Why this matters
Chatbot answers need proof, not vibes
Many small businesses add chat widgets, AI assistants, or FAQ answer blocks before they have a clean source of truth. The result is risky: the bot may make up hours, discounts, service-area coverage, warranty terms, refund rules, emergency availability, or next steps that the business cannot honor.
This worksheet gives owners, marketers, and operators a simple way to approve answer cards before they are reused by website chat, AI-search pages, or internal AI drafting workflows.
CHATBOT-ANSWER-SOURCE-READY
Copy/paste artifact
Chatbot answer source-of-truth card
Question / intent:
Approved short answer:
Approved long answer:
Source URL or document:
Last reviewed date:
Owner / reviewer:
Allowed channels: website chatbot / FAQ page / sales email / support reply / AI-search page / internal only
Escalate when:
Do NOT answer when:
Required disclaimer or safe wording:
Related internal page:
Related public page:
CRM/support label if a lead asks this:
Next review date:Answer categories
Start with the answers that can hurt revenue or trust
| Answer type | Source to verify | Stop rule |
|---|---|---|
| Pricing, fees, discounts | Current pricing page, quote policy, owner-approved offer sheet. | STOP PUBLISHING if price, discount, tax, financing, or deposit details are not current. |
| Service area and availability | Service-area page, dispatch calendar, branch/location rules. | STOP PUBLISHING if the answer promises coverage, arrival windows, or appointment openings without a live source. |
| Warranty, refund, guarantee | Policy page, contract, warranty document, manager approval. | STOP PUBLISHING if the bot might promise coverage, refunds, free service, or legal terms. |
| Emergency or safety questions | Emergency policy, licensed/professional guidance path, escalation owner. | STOP AUTOMATION and route to a human when safety, medical, electrical, water damage, lockout, or urgent conditions appear. |
| Eligibility or best fit | Package/service-fit page, intake rules, not-a-fit wording. | STOP PUBLISHING if qualification criteria are incomplete or could create discrimination, compliance, or false-promise risk. |
AI audit prompt
Check chatbot answers before publishing
You are auditing website chatbot / AI-search answers for a small business.
Use only the source cards below. Do not invent pricing, hours, service area, warranty, refund, eligibility, emergency response, credentials, discounts, or availability.
For each draft answer, return:
1. approved as-is / needs edit / do not publish,
2. unsupported claims,
3. missing source URL,
4. escalation or human-review trigger,
5. safer revised answer using only approved facts.
Source cards:
[PASTE SOURCE CARDS]
Draft answers:
[PASTE DRAFT ANSWERS]Human review
STOP PUBLISHING guardrails
- Do not publish chatbot answers that create new prices, guarantees, discounts, service-area claims, credentials, certifications, or availability promises.
- Do not let a bot give emergency, safety, diagnosis, legal, tax, medical, or financial advice. Route to a human or appropriate emergency/professional source.
- Do not summarize old policy pages unless the owner has reviewed the source date.
- Do not expose private customer examples, internal notes, or confidential vendor/pricing terms in public answers.
Next step
Build a safer AI-search answer system
Pair this worksheet with the AI Search Source Link Map, Support Policy Checklist, and the free AI search visibility audit.
For the full website/AI-search improvement system, see SEO After AI.