An SEO company should be evaluated before a severe traffic decline forces an emergency review. Rankings, traffic, and conversions matter, but they are lagging outcomes that can be influenced by seasonality, brand demand, product changes, prior work, competition, and search updates.
The more reliable evaluation begins with governance: what problem was selected, what work was approved, who completed it, what changed, what evidence supports it, and what result will determine the next decision.
A company can hold strong rankings while remaining one algorithm update away from total invisibility only as a risk scenario, not a guaranteed outcome. Likewise, an agency should not be rejected merely because it does not use fashionable terms such as entity architecture or AI search visibility.
It should be able to explain how the site represents the business, answers important queries, earns legitimate references, manages technical access, complies with applicable rules, and measures commercial value.
The source refers to a playbook from five years ago; preserve that period as a warning against stale assumptions rather than proof that any specific method is obsolete. This guide provides one operating audit with six workstreams: deliverable evidence, page and topic planning, authority evidence, compliance, Google AI and other AI search observation, and technical prioritization. The output is a keep, correct, renegotiate, or replace decision supported by records rather than frustration.
Key Takeaways
- 1A polished monthly report is insufficient when the agency cannot connect metrics to completed work, decisions, owners, and next actions.
- 2Entity and topic planning should clarify the business, services, people, locations, and user questions without relying on an invented protocol.
- 3E-E-A-T work should document accurate experience, expertise, authorship, sourcing, policies, and external evidence rather than promise AI recognition.
- 4Every deliverable should map to a defined user decision and a measurable business or search objective within the organic performance audit.
- 5High rankings can coexist with weak governance, poor intent fit, fragile links, or unmanaged compliance risk.
- 6Distinguish high-volume content production from a maintained editorial system with evidence, review, and clear page purpose.
- 7In law, finance, healthcare, and other regulated sectors, the agency must work within the client's current legal, privacy, and advertising requirements.
- 8Technical SEO should produce a prioritized remediation backlog with evidence and impact, not a large audit used to avoid strategic accountability.
1Can the Agency Prove What It Delivered?
The first audit is a reconciliation between contracted scope, planned work, completed work, and reported outcomes. Impressions, clicks, rankings, and conversions describe performance; they do not prove that the agency caused it.
A report with zero explanation of completed actions is incomplete, but a detailed activity log is also not proof of value. Request the work items, approval status, affected URLs, release dates, owners, source files, quality checks, and unresolved blockers.
For content, review briefs, drafts, reviewer comments, sources, publication records, and internal links. For technical work, review issue evidence, change tickets, before-and-after tests, deployments, and regressions.
For links or digital PR, review placements, source quality, disclosures, destination URLs, and commercial terms. Do not demand a one-to-one relationship between every billed hour and a visible artifact when analysis, coordination, or monitoring is legitimate; require enough documentation to verify the work and its purpose.
Schema changes should be assessed for accuracy and visible-content consistency, not as an authority enhancement by default. In healthcare, legal, or financial work, the record should also identify who approved material claims. The output is a monthly reconciliation showing delivered, delayed, rejected, superseded, and next work.
2Does the Agency Have a Coherent Page and Topic Plan?
Modern SEO requires clear representation of what the organization is, what it offers, who is responsible, where it operates, and which questions each page answers. This can be called entity and topic planning, but the label is less important than the deliverable.
For a law firm, concepts such as Personal Injury, Medical Malpractice, and a State Bar reference may belong in different page types and evidence contexts. The agency should explain which pages target services, which educate users, which verify people or locations, and how internal links support the journey.
Structured data can describe visible, supported information, but advanced Schema.org types do not create Knowledge Graph recognition by themselves. SameAs links should point to profiles that genuinely represent the same person or organization.
Content should use the specific language clients use when accurate, without assuming high-volume terms attract only tire kickers. For a financial firm, 'tax-advantaged wealth transfer for high-net-worth individuals' may justify a focused page only when the service is real, the audience is appropriate, and qualified reviewers approve the claims.
The agency should maintain a content inventory, page purpose, overlap analysis, source plan, and internal-link map. The output is a clear architecture rather than an invented entity-health score.
3Is Experience and Expertise Supported by Evidence?
E-E-A-T is part of Google's quality guidance, not a single score that an agency can switch on. Adding an author bio may improve transparency, but it is only one component of a trustworthy page. The agency should inventory the client's real experience, professional roles, licenses, certifications, publications, speaking work, research, case examples, policies, and correction process.
Every item should be current, permissioned, and relevant to the page. External links should support material claims and should not be presented as proof merely because the source is prestigious. The source refers to 500-word posts as a low-effort alternative to expert interviews.
Retain that number as a content example, not as a quality threshold. An agency may use writers, editors, and subject-matter experts in different combinations; the critical requirement is accurate attribution and review.
Google AI Overviews and other systems may cite many source types, but no evidence locker guarantees external validation or core-update resilience. Review the About page, author pages, editorial policy, privacy information, terms, references, and backlink associations. The output is an evidence inventory with source, owner, expiry or review date, permitted claims, and page usage.
4Does the Workflow Control Compliance and Claims Risk?
Industry awareness should be evaluated through the agency's workflow, not its confidence. Terms such as guaranteed, best, or top-rated may require evidence, qualification, or removal depending on the jurisdiction, service, and platform.
