A clear method for a changing search landscape.

Start with intent, not a flattering prompt

Use customer questions, sales conversations, service context, and approved expertise to build a question map. Separate discovery, comparison, trust, and readiness-to-inquire questions. Treat unverified question ideas as hypotheses, not measured search demand.

• Unbranded questions examine category discovery and alternatives.
• Branded questions examine what a system says when your business is named.
• Use sensible variations and record their wording. Do not substitute a favorable variation without disclosing it.

Define the observation scope

Agree platforms, product experiences, markets, language, question sample, repetition, and review cadence. Search-enabled answers and other model responses are different testing conditions. API outputs and consumer interfaces should not be treated as interchangeable.

• Record date, time, available model/product label, search mode, market assumptions, and conversation context.
• Keep raw answer evidence and supplied citations within the agreed retention and access policy.
• The free diagnostic is deliberately narrower than a full baseline.

Inspect sources and verify facts

Read visible citations where supplied. Compare claims with current business facts approved by the client. Classify evidence as brand-controlled, independent publisher, public profile, user-generated, or another relevant category without implying that every influence on an answer is known.

• Observed fact: an answer contains a claim or links to a URL.
• Verified accuracy finding: the claim agrees or conflicts with approved evidence.
• Hypothesis: a particular information gap may contribute to an incomplete answer.
• Unknown: any internal model influence or source not visible in the observation.

Prioritize a feasible next action

Rank work by relevance to real buyer decisions, seriousness of inaccuracies, quality of available evidence, control over implementation, and effort. Explain the reasoning. A high-priority item is not a prediction of guaranteed visibility gain.

• Resolve meaningful factual errors and access problems.
• Clarify services, markets, suitability, and differentiation.
• Cover unanswered comparison, process, and trust questions.
• Use applicable structured data only for true visible information.

Approve, implement, and retain the change record

Confirm scope, factual input, permissions, and approval responsibility. Record what changed, where it changed, and who approved it. Check publication and accessibility before interpreting any follow-up results.

Repeat and explain the limits

Use comparable questions and conditions wherever feasible. Disclose sample changes, platform changes, missing observations, and variation. A mention is not a linked citation, a citation is not a recommendation, and none independently proves a qualified lead or sale.