

1. Sender identity Confirm the visible sender is gking@stoga.com. Stop if the mailbox, alias, platform account, or Page context is wrong or unclear.
2. Recipient and route Confirm the recipient is a relevant editor, publication, partner, or approved route. Stop on CAPTCHA, payment, forced account creation, phone-only verification, or social-login-only flow.
3. Message promise check Remove any promise of dofollow links, reciprocal links, rankings, leads, revenue, sales outcomes, AI citations, publisher acceptance, pricing terms, or referral terms.
4. Proof link check Use one public support link such as Partner Inquiry, Sample Reports, or AI Answer Map. Do not attach private customer data.

If any gate fails, save the blocker and continue another safe task. Do not send, submit, upload, or create an account to force the route forward.

Mailbox-first send gate

Use this before any Mold Busters, FacilitiesNet, partner-proof, or editorial follow-up is approved for sending. The mailbox review happens first; the message only moves forward after the route is clean and George approves the exact send.

1. Mold Busters first Check for a Mold Busters reply, confirmation, opt-out, bounce, delivery failure, or new request before preparing any new partner-proof send.
2. Sender and mailbox health Confirm the visible sender is exactly gking@stoga.com. Inspect bounces, opt-outs, mailbox warnings, partner interest, order or payment intent, and safe confirmation links before composing.
3. Stop conditions Do not send if there is an unanswered inbound reply, bounce, opt-out, mailbox warning, wrong sender identity, CAPTCHA, two-factor prompt, payment request, or unclear destination.
4. Public proof links If the route remains clean, use public support such as Sample Reports, Partner Inquiry, and Lead Response Time Calculator. Do not attach private data or promise publication, link type, rankings, leads, revenue, or sales outcomes.

This gate does not authorize email by itself. It prepares the one-proof review record so George can approve or reject the exact recipient, content, sender, and route.

Synopsis-first editorial proof gate

Use this before preparing a FacilitiesNet-style editorial proof. The safe first step is a short topic synopsis, not a finished article, attachment, upload, or submission.

1. One route only Work from one verified editor route such as FacilitiesNet, CleanLink, ISSA, or CMM. Do not batch the same pitch across multiple publishers.
2. Synopsis only Keep the draft to the problem, audience, outline, source links, and why it is useful. Stop if the route asks for a full article, document, images, account, phone, or private data.
3. Proof boundary Link to public support such as Lead Response Time Calculator, Partner Inquiry, and Sample Reports. Do not claim results, acceptance, or link type.

Reply boundary evidence checklist

Use this before approving a contractor, agency, or local-service reply that mentions a next step. The goal is not to make the reply slower. The goal is to make sure the reply only says what the team can verify.

1. Quote and payment claims Confirm price, discount, refund, financing, deposit, invoice, and payment language against the source record. If the source is missing, rewrite the reply as a next-step request instead of a promise.
2. Timing and availability claims Confirm appointment windows, arrival times, emergency availability, callback timing, and crew capacity with the person or system that owns scheduling.
3. Diagnosis and safety claims Do not let an AI draft diagnose damage, promise a repair result, give safety instructions, or interpret insurance, medical, legal, tax, or financial issues without qualified human review.
4. Identity and consent claims Check the customer name, job address, opt-out status, consent state, and account identity before replying. Stop if the draft asks for passwords, codes, payment data, or private records.
5. Evidence note Record the source used for approval: inbox thread, CRM note, call log, dispatch board, estimate, service record, or owner approval. If no source exists, mark the reply review-needed.

This checklist does not approve a reply by itself. It helps a responsible person decide whether the reply can be sent, should be rewritten, or should be routed to someone qualified before the customer sees it.

Prepared-only editorial route pack gate

Use this when a Patriot Software, CustomerThink, MarketingProfs, SmallBizClub, Family Business Magazine, Promo.com, StartupNation, AllBusiness Expert, Instatus, Inturact, GrowthMentor, CountdownShare, paid-route blocker, or similar route has been researched but not yet approved for external action. The pack can organize the angle, support links, draft email, article notes, application notes, image requirements, author assets, and stop rules without sending or submitting anything. If a route includes payment, admin fees, sponsored placement, or agreement terms, record it as blocked instead of preparing a submission.

