How to choose an SEO company becomes much easier when you stop comparing promises and start comparing evidence, operating discipline, and fit. Most agencies can show polished dashboards. Fewer can explain exactly what they found, what they changed, why it mattered, how they validated the work, and where the evidence remains uncertain.
The right partner does not need to know every answer during a sales call. It should know how to investigate, document assumptions, protect your site, and connect search activity to qualified demand. The selection process should test those habits before a contract is signed.
Define the business problem before requesting proposals
“We need more SEO” is too broad to scope. Write a short problem statement that names the business goal, current constraint, priority audience, important products or locations, and known measurement limits.
For example:
We need to increase qualified demo opportunities for two enterprise products in North America. Organic traffic is growing, but CRM attribution is incomplete and sales says many inquiries are poor fits. We also want to understand how our brand appears in AI-generated answers. The first phase must diagnose demand, page ownership, conversion paths, and measurement before content production expands.
This brief invites a diagnostic response. It also makes vague volume promises easier to spot.
Ask for proof in layers
A case study is useful, but it is only one layer of proof. Evaluate an SEO company across four layers.
1. Proof of process
Can the agency show how it moves from evidence to a prioritized plan? Ask for a redacted audit finding, implementation ticket, editorial brief, validation record, and monthly report. You are assessing whether the team can communicate clearly enough for your developers, writers, analysts, and executives to act.
2. Proof of execution
Ask who performs technical changes, content revisions, analytics work, digital PR, and quality assurance. “We handle it” is not a staffing model. You need roles, approval boundaries, turnaround expectations, and a process for changes that could affect revenue or brand risk.
3. Proof of outcomes
Look for results tied to a defined period, market, scope, baseline, and metric. Traffic growth can be meaningful, but it is not interchangeable with qualified leads or revenue. A strong case explains what the agency influenced and what else changed during the period.
4. Proof of judgment
Present a messy scenario. What would the agency do if rankings rose while leads fell? How would it respond to a traffic decline after a redesign? What if an executive wants 100 new articles before analytics works? The answer should show prioritization, not obedience to a tactic.
Verify testimonials and reputation carefully
Reviews can help you understand communication and reliability, but they are not self-validating. Look for specificity, dates, reviewer identity where public, and consistency across platforms. Ask references what went wrong during the engagement and how the agency responded.
The U.S. Federal Trade Commission announced its final rule on fake reviews and testimonials in August 2024, with the rule taking effect on October 21, 2024. The rule addresses practices including fake or false reviews, certain purchased sentiment, and undisclosed insider testimonials. It is a legal standard with defined scope, not a general quality score for agencies, but it reinforces the need to treat reputation evidence seriously. Read the FTC announcement and rule context.
Questions that reveal real capability
“What would you need to learn before recommending a strategy?”
A good answer should include business priorities, margins or customer economics where appropriate, sales qualification, analytics, site history, technical constraints, internal resources, and prior changes. Be cautious if the strategy is already fixed before discovery.
“How do you decide what to fix first?”
Listen for impact, evidence strength, effort, dependencies, and risk. The agency should distinguish a repeated template defect from an isolated warning and a commercial blocker from a cosmetic preference.
“Who will work on the account?”
Ask for roles, relevant experience, expected time allocation, meeting participation, and escalation paths. Clarify whether the people in the pitch will remain involved. If specialists are shared, ask how priorities are scheduled.
“What do you implement directly?”
Some companies need strategic direction for internal teams. Others need hands-on implementation. Neither model is inherently better, but the contract should name deliverables, access, approvals, and acceptance criteria. An implementation partner should preserve backups, review production risk, and verify public output after a release.
“How will reporting connect to lead quality?”
The answer should go beyond rankings and sessions. Look for conversion definitions, CRM stages, call quality, assisted journeys, page groups, and known attribution gaps. The agency should be comfortable reporting “unknown” when the data does not support a conclusion.
“How do you measure AI-search visibility?”
A credible answer separates mentions, citations, referrals, leads, and revenue. It records prompt, platform, date, market, and variability. It does not promise a fixed number of citations or imply that an AI mention caused a sale.
This distinction matters because search behavior is changing. Pew Research Center reported in July 2025 that, among 900 consenting U.S. adults in its March 2025 browsing dataset, traditional-result clicks occurred on 8% of visits with an AI summary and 15% without one. Source links in the summary received clicks in 1% of visits. The findings describe that panel and period, not every query or customer segment. Review the Pew study.
“What would make you recommend against creating content?”
Strong agencies know when the constraint is elsewhere. They may first repair indexing, consolidate overlapping pages, fix product facts, strengthen proof, or improve the conversion path. An agency paid mainly for production should still be able to say when more pages would add maintenance cost without serving buyers.
“How do you handle a failed test or missed target?”
Look for a process: preserve the baseline, confirm implementation, inspect segment-level effects, identify outside changes, document what was learned, and adjust. Avoid teams that relabel every miss as an algorithm update or quietly replace the target.
What a useful proposal should include
- A clear interpretation of your stated problem.
- Initial hypotheses labeled as hypotheses.
- Discovery work required before firm recommendations.
- Deliverables, owners, frequency, and acceptance criteria.
- Implementation responsibilities for both teams.
- Measurement definitions and attribution limitations.
- Dependencies such as development time, legal review, or CRM access.
- Pricing, term, renewal, pause, and termination conditions.
- Risks and explicit exclusions.
Ask the proposer to walk through one month of work. Which decisions will be made? What will your team receive? What has to happen before the next step? This makes a generic package concrete.
