How to choose an AI SEO agency: 12 questions to ask before you sign

The AI SEO and GEO market contains agencies, consultants, software platforms and traditional SEO providers using overlapping language. The safest way to choose is to ignore the category label for a moment and inspect the operating method.

These twelve questions expose whether the provider can move from a buyer problem to evidence, implementation and accountable reporting.

1. Which commercial decision will the work support?

“Improve AI visibility” is not a decision. A better brief names the market, offer, customer questions and business outcome that justify the work. Ask the agency to state the decision in one sentence.

2. Which search and AI surfaces are in scope?

The proposal should name each surface rather than saying “all major platforms.” Search results, map surfaces, AI answers and citation-led engines behave differently. Scope should also name the language, market and relevant location conditions.

3. How is the query panel built?

A fixed number of prompts is not automatically rigorous. Ask how the provider covers awareness, comparison, problem, service, brand and decision questions—and how it prevents the panel from changing whenever the result is inconvenient.

4. What is recorded at baseline?

Presence alone is too thin. A useful baseline can include brand mentions, citations, shortlist position, narrative accuracy, competitors and sources. The provider should date the observation and preserve enough detail to repeat it.

5. How are observations separated from causes?

If a citation changes after a page edit, that sequence does not prove the edit caused the change. Ask how the agency labels observations, hypotheses and confidence. Calm uncertainty is a quality signal.

6. What happens after the audit?

Clarify whether the agency produces recommendations, implementation-ready tickets or actual production changes. If implementation is included, the scope should name access requirements, owners, dependencies and acceptance criteria.

7. Which technical areas can the team implement?

Ask for specific capabilities: crawlability, indexation, rendering, structured data, internal linking, information architecture, analytics or migrations. “Technical GEO” without production competence can leave you with a document nobody can ship.

8. How does content work connect to buyer questions?

The answer should go beyond publishing volume. Look for content-gap mapping, extractable answers, supported claims, internal linking, service-page clarity and evidence. More articles are not automatically the remedy.

9. How are tools used and verified?

Tools can accelerate collection and monitoring. Ask what is automated, what is manually reviewed and what the tool cannot observe reliably. A dashboard should support the method, not become the method. Use our comparison of AI SEO tools and implementation agencies to identify which operating capability the proposal is actually selling.

10. What will the report show?

A good report separates shipped work, technical verification, search and AI observations, commercial outcomes and open dependencies. Ask to see the report structure with client information removed, or use the SEO and AI Search reporting workbook as a neutral comparison frame.

11. What is explicitly outside scope?

Boundaries protect both parties. Typical exclusions might include digital PR, web development, multilingual delivery, engineering access, ongoing monitoring or content production. If the proposal contains no exclusions, the delivery assumptions are probably unresolved. The SEO agency contract scope checklist lists the operational boundaries that should be explicit before legal review.

12. What claims will the agency refuse to make?

The provider should be comfortable refusing guaranteed rankings, citations, recommendations and universal timelines. Search and AI systems are external systems; the agency controls the quality of its work, not the final inclusion decision.

A practical comparison method

Score each proposal on five areas: decision clarity, reproducibility, implementation depth, reporting discipline and boundaries. Weight implementation more heavily when your team lacks internal capacity. Weight enablement more heavily when your team will ship the roadmap.

Compare like with like. A software subscription, a strategy audit and an implementation engagement solve different jobs even when all three use the phrase “AI visibility.” Our own services architecture separates Diagnose, Build and Operate for this reason, and our pricing publishes the boundaries of each starting point.