# GEO Audit Fieldbook

Use this fieldbook to turn a GEO checklist into reproducible evidence and an owned implementation backlog. It is not an AI-platform scorecard and does not predict citations.

## Audit record

| Field | Entry |
|---|---|
| Business / property | |
| Audit owner | |
| Review window | |
| Priority products / services | |
| Markets and languages | |
| Primary conversion | |
| Technical reviewer | |
| Editorial reviewer | |
| Business approver | |

## Evidence labels

Use one label for every finding:

- **Verified:** directly reproduced in a page response, rendered page, platform control, official report, server log, analytics export, or current first-party documentation.
- **Observed:** recorded in a defined prompt sample, but variable and not a platform rule.
- **Inferred:** a plausible explanation supported by indirect evidence.
- **Unknown:** evidence is missing or contradictory; monitor or design a test.

## Part 1: Scope and prompt baseline

| Topic / decision | Priority URL | Conversion | Market | Language | Competitors | Owner |
|---|---|---|---|---|---|---|
| | | | | | | |

### Prompt cohort

| Prompt ID | Exact prompt | Intent | Platform / mode | Market | Language | Session state | Repeats planned |
|---|---|---|---|---|---|---|---|
| P-01 | | Category / problem / comparison / recommendation / implementation / branded | | | | Fresh / signed-in | |

### Observation runs

| Prompt ID | Run date | Run number | Brand mentioned | Owned URL cited | Citation URL | Accuracy | Competitors | Evidence link / capture |
|---|---|---|---|---|---|---|---|---|
| P-01 | | 1 | Yes / No | Yes / No | | Accurate / partial / wrong | | |

Do not silently change prompt wording, platform mode, market, language, or session state between comparable runs.

## Part 2: Google Search eligibility

| URL / property | HTTP status | robots | noindex | canonical | indexed canonical | snippet eligible | AI control | AI report | Evidence |
|---|---|---|---|---|---|---|---|---|---|
| | | | | | | | Include / exclude / inherit | Present / absent / no data | |

If the generative AI report is absent, record whether access, low impressions, or the property control could explain the absence. Do not record an absent report as zero visibility without evidence.

## Part 3: Crawler policy and delivery

| User agent | Job | Intended policy | robots verified | CDN / WAF verified | Test result | Server-log evidence | Retest date |
|---|---|---|---|---|---|---|---|
| OAI-SearchBot | ChatGPT Search crawler | | | | 200 / 403 / 429 / 5xx | | |
| ChatGPT-User | User-initiated fetcher | | N/A / note | | | | |
| GPTBot | Potential training crawler | | | | | | |
| PerplexityBot | Perplexity Search crawler | | | | | | |
| Perplexity-User | User-requested fetcher | | N/A / note | | | | |

## Part 4: Source usefulness

| URL | Primary reader question | Direct answer | Original value | Limits stated | Decision aid | Appropriate next step | Finding |
|---|---|---|---|---|---|---|---|
| | | Pass / fail | Data / method / experience / tool / none | Pass / fail | Table / steps / template / none | | |

## Part 5: Claim, author, and entity ledger

| Claim / fact | URL | Source of truth | Date checked | Scope / conditions | Conflict | Action | Owner | Retest |
|---|---|---|---|---|---|---|---|---|
| | | | | | | Keep / revise / test / remove | | |

Canonical entity values to compare: organization name, legal identity, primary URL, logo, contact details, markets, service and product names, pricing or policies, and named experts with current roles.

## Part 6: Structured and machine-readable surfaces

| Surface | Visible-content match | Valid | Canonical links | Current facts | Special-ranking claim removed | Evidence / action |
|---|---|---|---|---|---|---|
| Organization schema | | | | | N/A | |
| Article / TechArticle schema | | | | | N/A | |
| Breadcrumb schema | | | | | N/A | |
| llms.txt | | N/A | | | | |

## Part 7: Citation and source map

| Prompt ID | Cited source | Source class | Owned / third party | Claim supported | Current | Brand relationship | Gap / opportunity |
|---|---|---|---|---|---|---|---|
| | | Documentation / publisher / directory / review / forum / academic / other | | Yes / partial / no | | | |

## Part 8: Measurement and business outcomes

| Metric | Definition | Baseline | Observation window | Data source | Limitation | Owner |
|---|---|---|---|---|---|---|
| Google generative AI impressions | Links shown in supported Google generative AI features | | | Search Console | Google only; impression metric | |
| Brand mention rate | Prompts with a brand mention / comparable prompt runs | | | Prompt cohort | Variable sample | |
| Owned citation rate | Runs with an owned citation / comparable prompt runs | | | Prompt cohort | Variable sample | |
| Factual accuracy | Accurate uses / observed brand descriptions | | | Review rubric | Human classification | |
| AI referral sessions | Sessions attributed to documented AI referrals | | | Analytics | Attribution limitations | |
| Qualified conversions | Qualified outcomes associated with AI referrals or assisted journeys | | | CRM / analytics | Journey and attribution limitations | |

## Finding register

| ID | Observation | Evidence label | Captured evidence | Consequence | Priority | Recommended action | Owner | Due date | Retest | Status |
|---|---|---|---|---|---|---|---|---|---|---|
| F-01 | | Verified / observed / inferred / unknown | | | P0 / P1 / P2 / monitor | | | | | Open |

Priority guide:

- **P0:** eligibility blocker or material trust risk.
- **P1:** weakness preventing a priority page from becoming a strong source.
- **P2:** clarity or coverage improvement after fundamentals are fixed.
- **Monitor:** insufficient evidence or changing platform behavior.

## 30/60/90-day backlog

| Window | Finding IDs | Deliverable | Owner | Dependency | Success evidence | Retest date | Status |
|---|---|---|---|---|---|---|---|
| Days 1–30 | | Eligibility and trust blockers | | | | | |
| Days 31–60 | | Priority source and entity repairs | | | | | |
| Days 61–90 | | Citation gaps, measurement, and experiments | | | | | |

## Executive summary

### Directly verified


### Observed in the prompt sample


### Inferred but not proven


### Unknown or blocked by missing evidence


### Decision

Scale / iterate / stop / re-audit after prerequisites

### Next review date and owner