How should you evaluate a GEO supplier with a small pilot?

Use the pilot to make a procurement decision

A small GEO pilot should answer a decision the buyer can act on: whether the supplier can understand the organisation, produce accurate and reviewable work, operate the agreed controls, and support a proportionate next phase. It is not a short campaign that must prove a ranking gain or secure an AI citation within an arbitrary window.

Define the decision before the work. GOV.UK testing guidance recommends planning pilots early, tying objectives to the wider project, using clear success criteria, making the test representative and setting an objective evaluation method. Applied to GEO procurement, that means a bounded test with named inputs, outputs, evidence and decision rights—not an open-ended trial or a disguised full rollout.

Write one six-part pilot plan before starting

The buyer and supplier should complete the plan together. Keep the sample small enough to review properly, but broad enough to expose the material risks in the intended engagement. Every row should have an owner, due date and agreed record location.

Choose a representative question and page sample

Start from real procurement or stakeholder questions rather than invented high-volume keywords. Include at least one straightforward question, one comparison or qualification question, and one question where a missing caveat could create a material error. For a bilingual or multi-market engagement, include the languages and local facts that the future scope must actually govern.

Map those questions to a named content and technical sample. Depending on scope, that may include an organisation page, one service or course page, an evidence page, a contact or conversion path, structured data, language alternatives and a small set of crawl or index-readiness checks. Record why each item was selected; a supplier should not be able to replace a difficult agreed item with an easier showcase after the baseline is captured.

• Use current, buyer-approved facts and identify the factual owner for each material claim.
• Freeze a dated baseline before recommendations or edits begin.
• State which changes the supplier may implement and which remain recommendations for the buyer.
• List unavailable access, missing data and buyer dependencies instead of treating them as silent supplier failures.
• Avoid drawing programme-wide conclusions from a sample that excludes the hardest markets, languages or page types.

Score evidence and delivery the supplier can control

Pre-agree the scoring method. Useful criteria may cover factual accuracy, source traceability, completeness against the brief, technical validity, clarity of prioritisation, bilingual consistency, change control, correction handling and handover quality. Define the evidence required for each score and what counts as a material failure—for example an unsupported institutional claim, an unauthorised production change or an unrecorded use of sensitive data.

Report external outcomes separately. The pilot can record sampled AI answers, citations, crawling, indexation, rankings, traffic or enquiries using a dated method, but these are observations affected by external platforms, users, competitors, the buyer and time. Do not convert a short-window increase into a guaranteed supplier effect, and do not treat a lack of movement as proof that usable delivery evidence is absent.

End with a documented go, modify, retest or stop decision

Hold a decision review using the agreed evidence pack. Record the objective, sample, completed and missing evidence, scores, material exceptions, unresolved risks, buyer dependencies and reviewer conflicts. The decision should be one of four explicit outcomes: scale the work, proceed with conditions or a changed scope, run a targeted retest, or stop and close the pilot.

A successful pilot still needs a new or confirmed scope, governance rhythm, service levels, responsibilities, access model, measurement plan and commercial approval before wider delivery. An unsuccessful pilot still needs orderly handover: return or remove access, identify retained and deleted data, deliver buyer-owned records, close actions and state what may or may not be reused. GOV.UK's current Sourcing Playbook likewise places piloting within an evidence-based delivery-model and commercial decision rather than treating it as automatic authority to scale.

How SALIENS supports a bounded GEO pilot

SALIENS provides GEO audits, AI visibility monitoring, content and technical improvements, and ongoing GEO delivery. A scoped pilot can include a bilingual requirements brief, dated baseline, representative question and page sample, source-linked audit or improvement outputs, approval records, an evidence pack and a closing decision review. The exact sample, access and deliverables should be agreed before work begins.

Email info@oxfordintelligence.co.uk with your website, markets and languages, the decision the pilot must support, known risks, available access and preferred review date. SALIENS is operated by Oxford Intelligence Limited. A pilot can test delivery quality and produce decision evidence; it does not guarantee crawling, indexation, AI mentions, citations, rankings, traffic or enquiries. This guide is a procurement framework, not legal or commercial advice.