The agency should not assume tactics suitable for a local bakery are appropriate for a medical clinic, legal practice, or financial firm. The client remains responsible for obtaining qualified legal and compliance advice, while the agency should understand the review path and avoid publishing outside approval.
Legal solicitation rules vary. HIPAA may affect certain healthcare data practices, but its exact application depends on the organization, data, vendors, and workflow. GDPR and CCPA obligations also require current legal interpretation rather than a generic checklist.
A strong process can slow publication when evidence or approval is missing, and that delay should be visible in the project log. Review whether high-impact pages have sources, fact checks, subject-matter review, legal or compliance approval where required, version history, and correction ownership.
Link building should avoid schemes that create manual-action, contractual, or reputation risk. The output is a claims register, review matrix, privacy and vendor checklist, approval record, and escalation process.
5Can the Agency Evaluate Google AI and Other AI Search Visibility?
Search Generative Experience (SGE) was a historical experimental name. Current Google product references should use Google AI Overviews or Google AI features. Perplexity and other AI search products also present synthesized answers and citations, but their presence does not eliminate traditional search, direct visits, or other channels.
An agency should be able to explain how it observes these surfaces and how the observations affect content decisions. Self-contained answers, direct headings, clear sources, accurate structured data, and concise summaries may improve readability and machine interpretation, but no format guarantees citation.
The source uses rank #1 as the older objective, calls AI citation the opportunity of the next five years, and contrasts it with 2015-style SEO. Preserve those numbers as editorial comparisons, not verified forecasts or a claim that rankings no longer matter.
Ask the agency to save the exact prompt or query, date, model or product, cited URLs, and whether the brand was mentioned, cited, recommended, or omitted. Compare citation trends with search demand, referral traffic, conversions, and source accuracy.
The output is an AI visibility observation log and a prioritized content correction plan, not an AI-ready certification.
6Can the Agency Explain and Prioritize Technical SEO?
Technical SEO is necessary, but a large audit can obscure the decisions that matter. Missing alt text and duplicate meta descriptions may be important for accessibility or page quality in specific contexts, yet they should not be dismissed as having almost zero revenue impact without analysis.
Ask the agency to distinguish defects, opportunities, monitoring items, and low-priority cleanup. Core Web Vitals, rendering, crawl behavior, indexation, canonicalization, internal architecture, sitemaps, redirects, and server responses may all deserve review depending on the site.
Crawl budget optimization is not automatically important for every website. A 100-page audit should identify the top three actions with evidence, affected templates, expected user or search benefit, effort, risk, and validation.
Automated tools can generate useful raw findings; the agency's value is interpreting, prioritizing, testing, and integrating them with the business. Page speed in a healthcare portal can affect user access and task completion, but it should not be converted into an unsupported ranking or patient outcome claim.
The output is a prioritized backlog with before-and-after evidence, release ownership, rollback plan, and post-deployment monitoring.
7What Most Guides Get Wrong
Generic transparency advice often produces more meetings without creating better evidence. The client does not need access to every internal conversation, but it does need a verifiable record of approved scope, completed work, material changes, dependencies, and outcomes.
Another error is presenting SEO as a choice between keywords and entities. Keyword research, topic coverage, structured data, business information, authorship, internal links, and external references can all contribute to a coherent site.
No agency should be judged by whether it repeats terms such as knowledge graph integration. Judge whether its recommendations are accurate, relevant to the business, technically feasible, compliant, prioritized, and measurable.
Clarity is operational: the client should understand what was done, why it mattered, what remains uncertain, and what happens next.
8Why Agency Clarity Matters More Than Report Volume
Large reports and spreadsheets can demonstrate activity without helping a client make a decision. The useful standard is not minimal reporting; it is sufficient evidence presented in a form that connects work to objectives, risks, outcomes, and next actions.
A client does not need every tracked keyword in every meeting, but it should be able to inspect the query set, methodology, and raw data when those details affect the decision. I would now evaluate an agency by the quality of its reasoning, documentation, review process, prioritization, and willingness to state uncertainty.
A documented process does not guarantee sustainable rankings or protect against every algorithm change. It does make work easier to audit, correct, transfer, and defend in high-scrutiny environments.
9Your 30-Day SEO Agency Evaluation
Evidence Day 1-7
Request the approved scope and a complete log of technical, content, link, and reporting work completed during the last 90 days, including owners, dates, URLs, approvals, and blockers.
Outcome: A reconciliation of contracted work, delivered work, delayed work, evidence, and reported outcomes.
Content Day 8-14
Review your top 10 pages for page purpose, authorship, sources, supported structured data, internal links, current accuracy, and relevant E-E-A-T evidence.
Outcome: A prioritized list of content, evidence, architecture, and governance gaps.
AI Day 15-21
Request a current Google AI and other AI search observation document rather than a historical SGE strategy, including exact queries, dates, citations, and correction priorities.
Outcome: Evidence of whether the agency can monitor current AI discovery surfaces without relying on hype.
Business Day 22-30
Review Share of Voice for an agreed set of high-intent service terms and compare it with qualified leads, assisted conversions, revenue influence, and query mix.
Outcome: A decision on whether organic growth is aligned with the business goals used to evaluate the agency.