Our position: weak onboarding creates synthetic prompts that measure the agency's imagination instead of the client's market.
What you should leave with
- Interview sales before finalizing prompts.
- Resolve entity aliases and locations early.
- Document claims the client can substantiate.
- Approve scope and classification policy together.
What business information is required before testing?
Collect the products, services, regions, valuable customer segments, disqualifiers, sales objections, and decisions the client wants to influence. Ask which outcomes matter most and which claims can be supported publicly.
A product list alone does not describe a buying decision. Ask what causes a prospect to choose, delay, or reject the client: security, speed, specialty, geography, price model, integration, credentials, or risk. Those constraints become the backbone of high-intent prompts.
Document exclusions as carefully as targets. A law firm may not want low-value case types; a SaaS company may not support a required integration. The audit should not reward recommendations to buyers the company cannot serve.
- Priority offers and margins
- Buyer segments and constraints
- Service regions and exclusions
- Claims with supporting evidence
Evidence used in this section
Which entity and competitor details prevent bad results?
Record the legal name, public brand, product names, former names, domains, profiles, locations, and common misspellings. Approve a direct competitor set and a separate list of substitutes that buyers may consider.
Entity ambiguity can turn a clean string match into a false positive. Capture parent-child relationships, product brands, franchises, and similarly named businesses. Define whether a product mention counts for the company and how location-specific recommendations are classified.
Do not let the client choose competitors only by aspiration. Include companies that repeatedly appear in sales conversations, search results, directories, reviews, and early prompt sampling. Keep the approved core stable while allowing an ‘emerging competitors’ field for discovery.
- Canonical names and aliases
- Owned domains and profiles
- Direct competitors and substitutes
- Entity-match exceptions
Evidence used in this section
How should the onboarding workshop run?
Run a 60-minute workshop with sales, marketing, and a subject-matter expert; draft prompt families live; then return a versioned scope for approval. Resolve access, dependencies, and legal review before collection begins.
Use customer questions, call notes, reviews, and search data as raw material. Convert them into decisions with a buyer, category, constraint, and desired outcome. Keep prompts natural; do not stuff the brand name into tests meant to measure unprompted recommendation visibility.
End with a responsibility map. The agency needs owners for factual approval, technical changes, content, directory profiles, and third-party outreach. A strong diagnosis is wasted when no one can authorize the repair.
- STEP 1
Interview
Capture buyer language, objections, constraints, value, and known misinformation.
- STEP 2
Map
Resolve entities, offers, locations, competitors, substitutes, and evidence assets.
- STEP 3
Draft
Build prompt families and assign commercial value before seeing results.
- STEP 4
Approve
Sign off scope, policy, access, owners, dependencies, and delivery dates.
Evidence used in this section
What does onboarding completion look like?
Onboarding is complete when the audit can run without guessing: scope is approved, prompt families cover priority decisions, entity rules are written, competitors are defensible, access works, and reviewers are scheduled.
Use a readiness gate instead of starting on the calendar date regardless of missing inputs. Mark each dependency as ready, accepted risk, or blocked. If the client cannot validate product facts, report that limitation rather than inventing a clean baseline.
Preserve the onboarding snapshot with the audit version. When products, regions, or positioning change later, the team can explain why prompts changed and avoid treating a business-model change as a visibility trend.
| Gate | Ready when | If missing |
|---|---|---|
| Market model | Segments, decisions, and exclusions approved | Reduce scope |
| Entity policy | Aliases and ownership rules documented | Flag low confidence |
| Execution access | Owners and systems available | Separate audit from fixes |
Evidence used in this section
What should onboarding avoid?
Avoid copying a universal prompt list, accepting unsupported positioning claims, using only marketing's competitor list, and beginning collection before entity rules are settled. These shortcuts create polished but irrelevant findings.
Do not ask clients to predict model-ranking factors. Ask for evidence: which pages substantiate the claim, which third parties validate it, and where customers already compare options. The agency's job is to test the public evidence environment, not translate speculation into a strategy.
Keep the questionnaire short enough to finish and use the workshop for nuance. Forty mandatory free-text fields often produce rushed answers; a focused form plus a skilled conversation creates better prompt inputs.
- No generic prompt import
- No unverified superlatives
- No hidden entity assumptions
- No testing before scope approval
Method boundary: Sensitive customer or sales information should be minimized. Collect only what is necessary to design the test and apply the client's data-handling policy.
Evidence used in this section
Questions that change the decision
Frequently asked questions
Who should attend the onboarding call?
Include a marketing owner, someone close to sales or customers, and a subject-matter expert who can approve claims. Add technical or local-operations owners when fixes will depend on them.
Should the client write the prompts?
The client should provide buyer language and validate relevance. The agency should normalize coverage, remove leading wording, and maintain a versioned prompt design.
Do we need analytics access before the audit?
It is helpful for prioritization and later attribution but not required to observe AI answers. Record missing access as a limitation and avoid unsupported business-impact claims.
How long should onboarding take?
A focused single-market client can often complete it in one workshop plus asynchronous approval. Complex locations, products, languages, or compliance review need a longer readiness phase.
Primary sources and research
Platform documentation supports factual statements. Where we describe an audit method or prioritization rule, that is AnswerMentions' operating judgment and is labeled as such.
- [1]Google Search Central: creating helpful, reliable contentGoogle recommends original information, substantial analysis, clear sourcing, and content that leaves a visitor feeling they learned enough to achieve the goal.
- [2]FTC: advertising and marketing basicsThe FTC states that advertising claims must be truthful, non-deceptive, and supported by evidence when appropriate.
- [3]OpenAI Help: accuracy and citationsOpenAI warns that ChatGPT can produce incorrect facts and fabricated references, so consequential claims should be checked against reliable sources.
- [4]Perplexity Help Center: how sources workPerplexity explains that it searches the web, identifies sources, and synthesizes an answer with citations, making source inspection central to evaluation.
- [5]NIST: AI Risk Management FrameworkNIST frames AI risk work around governance, mapping, measurement, and management, a useful model for separating observations from decisions.
- [6]Aggarwal et al.: Generative Engine OptimizationThe KDD 2024 paper evaluates generative-engine visibility in a controlled benchmark; it is evidence that visibility can be studied, not a universal ranking recipe.