Red flags in SEO sales
Guaranteed rankings or revenue
No agency controls search results, competitors, buyer demand, or your sales team. A partner can commit to work quality, response times, deliverables, and validation. It cannot responsibly guarantee a particular ranking or revenue result.
Secret methods that cannot be explained
Proprietary tools can be useful. The decisions made with them still need to be understandable. You should know what will change on your site, which assets will be created, and what risks are involved.
Activity presented as impact
Counts of articles, links, audits, and meetings describe output. They do not prove progress. Every activity should connect to a diagnosed constraint and an intended measure.
Unqualified case-study claims
Be wary of charts without dates, labels, or baselines. Ask whether the chart covers all traffic or a segment, whether other channels changed, and whether results are typical. Do not assume a result in one market transfers to yours.
AI citation promises
AI systems and search products change, responses vary, and platforms control selection. An agency can improve the accessibility, clarity, evidence, and usefulness of your public information. It cannot force a recommendation.
An original comparison scorecard
This scorecard is an original evaluation aid, not a certification model. Weight each category according to your needs, score from one to five, and require notes for every score.
| Category | What to evaluate | Example weight |
|---|---|---|
| Diagnosis | Business understanding, evidence, prioritization | 20% |
| Execution | Roles, implementation, QA, change control | 20% |
| Measurement | Lead quality, CRM connection, limitations | 20% |
| Relevant capability | Technical, content, local, enterprise, or AI needs | 15% |
| Communication | Clarity, cadence, escalation, documentation | 10% |
| Commercial fit | Pricing, term, resources, contract flexibility | 10% |
| References | Specific, verifiable feedback from comparable work | 5% |
Do not let a total score erase a fatal issue. A team with poor data handling or unclear production access may be unsuitable even if it presents well elsewhere.
Hypothetical example: choosing between three proposals
This example is hypothetical. A regional healthcare software company receives three proposals.
Agency A promises 50 articles and first-page rankings. It does not ask about sales qualification or compliance review. Agency B proposes a three-month technical project but excludes content and analytics. Agency C begins with measurement validation, topic ownership, technical sampling, customer-interview inputs, and a prioritized six-month plan. It identifies legal review and developer capacity as dependencies.
Agency C may not have the most dramatic forecast. It has the clearest chain from evidence to execution. If the company needs a partner to own implementation across those functions, C is likely the stronger fit. If the internal team can handle content and analytics, B might still be useful as a bounded specialist. The decision depends on the operating need, not the longest deliverable list.
Reporting standards to agree on before signing
Define the baseline date, source systems, conversion events, qualified-lead definition, page groups, target markets, and reporting cadence. Require change logs. Ask for commentary that distinguishes observations from explanations. Decide who can approve website changes and how rollbacks work.
If AI visibility is included, define the prompt set and observation method. If link acquisition is included, establish quality and disclosure standards. If content is included, define expert input, sourcing, review, authorship, and update ownership.
Use a paid discovery phase when the scope is genuinely unknown
A complex site may not support a responsible fixed strategy after one sales call. A bounded paid discovery can be a better test than a long retainer based on assumptions. Define the questions, systems to inspect, stakeholder interviews, sample limits, deliverables, timeline, and price. The output should belong to your company and be usable even if you select another implementation partner.
Discovery is not permission for indefinite analysis. It should end with a diagnosis, prioritized options, dependencies, measurement plan, and recommended scope. Ask which conclusions will remain provisional and what evidence would resolve them.
Review access, security, and change control
SEO work can require powerful access to the website, analytics, tag management, business profiles, content systems, repositories, and third-party tools. Before granting it, agree on company-owned accounts, role-based permissions, credential handling, approval requirements, backups, production windows, and offboarding.
Ask how the agency records changes and validates them. For a redirect release, for example, the team should preserve the before-state, test the intended rules, confirm important destinations, inspect public behavior, and retain a recovery path. For content, it should preserve sources and approvals. A strong answer protects both performance and accountability.
This review is especially important when an agency uses subcontractors or automated publishing. Ask where data is processed, who can reach production, whether your information trains external systems, and how generated material is checked. Legal and security teams may need to assess the answers.
When Fuel Online may fit
Fuel Online’s SEO services are relevant when a company wants technical, content, authority, and measurement work connected to commercial priorities. Larger organizations can review the enterprise SEO service. Teams that specifically need answer-system discovery can examine AI SEO services. Those pages should be the start of a fit discussion, not a substitute for scope.
Frequently asked questions
How many SEO companies should we interview?
Three to five serious candidates is usually enough to compare approaches without creating an unmanageable process. A complex procurement may need more. Use the same brief, questions, and scoring criteria for each finalist.
Should we choose an industry specialist?
Industry knowledge can shorten discovery and improve risk awareness. It can also produce recycled strategies. Evaluate both relevant experience and the team’s ability to question assumptions. Regulated or technically complex markets may require deeper specialization.
What contract length is reasonable?
SEO needs enough time for diagnosis, implementation, and measurement, but a long commitment should come with clear deliverables and exit terms. Ask what can realistically be validated at 30, 90, and 180 days. Avoid treating the contract term as proof that results require secrecy.
How should we check references?
Ask about communication, missed expectations, staff continuity, implementation quality, reporting honesty, and how the agency handled disagreement. Comparable complexity matters more than an identical logo.
What should we own when the relationship ends?
Your company should retain access to analytics, search accounts, content, creative, documentation, source files, change logs, and accounts created on its behalf, subject to the contract. Clarify licenses for proprietary software and data exports before signing.
The best selection process rewards clarity, evidence, and operational fit. Choose the SEO company that can show how decisions are made and how work will be verified, especially when the honest answer is that more evidence is needed.
