Choosing and contracting a search engine marketing agency in 2026

You have a budget to commit, a list of providers and no written procedure. This page is for the marketing director or executive who has to choose a search engine marketing agency and frame the RFP.

It walks through the buying process for a search engine marketing engagement: the document you send out, the schedule you hold to, the grid you score with, the questions you ask at the pitch, the clauses you have written in. By the end, you know what to ask for, how to separate two proposals and what has to be in the contract.

Clarifying the need before launching the RFP

The quality of the proposals you receive depends on the document you send out. If the question is still what you are buying from an SEO agency, which engagement models exist and what they cost, that upstream framing is what has to be settled first.

Before opening any discussions, write a one-page “assignment framework”. Without it, you will be comparing proposals that cannot be compared (different scopes, workloads and priorities):

• Main objective: acquisition (leads), e-commerce (revenue), awareness (share of voice), reducing dependence on paid, launching an offer…
• Scope: domain(s), subdomains, estimated number of URLs, key templates (product, category, “solution” pages, blog, local pages), an existing blog to optimize
• Technical constraints: CMS, JavaScript rendering, development resources available, deployment frequency, legal constraints (GDPR, regulated sectors), brand sign-off
• Markets and languages: geographies, multilingual setup, local variations (terminology, local pages, logistics)
• Deadlines: migration or redesign, product launch, seasonal peaks

The scope you write there does not follow from your objectives alone, but from what your teams can absorb. The concrete decision criterion is your execution capacity. With development and content teams available, you can favour an agency that is more “strategy + QA”. If you are short of bandwidth, look for a partner able to produce, orchestrate and document.

Settle this before the RFP: a candidate who has to guess who will produce the content will price wide, or price wrong. This framing also tests the provider's maturity: a good agency clarifies the grey areas (access, data, dependencies) rather than pricing by guesswork.

The brief: what to supply in order to receive comparable proposals

Without producing a heavy document, supply at least:

• Business objectives and success KPIs (SEO + GEO)
• Scope (domains, markets, languages, estimate of templates and URLs)
• Context (migration or redesign history, sign-off constraints, internal resources)
• Possible access (read access to Search Console and Analytics, security constraints)
• Deadlines and milestones

The “access” item is the one that gets forgotten and the one that costs most afterwards: state in the brief:

• What you can open
• At what level
• Within what time
• What you will not open

Specify the response format you expect, so the scoring grid applies without reprocessing.

You also need to know whether the RFP covers one scope or two: organic search only, or paid search as well? Both answers are defensible, but they do not call for the same document. If you are covering both, three points have to be written down:

• Who owns the boundary: who decides that a query is covered by organic, by paid or by both, and how often that decision is reviewed.
• Which data flows between the two scopes, and who consolidates it into a single report.
• What happens if one of the two goes to another provider: interfaces, synchronization rhythm, owner of the switching plan.

Without that boundary, the proposals will not be about the same object, and the price gap will mean nothing. When the boundary between the two channels has to be written down, their articulation becomes the object of the contract: that is the case of a paid search agency set against organic.

What to require in the proposal, and how to test it

Expect, at a minimum, a clear chain between diagnosis → prioritization → execution → quality control → measurement → iteration. A proposal that skips one of these steps is not a method, it is a catalogue. And at every step, ask for:

• Evidence
• Priorities
• Acceptance criteria
• Documentation that several teams (marketing, content, development) can use

The initial audit: findings, evidence and a prioritized roadmap

Require an audit that leads to executable decisions, not a generic checklist. A useful audit combines three elements:

• Findings: crawling, indexing, performance, content, authority.
• Evidence: Google Search Console and Analytics extracts, sample URLs and templates.
• Prioritized roadmap: impact/effort/risk, acceptance criteria.

Prioritization is not a refinement of presentation: Google relies on more than 200 ranking factors (Hubspot, 2026). Without it, you end up with a list of actions nobody can execute, and you have paid for an inventory. The orders of magnitude that let you challenge a proposal are gathered in our SEO statistics.

And to check the document you are handed, rely on what an agency SEO audit covering SEO and GEO has to deliver: its approaches, its depth and its expected deliverables. An audit whose roadmap carries no dates cannot be accepted.

The technical definition of done and the link building framework

Ask for a minimum “definition of done” on the technical side: what counts as delivered, how it is tested, and how side effects are avoided. On the same principle, link building has to be framed as a process, not as “buying links”. Have these six points written down:

These six requirements go into the contract as they stand. Link building is the line of the quote where the gap between two proposals is hardest to read: without written selection criteria, you cannot refuse a single link delivered. If it weighs heavily in the proposal, it gets framed line by line, against the selection criteria, sign-off and impact tracking of an off-site SEO agency.

