GEO agency RFP template: copy-ready sections and scoring rubric
A copy-ready GEO RFP template for AI visibility services: scope, measurement, governance, pricing and exit sections, plus a weighted rubric to score bids.
- A good GEO RFP fixes the prompt set, engines, markets and run method up front, so every bid is measured the same way.
- Weight measurement rigor and proof of shipped change highest; give price about 15 of 100 points.
- Require raw data exports, an error ledger and a handover clause, so your history survives a change of vendor.
A GEO RFP template should do three things: define the measurement method yourself, ask for proof of shipped changes rather than promises, and score every bid against the same weighted rubric. Below are eight copy-ready sections you can paste into your procurement document, followed by a 100-point scoring rubric and the mistakes that make GEO RFPs hard to compare.
Edit the bracketed fields, delete what does not apply, and keep the structure.
01Why do GEO RFPs need a different structure from SEO RFPs?
Classic SEO RFPs ask for keyword rankings and traffic forecasts. Answer engines do not have stable rankings. In a SparkToro and Gumshoe study of nearly 3,000 runs, the same prompt produced the same list of brands less than once in a hundred runs.[1]
That means a vendor can make almost any number look good by choosing its own prompts, engines and run counts. Your RFP has to fix those inputs so bids are comparable. The rest of the structure follows from that.
You write the prompt set and the measurement method. Vendors propose how to improve the numbers, not how to define them.
02What sections should a GEO RFP include?
Eight sections cover the decisions that matter. Each block below is written so you can paste it directly into your RFP.
- 11. Background and objectivesWho you are, which products and markets are in scope, and the business problem: missing from answers, misrepresented, or losing to named competitors.
- 22. Scope of surfacesEngines, app stores and marketplaces in scope, plus markets and languages.
- 33. Measurement methodYour prompt set, run counts, locations and reporting cadence. Vendors must use it.
- 44. Execution and ownershipWho drafts, who approves, who publishes, on-site and off-site.
- 55. Governance and complianceBrand, legal and regional sign-off, audit trail, prohibited tactics.
- 66. Team and toolingNamed people, hours, tools used and who holds the licenses.
- 77. CommercialsPricing model, pass-through costs, term and termination.
- 88. Data, handover and exitWhat you own and receive at the end of the contract.
Section 1: Background and objectives (copy-ready)
> [Company] sells [products] in [markets]. Buyers increasingly research us through AI assistants and AI search features. We are seeking a partner to (a) measure how often and how accurately we appear in AI answers for a defined set of buyer questions, (b) correct inaccurate information at its source, (c) increase our share of answer against [named competitors], and (d) operate this program continuously through our existing CMS and approval processes. Success will be judged on measured change against the baseline defined in Section 3, and on the volume and quality of changes shipped.
Section 2: Scope of surfaces (copy-ready)
> In scope: [ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Microsoft Copilot, Claude], [app stores / marketplaces if relevant]. Markets: [list]. Languages: [list]. Bidders must state which surfaces they can measure directly, which through third-party tools, and which they cannot cover.
Coverage differs widely between tools, so ask. Ahrefs Brand Radar, for example, lists six AI platforms plus Claude for custom prompts,[2] while Profound's Enterprise plan lists up to nine answer engines.[3]
Section 3: Measurement method (copy-ready)
> We will supply a prompt set of [N] questions per market and language. Bidders must: run each prompt at least [N] times per engine per measurement cycle; state logged-in or logged-out conditions and locations; report mention rate, share of answer against named competitors, factual accuracy and cited sources; and deliver raw exports (CSV or JSON) for every cycle. Bidders may propose additional prompts, reported separately.
For background on defensible metrics, see our guide to AI share of voice.
Section 4: Execution and ownership (copy-ready)
> Describe, for each work type (on-site content, technical fixes, structured data, review profiles, directories, digital PR, community engagement, documentation), who drafts, who approves and who publishes. State whether you can work inside our CMS [name] and ticketing system [name], and your expected volume of shipped changes per month at the proposed fee.
Section 5: Governance and compliance (copy-ready)
> Describe your approval workflow for brand, legal and regional review, the audit trail you maintain, and how you handle regulated claims. Confirm in writing that you will not use fake reviews, undisclosed paid placements, undisclosed conflicts of interest or invented personas, and that all community participation will be disclosed.
Our article on AI visibility governance covers the controls legal teams usually ask for.
Section 6: Team and tooling (copy-ready)
> Name the lead strategist and core team, their allocated hours, and native-language reviewers for each in-scope language. List every measurement tool you will use, who holds each license, and whether license costs are included in your fee.
