An AI search engagement needs a defined scope, initial measurements, an action plan, and a review process. This chapter describes a structure that an agency or internal team can adapt.Research sources and limitations. Identifies the available report data and the analysis still needed to support earlier draft figures.
Agree on the customer questions and success measures before producing deliverables. Record what the available tools can and cannot establish.
Onboarding and discovery
Document products, services, customer groups, competitors, supported claims, and publication constraints. Ask which business decisions the engagement should support. Keep assumptions separate from verified facts.
Build the prompt set and record repeated initial answers. Compare search conditions where supported. Retain the full responses and configuration details so later measurements can be reviewed.
Agree on access, content approval, responsibilities, and the release process. Use those constraints to make the plan executable.
Research and planning
Write the plan so another qualified team can carry it out. Each task needs enough detail to identify the intended change and verification method.
Combine the access and accuracy audit with competitor and source comparisons. Identify confirmed problems, plausible explanations, and unknowns. Use customer value and effort to prioritize them.
List pages to revise or create, factual corrections, technical changes, and relevant publication opportunities. Add a paid test only when it answers a defined business question. Name the owner, dependency, expected deliverable, and check for each task.
Implementation
Resolve dependencies first. Confirmed access problems may need engineering work before content can be retrieved. Content and independent communication can proceed in parallel when they do not depend on the same unresolved facts.
State whether the engagement includes implementation or only research and planning. A handoff should include the evidence, priorities, and verification steps. The First 90 Days provides a sample schedule.
Reporting and the operating cadence
Choose collection and reporting intervals according to the category and decisions. Weekly collection and monthly reporting can be a starting point, but the schedule is not a universal requirement.
Report the measurements, completed work, business outcomes, and next priorities. Include failed requests, sample sizes, and known model or source changes. Do not claim campaign impact solely because a metric changed after publication.
Review the plan periodically. Continue work supported by the evidence, revise unsuccessful assumptions, and document why priorities changed.
Define what the client receives
| Deliverable | Acceptance check |
|---|---|
| Prompt inventory | Exact wording, versions, source of demand, and test groups recorded |
| Initial measurements | Full answers, settings, attempt counts, and scoring rules retained |
| Audit | Each finding includes evidence, affected URLs, and a proposed check |
| Corrections | Approved changes published and checked against business facts |
| Follow-up report | Comparable results, known changes, limitations, and next actions |
Agree on these deliverables before promising a schedule. Engineering access, product facts, and approval times can determine when a correction is possible. A fixed recommendation increase is not a deliverable the team can control.