# AI Visibility Audit Proposal Prepared by [agency legal name] ("the agency") for [client legal name] ("the client") on [date]. Valid until [expiry date]. Have local counsel review contractual language before signature. This template is general commercial language, not legal advice. ## 1. Engagement objective The agency will assess how [audience, for example "US homeowners choosing an installer"] can encounter and understand [client brand] while deciding [the buyer decision in scope] across [declared platforms], in [geography] and [language]. The engagement produces a reproducible baseline of how the brand appears in a declared sample of AI-generated answers, identifies material technical, source, and representation gaps, and prioritizes actions that improve public clarity and buyer usefulness over the next [period]. This engagement does not promise any ranking, mention, citation, recommendation, traffic, lead, or revenue outcome on any platform. ## 2. Scope **Brand and offers in scope.** [One legal or trading name; the named offers; approved locations and service areas; intended audiences. Everything else is out of scope.] **Platforms and surfaces.** [Exact platforms and product surfaces, for example "ChatGPT search, signed out, US location context" and "Google AI Mode, US English". Include account state and location context per platform.] **Prompt panel.** [Count] prompts derived from [the buyer decisions in scope], separated into [strata, for example "problem discovery, category discovery, comparison and fit, validation and risk"], with [count] repeated runs per prompt per platform during [date window]. The panel is frozen before the first saved observation. **Priority pages.** [List of URLs, or the selection rule, for example "the ten pages that carry the material facts for the offers in scope"]. **Competitors.** [Named competitors, or the selection rule written before observation]. Competitor observations use the identical frozen prompt panel and the same denominator. **Technical review.** Public response checks, robots and crawler-access rules, indexability and snippet controls, canonical destinations, rendered-text checks, and structured-versus-visible fact checks for the priority pages. ## 3. Method The agency will: 1. Freeze the prompt panel, platforms, run conditions, and competitor selection rule before the first saved observation. 2. Preserve exact prompt text, full answer evidence, run date and time, platform, surface, and account state for every observation. 3. Classify brand mentions and cited sources separately, and never sum them into one number. 4. Grade every material statement about the brand against a client-approved fact ledger. 5. Label every finding as observed, verified, inferred, or unknown, and keep unknowns visible. Like-for-like trend comparisons use the same declared method version. Any material method change is disclosed and starts a new baseline rather than silently rewriting the old one. ## 4. Client inputs The client will provide, by [date]: 1. One authorized project owner empowered to approve facts and accept deliverables. 2. Written approval of the brand, offers, pages, platforms, and competitors in scope. 3. An approved fact ledger covering names, offers, fit and non-fit conditions, locations, pricing language, credentials, and material claims, each with a source. 4. Access to agreed first-party tools [list, for example search console and analytics] where included. 5. Responses to factual questions within [n] business days. Delayed or conflicting inputs extend the delivery window day for day and may trigger the change-control process in section 8. ## 5. Deliverables - Scope and method record (the reproducibility layer) - Approved fact and entity ledger, reviewed and dated - Technical eligibility record for [count] priority pages with dated evidence - Frozen prompt panel ([count] prompts, versioned) and complete observation log - Mention, citation, representation, and source baseline, with competitor comparison on the same denominator - Prioritized findings, each traced to saved evidence - Client report ([format]) and a [length] handoff session - 90-day action roadmap with owners and acceptance checks per item ## 6. Exclusions Not included unless added in writing through change control: - Any guarantee of ranking, mention, citation, recommendation, sentiment, traffic, leads, or revenue - A census of all prompts, users, models, locations, or answers; the audit reports a declared sample - Access to, or claims about, platform-internal ranking, retrieval, or training systems - Implementation: content production, development, schema changes, profile updates, or outreach - Continuous monitoring after the declared observation window - Additional languages, regions, brands, business units, platforms, or competitor sets - Legal, medical, financial, or regulatory advice ## 7. Acceptance criteria The engagement is complete when: - every declared scope field in section 2 is filled and approved - the frozen prompt panel and method version are preserved and dated - every material finding traces to saved evidence - technical, source-use, representation, and outcome evidence are reported separately - every roadmap item names an owner and an observable completion check - method limits and unknowns are stated in the client-facing report - deliverables open in the agreed formats ## 8. Timeline and change control Planned active delivery window: [start condition, for example "five business days after client inputs in section 4 are complete"] through [target handoff date], subject to the agreed client review periods. Adding prompts, runs, platforms, priority pages, competitors, languages, locations, deliverables, or stakeholder sessions changes the fee and the delivery window. The agency will present the effect in writing before beginning added work. No added work begins on a verbal request. ## 9. Fee and payment Fixed fee: [amount and currency], based on the scope in section 2. Payment schedule: [for example "50 percent on signature, 50 percent on delivery of the client report"]. Additional work: [hourly rate or change-order method]. Late payment: [terms]. ## 10. Responsible-use statement Generative answers vary by platform, model, location, account state, wording, and time. This audit reports a controlled, declared sample and available first-party evidence; it does not measure all AI answers and cannot guarantee how any platform will behave. Recommendations are designed to improve public clarity, technical eligibility, evidence quality, and buyer usefulness even where sampled answers do not change. ## Signatures For [client legal name]: name, title, date, signature. For [agency legal name]: name, title, date, signature.