What a GEO audit should deliver before ongoing work

Buy a decision-ready baseline, not a visibility score alone

Before committing to ongoing GEO work, your team needs to know what was checked, what the evidence shows and what to do next. A headline score cannot answer those questions without its underlying records. Use the checklist below to discuss acceptance criteria with a prospective supplier; it is a suggested procurement framework, not a universal industry standard or an addition to any existing contract.

The practical test is simple: could a colleague who did not attend the presentation trace a finding back to its evidence and assign the next action? If not, ask for the missing context before treating the report as a basis for further investment.

1. An agreed scope and a reusable question register

Ask for the brand, domain, audience, market and language covered, together with the actual questions used. Separate questions that name your brand from discovery questions that do not. For example, an audit of an institutional service might ask how a buyer would compare providers before it tests what an answer says about a named supplier. These are different tests and should not be combined without explanation.

Record the agreed engines and interfaces, observation dates, search settings where available, rounds and relevant session conditions. Mark settings the supplier cannot observe as unknown. Keep a fixed core of questions for later comparison and log any changes. An API result should be labelled as such, not presented as an observation from a consumer interface.

• Is the full question list included, with branded and non-branded questions labelled?
• Can you see the planned coverage and the checks actually completed?
• Are unavailable platforms and unobserved settings clearly identified?

2. Inspectable answers with separate outcome fields

Request dated answer records and the source URLs shown, where retention is permitted. Record brand mentions separately from linked citations to your own domain; keep third-party citations separate too. Check important descriptions against approved company facts, because a mention can still describe the wrong service or market.

Each percentage needs an explicit denominator. A completed answer with no mention is different from an unsuccessful request or a check that never ran. As an illustrative example, if 24 of 30 planned checks complete and six completed answers mention the brand, the observed mention rate is 6/24, while completion is 24/30. Neither figure describes the missing six answers. This is a calculation example, not a SALIENS client result.

3. Page-level findings, not a generic technical checklist

For each important website finding, ask for the affected URL, the observed issue, its evidence and the proposed next action. Distinguish a reachable page from a page confirmed as indexed; if the supplier lacks access to search-console evidence, the report should say indexing is unverified.

Google's AI Search guidance says its existing SEO foundations apply to AI Overviews and AI Mode. Supporting pages must be indexed and eligible for a search snippet; there is no special AI file or schema requirement, and eligibility does not guarantee inclusion. A useful audit therefore examines relevant crawl access, textual information, internal links and consistency between structured data and visible content. The Google source below was checked on 13 September 2026.

4. A prioritised action register and a clear handover

Ask the supplier to connect each recommendation to a buyer problem: unclear service scope, inconsistent facts or a technical access issue, for example. An action register should identify the supporting finding, affected page, priority rationale, proposed owner, dependencies and how completion will be checked. Effort estimates and implementation responsibility should be discussed explicitly.

Acceptance should concern evidence and agreed deliverables, not a promised ranking. Ask what the audit does not include, who approves factual corrections and whether your own team must implement them. A later comparison also needs the original question set and conditions; a differently configured test cannot by itself establish improvement caused by the work.

• Scope: can the report be reconciled with the written brief?
• Evidence: can each material finding be traced to an answer or page?
• Coverage: are completed, failed and unrun checks separated?
• Actions: does each priority have a reason, owner and completion check?
• Handover: are exclusions, access needs and the next decision clear?

How SALIENS' Baseline Audit fits

SALIENS helps brands become discoverable and accurately represented in search and AI-generated answers, giving potential customers the information they need to make informed enquiries. Its current Baseline Audit covers one brand, one primary domain and one market-language profile, with 30 prompts across three engines and three rounds, three competitors and up to 20 pages. The listed outputs include a baseline report, priority plan and one 60-minute review. Content production and technical implementation are not included; final scope is confirmed in the written order.

To discuss an audit, email info@oxfordintelligence.co.uk with your website, target market and language, and the buying questions you want to understand. Say whether you need diagnosis only or also have an implementation gap. SALIENS also offers monitoring, content and technical improvement, and ongoing GEO delivery under separately agreed scopes. The linked service guide explains the options. No ranking, AI citation or enquiry outcome is guaranteed.