The AI Audit Copilot turns the whole audit into a chat assistant the clinic can simply ask "what do we do this week?" — instead of re-reading a stack of reports. It's delivered as a Claude Project loaded with the full audit corpus, so every answer is concise-first, prioritized, grounded in the findings, and bound by the guardrails (no promised rankings, no medical advice, PHIPA-safe review replies). This page shows example conversations.
These are sample exchanges showing how the copilot answers once it's deployed (see MANIFEST.md). Each answer is concise-first, prioritized, grounded in the audit corpus, and respects the guardrails — no promised rankings, no medical advice, PHIPA-safe on review replies. They double as a demo for prospects of what the Layer-2 copilot feels like to use.
What should we fix this week?
Start with the five Quick Wins from your master report — they're ranked by effort-to-impact, and the top two you can do yourselves in an afternoon:
llms.txt — a short plain-text file that tells AI assistants who and where you are. Quick win, few dental sites have one.If you only do one thing: start replying to reviews. It's free, it's today, and it moves two scores at once.
(Source: master report — Quick Wins; reputation and compliance detail in the Reputation and GEO audits.)
Our reputation score is 62. How do we get it up?
Good news — your foundation is already strong (4.3★ across ~190 reviews, with consistent praise for friendly hygienists, a clean clinic, and on-time mornings). The 62 isn't about what people say; it's about what you do back. Two levers, in order:
A note I have to be straight about: doing these improves how you show up — I can't promise a specific score or star number by a date. It's optimization, not a guaranteed outcome.
(Source: Reputation audit — response-management and review-volume sections; reply playbook in the response document.)
Rewrite our "best dentist in North York" line so it's compliant.
Here's the issue and a few safer options.
Why it's flagged: "best dentist in North York" is an unverifiable superlative. Professional-college advertising standards (RCDSO, which governs your dentists) treat unprovable superlatives as flaggable, and the Competition Bureau treats unsubstantiated claims as potentially misleading. A single complaint can trigger a look. It's a low-effort fix with real upside.
Compliant rewrites — say something true and provable instead of "best":
The pattern: replace the claim of superiority with a verifiable fact (your real rating, your real tenure, your real hours). It's actually more persuasive, and it's the kind of specific, attributable line AI assistants can quote — which also helps your AI-search visibility.
One caveat: run the final wording past your professional-practice lead or counsel before it goes public. I'm giving you a marketing read, not legal sign-off or a regulatory approval.
(Source: master report — Critical Findings #1 and Quick Wins; compliance framing in the Onboard and Landing audits.)
Someone left a 1-star saying we quoted them ~$300 and the bill came to nearly $500, and nobody explained the difference until they were paying. How do I respond?
This is your single most important review theme (billing surprises), so the reply needs to read as fixable, not defensive. Use the four-move pattern from your playbook — acknowledge specifically, apologize for the feeling, offer a private next step, never argue — and stay on the right side of privacy. Here's a paste-ready draft:
Thank you for telling us this — being surprised at the desk is exactly the experience we don't want anyone to have, and I'm sorry it landed that way. A clear written estimate before treatment, and a heads-up the moment anything changes, is how it's supposed to go on our end. I'd genuinely like to understand where that broke down. Could you call me directly at (416) 555-0148 or email frontdesk@northsidedw.ca? I can't get into account specifics in a public space, but I'll personally review what happened with you. — Dana, Office Manager
The privacy line — important: notice the reply never confirms they're a patient and never repeats the dollar figures or discusses their treatment. Under Ontario's PHIPA, defending yourself in public with a patient's billing or visit details can itself be a reportable privacy breach — even when the reviewer brought it up first. The move is always: validate the feeling, take the facts offline.
Change a clause or two so it doesn't read templated, sign it with a real name and role, and aim to post within 48 hours. Want me to do a softer or firmer version?
(Source: Reputation response playbook — billing-surprise draft, four-move craft model, and the PHIPA guardrail.)
Sample exchanges generated against the deployed copilot's system prompt and the six-file audit corpus, demonstrating the concise-first, prioritized, source-cited answer style and the compliance/privacy/no-overpromising guardrails. Reviews and replies are fictional composites; drafts are paste-ready starting points, not legal advice or guaranteed outcomes.