Section 7: Commercials (copy-ready)
> Provide pricing as [monthly retainer / project / hybrid], with all pass-through costs itemized. State minimum term, notice period and any performance-linked components, including how performance is measured under Section 3.
Ask for tool costs separately, because they vary from published list prices to custom quotes. As of October 2026, Semrush lists its AI Visibility Toolkit at $99 per month,[4] while Profound publishes a free trial and custom Enterprise pricing only.[3]
Section 8: Data, handover and exit (copy-ready)
> On termination we will receive: all raw measurement exports; the full prompt set including any additions; the error ledger (each inaccurate claim, its traced source, the fix and confirmation date); credentials for any third-party profiles created on our behalf; and written playbooks for recurring work. No data or account created for us will be retained as vendor property.
03How should you score GEO RFP responses?
Agree the weights before any bids arrive, and have at least two people score each response independently. The rubric below puts proof of shipped change and measurement rigor ahead of price, because a cheap program that never ships changes costs more in the end.
| Weight | What a top score looks like | What a low score looks like | |
|---|---|---|---|
| Measurement rigor | 20 | Uses your method, repeated runs, raw exports, accuracy tracked | Screenshots, own prompts, no run counts |
| Proof of shipped change | 20 | Dated list of changes and before/after on the same prompt set | Anonymous case studies, traffic charts only |
| Execution model | 15 | Works inside your CMS and approvals; clear owners | Delivers recommendations for you to implement |
| Off-site authority | 10 | Concrete plan for reviews, press, directories, communities | Mostly blog content on your domain |
| Governance and compliance | 10 | Written approval flow, audit trail, prohibited-tactics pledge | No mention of legal or brand review |
| Market and language coverage | 5 | Native reviewers for each language in scope | Machine translation, one market |
| Commercials and value | 15 | Transparent fee, itemized pass-through, sensible term | Opaque bundle, long lock-in |
| Data ownership and exit | 5 | Accepts Section 8 in full | Vendor keeps data or accounts |
Score each criterion on a 0–5 scale, multiply by its weight, divide by five and add up. Discuss the scores only after everyone has finished, so the loudest reviewer does not set the result. Keep the individual score sheets; procurement and audit teams often ask for them later.
These weights are our editorial recommendation, not a benchmark. A heavily regulated business might move 5 points from commercials to governance; a 12-market rollout might raise coverage. Change the weights before the bids arrive, never after.
Illustrative example: scoring one criterion. A bidder scores 4 out of 5 on measurement rigor (weight 20). Its points for that row are 4 × 20 ÷ 5 = 16. Repeat for every row and add the results to get a total out of 100.
04What mistakes make GEO RFPs hard to compare?
The most common one is letting each bidder choose its own prompts. Each will pick questions it already wins, and you will end up comparing marketing.
The second is asking for "rankings." Position in an AI answer is too unstable to report alone; ask for mention rate across repeated runs instead, which the SparkToro study found far more meaningful.[1]
The third is ignoring crawler access. If your site or CDN blocks AI search crawlers, no vendor can get you cited. OpenAI, for instance, states that sites must allow OAI-SearchBot to be eligible for ChatGPT search results.[5] Ask bidders to audit this in their first week.
05Should you include operating partners, not just agencies?
Yes. The market now includes software platforms, SEO suites with AI modules, classic agencies and operating teams that embed in the marketing function. A well-written RFP lets all of them bid on the same terms, so you can see which model fits.
OwnlyReal is one of the operating-team type: tool-agnostic, measuring share of answer and shipping fixes through the client's CMS, ticketing and approval chains. The same rubric applies to it as to anyone else. You can see the model in more detail under why OwnlyReal.
For the questions to ask in the pitch meetings that follow, see how to choose a GEO agency.
Questions people also ask
What should a GEO RFP include?
Background and objectives, surfaces in scope, a measurement method you define, execution ownership, governance, team and tooling, commercials, and data ownership and exit terms.
Who should write the prompt set for a GEO RFP?
You should, with input from sales and product. If each vendor picks its own prompts, their numbers cannot be compared.
How should GEO bids be weighted?
A practical split is 40 points for measurement rigor and proof of shipped change, 25 for execution and off-site authority, 10 for governance, 15 for commercials and 10 for coverage and exit terms.
Should a GEO agency guarantee results in the RFP?
No. AI answers vary between runs and engines change models. Ask for committed work volume, measurement cadence and transparent reporting instead.
How long should a GEO pilot or first term be?
Long enough for changes to pass your approvals and for engines to recrawl them. Many buyers run a short paid pilot before a longer term, with the same method used throughout.