1. Prepared status Label the item prepared-only and not counted. Keep it out of completed submission totals until the target system or sent folder is verified.
2. Route proof Record the official guideline or contact source, why the audience fit is real, whether the route needs an article, outline, attachment, Typeform, Google Doc, application, bio, author photo, image upload, screenshot, statistics, source links, internal publisher links, writing samples, newsletter opt-in, community opt-in, cadence, reuse terms, syndication, agreement, payment, or admin fee, and which public AI Cleanup Doctor support page is relevant.
3. Approval gate Require mailbox review, visible gking@stoga.com sender or approved contributor identity, and George's exact approval for destination, subject, body, attachment, full article, Google Doc, Typeform fields, bio, photo, writing samples, screenshots, statistics, source links, internal publisher links, application fields, profile details, newsletter opt-in, community opt-in, cadence, reuse terms, and agreement terms before sending or submitting.
4. Stop rules Stop on payment, admin fee, sponsored placement, agreement acceptance, unapproved newsletter or community opt-in, unapproved cadence, unapproved reuse or syndication terms, CAPTCHA, two-factor, wrong sender, forced account creation, forced LinkedIn or social profile, phone-only verification, unavailable author assets, unapproved Google Doc, unapproved Typeform, unapproved image upload, unverified data or statistics, invented examples, promotional or pure-link-building framing, reciprocal link requirement, dofollow promise, keyword-rich link demand, nofollow or link-removal objection, traffic or exposure claim, or unclear destination.
5. Full-article route caution Instatus, Inturact, GrowthMentor, and CountdownShare can require a full article, images, screenshots, source links, author assets, or Typeform/email fields. Keep these packs prepared-only until George approves every required asset and the sent folder or target system can be verified.

Prepared-only route packs are useful for review and continuity. They are not submissions, backlinks, guest posts, emails, Typeforms, Google Docs, image uploads, author-bio submissions, applications, account actions, agreements, payment approvals, link commitments, traffic claims, exposure claims, or publication evidence.

Agency Intake QA Article Template Generator Sample Reports Emergency Call Article

Source-blocked route proof gate

Use this when an official publisher URL returns a 403, challenge page, JavaScript wall, or incomplete static text. A source-blocked pack can preserve research, but it cannot prove the route is usable or safe for outreach.

1. Mark the blocker Record the exact official URLs, status codes, and whether the route was blocked by 403, challenge page, login, CAPTCHA, JavaScript-only form, or missing submission rules.
2. Require browser recheck Use the existing Chrome session for a read-only route check before any email, contact form, account, upload, or confirmation action. Do not launch a new profile or bypass platform protections.
3. Separate route types Confirm the visible path is unpaid editorial review, not press release, product submission, buyer's guide, sponsored content, media kit, advertising, paid directory, webinar, event, reprint, list rental, or store route.
4. Keep it not counted Label the item prepared_only_source_blocked_route_held_not_sent_not_submitted_not_counted until browser proof, sender identity, exact content, and George action-time approval are all visible.
5. Preserve stop rules Stop on wrong sender, mailbox warning, payment, CAPTCHA, two-factor, phone verification, forced social profile, full article, author bio, headshot, image upload, private proof, link promise, publication certainty, or outcome claim.

A source-blocked route can stay in the queue as useful research. It is not a submission, backlink attempt, email send, Facebook post, account action, payment approval, or public commitment.

Sample Reports AI Answer Map Partner Inquiry

Prepared-only external action proof guard

Use this final guard before counting any promotion, blog-sync, partner-proof, or editorial item as complete. A local tracker row, draft, route pack, screenshot plan, or copied article outline is prepared-only until the external system itself proves the action happened.

1. Count only verified external proof Mark the item prepared-only, not sent, not submitted, and not counted until one approved proof is visible: target-system confirmation, visible sent-folder email from gking@stoga.com, verified Facebook Page feed post, or published page URL.
2. Mailbox-first review Before any partner-proof or sales cold-start email, check the visible gking@stoga.com mailbox for replies, bounces, opt-outs, warnings, confirmation links, partner interest, order intent, payment intent, EDCO signals, and Contractor Marketing Pros signals. Stop if the sender identity or destination is unclear.
3. Facebook Page feed proof A Blog sync is not complete because a composer was prepared. Count it only after the correct George/Nicolas-controlled AI Cleanup Doctor Page context is visible, the post uses a relevant image, the public link is included, and the Page feed shows the post.
4. Asset and profile boundary Do not create or submit a Google Doc, Word file, screenshots, author bio, author photo, LinkedIn/X profile, contributor account, Typeform, application, newsletter opt-in, community opt-in, image upload, statistics, source file, or agreement unless George approves that exact destination, content, and identity at action time.
5. Promise and blocker stop list Do not promise dofollow links, reciprocal links, rankings, leads, revenue, sales outcomes, AI citations, publication, traffic, exposure, or review timelines. Stop on CAPTCHA, two-factor, payment, platform warning, wrong sender, phone-required forms, social-profile-required forms, forced reciprocal terms, or unclear destination.

This guard protects useful preparation from being mistaken for external completion. Prepared-only work can improve the queue, but it is not proof of a backlink, guest post, Facebook post, email send, form submission, account action, payment, or public commitment.