The screenshot is the sales hook; the evidence ledger is the product. Agencies that sell unexplained scores will create reporting debt faster than recurring revenue.
What you should leave with
- Standardize collection and classification before white-labeling the report.
- Separate observed output, external evidence, and agency interpretation.
- Sell a bounded repair sprint before an indefinite monitoring retainer.
- Make every monthly task traceable to a prompt, claim, source, and retest.
What should the agency actually sell?
Sell a diagnostic baseline, a prioritized repair sprint, and optional monitoring as distinct products with explicit outputs and boundaries.
The baseline should answer whether the client is recommended, which competitors win, why they win, which sources support the answer, what is factually wrong, and what to repair. A one-time audit is easier to buy than an open-ended AEO promise and gives both parties evidence for deciding whether monthly work is justified.
The repair sprint should define the pages, profiles, schema, source corrections, approvals, and retest window. Monitoring becomes valuable after the client has a stable prompt set and enough repair activity to observe. Otherwise the agency charges to watch a flat graph.
| Offer | Client receives | Boundary |
|---|---|---|
| Baseline | Prompt ledger, score, source map, errors | Sampled answers, not total market share |
| Repair sprint | Owned tasks and implemented assets | External publication is not guaranteed |
| Monitoring | Changes, source gains/losses, retests | Model variation remains |
| White label | Brandable report and methods appendix | Agency owns client claims and approvals |
Evidence used in this section
How should prompts and scoring be governed?
Version the prompt set, define inclusion rules before collection, retain raw answers, and disclose weighting and confidence in every report.
Use the client's sales language, search data, win-loss notes, and service boundaries to draft prompt families. Freeze the baseline before looking at the results. If prompts are edited only when the client loses, the trend becomes a moving target. Record platform, date, geography, answer mode, exact wording, and source set.
Create a short classification manual for recommended, compared, cited, passing mention, absent, and factually wrong. Human-review ambiguous results and high-impact errors. A client should be able to trace a score movement to specific rows rather than trusting a chart generated by proprietary weighting.
- Prompt version and business rationale
- Raw answer and visible citations
- Brand alias and competitor normalization
- Human review for ambiguous or harmful claims
- Published formula, sample limits, and confidence
Evidence used in this section
What belongs in a client-ready report?
Lead with the commercial gap, then show the raw evidence, repeated recommendation reasons, missing source patterns, accuracy risks, and a sequenced work plan.
Executives need the size and consequence of the gap; operators need the underlying rows. Show recommendation coverage by prompt family and platform, competitor share, repeated source domains, and a handful of answer excerpts with dates. Distinguish platform output from the agency's inference and recommendation.
The plan should state owner, asset, evidence needed, dependency, expected impact, and retest date. Avoid filler such as 'build authority.' Name the comparison page, directory conflict, missing service fact, FAQ, case proof, or source relationship. Include what will not be done and why.
Evidence used in this section
How should monthly AEO delivery work?
Run a fixed operating cycle: detect a repeated gap, validate the source and business truth, implement the smallest credible fix, verify publication, and retest.
A monthly queue should not be filled from a generic content calendar. Group losses by cause: missing page, weak evidence, external inconsistency, factual error, crawl problem, or uncertain measurement. Work the highest-value repeated cause and preserve before-and-after evidence.
Coordinate approvals early. Legal, product, clinical, and local clients may need subject-matter review. Third-party corrections can take longer than owned-page changes, so report their status separately. When no credible fix exists, say so rather than manufacturing activity.
- STEP 1
Diagnose
Confirm the pattern across prompts, runs, platforms, and sources.
- STEP 2
Substantiate
Get the client's current fact and proof before writing or markup.
- STEP 3
Implement
Change the smallest owned or controllable asset that addresses the cause.
- STEP 4
Verify and retest
Check the live page, crawlability, schema, source, and answer outcome.
Evidence used in this section
How can agencies use the audit for prospecting without losing trust?
Use a narrow, reproducible finding in outreach and give the prospect enough evidence to verify it before asking for a sale.
A strong message states the tested prompt count, date, platforms, brand appearances, competitor appearances, and one repeated source gap. Attach a clean screenshot or preview link, but include the sampling caveat. That is more credible than claiming the prospect is 'invisible everywhere.'
Do not fabricate prompts around a predetermined failure or publish sensitive competitive reports without care. The free audit should demonstrate the diagnostic method and create a useful next step. The paid offer earns its price through deeper validation, implementation, and retesting, not by hiding every underlying fact behind a form.
Method boundary: A sampled result can start a sales conversation, but it should never be presented as exhaustive demand data or guaranteed lost revenue.
Questions that change the decision
Frequently asked questions
Can an agency white-label the methodology?
It can brand the client experience, but should preserve the scoring formula, evidence definitions, sample limits, and source disclosure. Hiding method details makes client review and trend interpretation weaker.
How often should client prompts run?
Frequency should match decision value and budget. Monthly can be enough for many local or mid-market programs; high-value launches or volatile categories may justify more frequent repeated runs.
Should AI visibility replace rank tracking?
No. Search rankings, Search Console, analytics, crawler logs, and answer audits observe different parts of discovery. The agency should connect them without pretending they are the same metric.
What makes the service retainable?
A clear queue of evidence repairs, publication work, external corrections, and controlled retests creates ongoing value. A dashboard alone is vulnerable once the client has seen the initial graph.
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: AI features and your websiteGoogle says AI features build on normal Search eligibility and may use query fan-out, so crawlability and specific supporting pages still matter.
- [2]Google Search Central: people-first contentGoogle asks whether content demonstrates first-hand expertise, original analysis, clear authorship, and a satisfying answer for the intended audience.
- [3]OpenAI: Introducing ChatGPT searchOpenAI describes answers with links to web sources, making the source set and freshness of public facts relevant to recommendation audits.
- [4]Google structured data policiesStructured data must describe visible, representative content and cannot substitute for trustworthy evidence on the page.