Reporting, the first thirty days of deliverables and generative visibility

An agency has to propose a rhythm, format and level of detail suited to your organization. Ask for reporting that supports decisions:

• What is progressing (and why)?
• What is stalling?
• Which actions have been delivered, tested and signed off?
• What is the next focus (impact/effort/risk)?

The deliverables to obtain within the first thirty days follow from those four questions:

• 90-day roadmap: themes, priorities, dependencies
• Prioritized backlog: impact/effort/risk, owner, status
• Briefs: intents, heading plan, supporting evidence, internal linking, quality criteria
• Dashboards: Search Console, Analytics, GEO visibility
• Meeting notes: decisions, trade-offs, actions, risks

The dashboard can no longer stop at rankings and clicks: 60% of searches end without a click (Semrush, 2025). Also ask for:

• The brand's share of citation
• The accuracy of mentions
• Source diversity
• Coverage of your strategic topics

This does not justify a separate RFP for generative visibility: it is built on the same foundation (indexing, content, authority) and steered in the same reporting. To place these measurements, see our GEO statistics.

Shortlist, scoring grid and comparison of the responses

Build a simple scoring grid (out of 5 per criterion, say), then have the stakeholders fill it in (marketing, product, development, management).

• Method: prioritization, evidence, QA, iteration
• Transparency: data, assumptions, limits, risks
• Resources: seniority, availability, continuity
• Deliverables: operational usefulness, clarity, documentation
• Compatibility: governance, cadence, communication

Having several departments score prevents the decision from coming down to the amount. Score individually before pooling, then discuss only the criteria where the scores differ by more than one point.

Breaking down a quote, then separating two well-scored proposals

Ask for a readable breakdown:

• Framing and audit time
• Production (briefs, content, optimization)
• Management (project manager, meetings, coordination)
• Link building (framework, selection, tracking)
• Reporting (dashboards, analysis, recommendations)

Without that transparency, you cannot arbitrate (reduce the scope, increase the pace, and so on). Each line has to carry its unit:

• What is billed by time spent
• What is billed by deliverable
• What is included in a recurring engagement

It is that reading that makes a quote enforceable. If the discussion drifts toward expected return rather than spend, calculating and steering SEO ROI gives the framework.

In most RFPs, two candidates come out one or two points apart: the grid will not decide. What separates them is the comparison of assumptions. So do not compare price alone, compare the assumptions (content volume, level of development support, link building pace, governance). A “cheaper” quote may rest on assumptions that are unrealistic on the client side.

Take the two proposals line by line and ask each the same question: what does this one assume of us that the other does not? The candidate whose assumptions are tenable with your resources wins, even if it is the more expensive one. Finally, check that the guardrails are there: limits, risks, dependencies, and how the plan is revised.

The pitch and the compatibility test

The pitch is there to tell a method from a sales argument. Five questions are enough, asked of every candidate in the same order:

• “How do you prioritize: which data, which criteria, which trade-offs?”
• “How do you avoid generic recommendations?”
• “What is your acceptance testing and quality control process (technical and content)?”
• “What are the dependencies on the client side (development, sign-off, legal), and how do you manage them?”
• “What risks do you identify in our context, and how do you mitigate them?”

Add a sixth, about the team: “who is responsible for what, and how do you ensure continuity over 12 months?” The answer is worth more than an org chart. To get past the prepared answers, rely on what lets you measure the real level of SEO expertise.

The pitch is also a compatibility test. Validate a RACI (who is Responsible, Accountable, Consulted, Informed) from the outset. Many failures come from a “delivered audit” that ends up in a backlog with no owner. Also ask:

• Which single channel for requests
• How often you meet
• Which deliverable formats
• How changes are documented

Without a RACI, a schedule and sign-off rules, you get deliverables… that are never deployed. So have the person who will own the backlog on your side named during the pitch.

How the RFP runs, and its schedule

The procedure comes down to seven steps:

Allow around eight weeks between sending the brief and the kick-off.

Two milestones are not negotiable: the closing of the questions window and the submission date. The window goes early, never the day before submission: a candidate who learns two days before the deadline that they will not get your performance data has no time left to redo their pricing.

The contract: scope, commitments, exit and access

Points to secure in writing:

• Duration and renewal terms
• Scope (what is included and excluded)
• Ownership of the deliverables (audits, briefs, dashboards, content)
• Exit: how to recover the deliverables, the documentation and the history

Exit is not verified at signature but at the first report: ask for the first month's deliverables in a format you can reopen without the agency.

Add what the generative share changes on the commitments side:

• Which structured formats are delivered
• How often content is refreshed
• Who monitors the brand mentions picked up in assistants' answers
• Within what deadline an inaccurate mention is corrected at source

The accuracy of those answers cannot be written into a contract, because it depends on third-party systems nobody controls; what is committed to is the monitoring and the corrective action.

