# AI Visibility Report: [Client] / [Period]

Prepared by: [Agency]
Method version: [v1.0, date the method was frozen]
Observation window: [start date] to [end date]
Report date: [date]

Replace every bracketed placeholder. Delete every guidance line (marked with >)
before sending. Nothing in this template is a real observation.

---

## 1. Decision and scope

> One paragraph. If the client cannot say what decision this report supports,
> the report will be read as a score and argued with instead of used.

- Business decision this report supports: [decision]
- Audience and buying context: [who, deciding what]
- Geography and language: [geo, language]
- Platforms observed: [ChatGPT, Google AI features, Copilot, ...]
- Prompt panel: [count] prompts, frozen on [date], in strata:
  [n] unbranded discovery / [n] comparison / [n] branded validation
- Repeated runs: [n] per prompt per platform
- Priority pages: [list or link to inventory]
- Exclusions: [what was deliberately out of scope and why]

## 2. Executive finding

> One page maximum. Three questions, in order: what happened, how do we know,
> what should happen next. Every number carries its denominator.

What we observed:
[1 to 3 findings, each with numerator, denominator, platform, and stratum.]

How we know:
[Method in two sentences, pointing at section 3 and the appendix.]

Recommended next action:
[Smallest useful action, owner, acceptance check, review trigger.]

## 3. Method and limits

> This section is the product. Do not shrink it into a footnote.

- This report is a controlled sample of observed answers, not a census of all
  AI answers or users.
- Prompts were selected from [buyer decisions / discovery interviews / search
  evidence], not from phrases known to favor the client.
- Account state, location, and personalization during observation: [state].
- Known limits this period: [model changes, platform changes, invalid rows].
- This report does not establish a platform ranking and does not predict
  future mentions, citations, traffic, or revenue.

## 4. Technical eligibility

> First-party checks on priority pages. Keep these separate from answer
> observations: a page can pass every check and still never be cited.

| Priority URL | Public response | Robots and index controls       | Canonical      | Facts in rendered text    | Checked on |
| ------------ | --------------- | ------------------------------- | -------------- | ------------------------- | ---------- |
| [URL]        | [200 / other]   | [allowed / blocked, which rule] | [self / other] | [pass / fail, which fact] | [date]     |

Earliest unresolved blocker: [none / description with evidence reference].

## 5. Mentions and citations

> Report each stratum and platform separately. Never merge branded and
> unbranded prompts into one rate.

| Platform   | Stratum   | Valid observations | Brand mentions | Owned citations | Mention rate | Owned citation rate |
| ---------- | --------- | ------------------ | -------------- | --------------- | ------------ | ------------------- |
| [platform] | [stratum] | [n]                | [n]            | [n]             | [x/n]        | [x/n]               |

Most-cited owned URLs: [list with counts]
Most-cited third-party URLs about the brand: [list with counts]

## 6. Representation accuracy

> Compare answer claims against the approved fact ledger dated [date].
> Classify every material brand statement; do not editorialize.

| Claim in answer | Fact ledger says | Label                                                          | Severity           | Evidence row     |
| --------------- | ---------------- | -------------------------------------------------------------- | ------------------ | ---------------- |
| [claim]         | [approved fact]  | [accurate / incomplete / outdated / conflicting / unsupported] | [material / minor] | [observation_id] |

Representation error rate: [material mentions with an error] / [material
mentions reviewed].

## 7. Sources and competitors

- Owned sources that appeared: [list]
- Third-party sources that carried the brand's facts: [list, with freshness]
- Sources answering the same prompts without the brand: [list]
- Competitors named, by stratum: [competitor: n of N observations]

## 8. Findings and priorities

> One block per finding. The chain below is what makes a finding auditable.
> Never let a recommendation appear without an observation above it.

Finding [n]: [one-sentence title]

- Observation: [what was recorded, with denominator and evidence rows]
- Interpretation: [what we think it means, labeled inferred where the
  platform does not expose the mechanism]
- Buyer consequence: [why a buyer would care]
- Recommendation: [smallest useful action]
- Owner: [who] / Confidence: [high / medium / low]
- Acceptance check: [what will be re-observed, and when]

## 9. Roadmap

> Sequence by dependency, not by effort. Technical blockers precede content
> work. Factual conflicts precede anything that amplifies them.

| Order | Action   | Depends on  | Owner   | Review trigger  |
| ----- | -------- | ----------- | ------- | --------------- |
| 1     | [action] | [finding n] | [owner] | [date or event] |

## 10. Appendix

- Prompt panel: [IDs and full text, or link]
- Observation log: [link to the evidence workbook]
- Fact ledger version: [date]
- Definitions: mention, citation, valid observation, material error
- Change log: [method or scope changes since the last report]