Best-efforts commitment and warning signs

Search engine marketing depends on factors you do not control (competition, algorithm updates, enriched SERPs), which imposes a logic of iteration. A best-efforts commitment (method, deliverables, cadence, QA, reporting) is generally more realistic than a firm ranking promise. An expected gain is stated as an assumption with its conditions, then measured.

The line between what is contracted like any other obligation and what cannot be guaranteed is detailed in the pillar on choosing an SEO agency. Three signs should alert you:

• Promises of a “guaranteed number one” with no analysis of your context
• Absence of a measurement plan (Search Console, Analytics, GEO)
• Deliverables that cannot be acted on: no evidence, no prioritization, no acceptance criteria

SLAs, data access and exiting the contract

In B2B, deal with this subject early:

• The SLAs: response times, meeting frequency, delivery times
• Confidentiality and data processing
• Access management: least privilege, traceability, read access to Search Console and Analytics

Finally, write down what conditions renewal and what triggers an exit. Three simple thresholds are enough to read a trajectory:

• Low CTR despite a good position: a title or meta problem, an intent mismatch, a SERP that is too enriched
• Impressions rising but clicks flat: a “zero-click” or AI Overviews effect, a need for GEO optimization
• Traffic rising but conversions falling: targeting that is too informational, commercial pages to strengthen, UX friction

These three cases call for iteration, not an exit: they describe a plan that can be corrected. The exit clause is triggered by something else:

• Deliverables that fail acceptance two cycles in a row
• A backlog that does not move for lack of documented decisions
• Reporting that stops answering the four agreed questions

Only 0.78% of clicks go to page 2 (Ahrefs, 2025). A page stalling at 11–20 therefore signals work to be taken up again, not a breach of contract. What is enforceable is execution: the consolidation work was in the plan and was not carried out.

Depending on what your RFP covers, several situations change how the scope is written:

• Outlets and catchment areas: if the scope covers physical locations, the evaluation criteria change → local SEO agency working a catchment area.
• Long sales cycle: if measurement has to reach the pipeline, the contract indicators change → B2B web agency: buying committees, proof content and pipeline.
• A multi-channel RFP: if it goes beyond search alone, the framing sits upstream of the channels → webmarketing agency and its data-driven management.
• A scope limited to one person: the question is no longer buying a service but the craft of whoever delivers it, and its continuity → assignments and deliverables of an SEO consultant.

One last task stays on your side: checking the audit deliverable a candidate hands you. Setting it against a map you maintain yourself is what separates a substantiated diagnosis from a well-presented checklist: that is what the 360 SEO and GEO audit module makes possible.

Frequently asked questions about choosing a search engine marketing agency

How do you choose a search engine marketing agency suited to my sector?

Start from your context (sales cycle, regulation, seasonality, competition) and expect a proposal that spells out assumptions, risks and priorities. A relevant agency reformulates your need and proposes a testable roadmap rather than a “standard pack”. Above all, check that it can name what it still does not know about your context and what it must verify once access is open: more revealing than a list of sector references.

What budget should you plan, and over what period, for measurable results?

The budget depends on the size of the site, on competition and on the pace expected: it can only be set once the scope is written, and only compared at equal execution assumptions. To place the spend, the median budget for an SEO assignment, all services combined, is €5,000 excl. VAT (La Fabrique du Net, 2026).

On duration, commit to a horizon compatible with the expected iteration pace, rather than to a one-off assignment, with renewal conditioned on written criteria.

How do you read results commitments without falling for unrealistic promises?

Favour commitments on method, cadence, deliverable quality, measurement and iteration: those are contracted without reservation, like deadlines. An expected gain is stated as an assumption with its conditions, then measured. Be wary, on the other hand, of unconditional ranking guarantees when algorithms change continuously.

Which deliverables should you expect within the first 30 days?

An audit with evidence, a prioritized backlog, a 90-day roadmap, a content plan (intents + briefs), a technical plan (quick wins + longer projects), and an initial dashboard (SEO + GEO). Have each delivered in a format you can reopen and use without the agency: that is the first full-scale test of your exit clause, and it happens before the stakes become real.

Which access and which data should you keep (Search Console, Analytics, content)?

Keep ownership and administration of Google Search Console and Google Analytics, as well as access to the content (CMS, documents, briefs). The agency should work with controlled access, ideally read-only where possible, and clear traceability. It is that rule that makes exit workable: without it, you recover deliverables without the history that explains them, and you start from scratch when you change provider.

How do you build GEO into the strategy without cannibalizing SEO?

By starting from SEO (indexing, content, authority) then adding a GEO layer: structured formats, direct answers, sourced data, refreshes, and citation measurement. GEO complements SEO; it does not replace it.

On the contract side, that translates into written deliverables and a written refresh frequency, plus a named owner for monitoring mentions and for the corrections to be made at source. It goes neither through a commitment on the accuracy of generated answers, which does not depend on the provider, nor through a separate budget line.