The AI Search RFP: A Free Template for Procuring AEO/GEO Services

What Should an AEO RFP Template Include?

An effective AEO RFP template contains seven sections: company background and objectives, scope of work, measurement and reporting requirements, vendor qualifications and mandatory disclosures, pricing structure, evaluation criteria with published scoring weights, and process logistics. Together these sections force vendors to expose their methodology, staffing, and measurement discipline before you commit budget. This article walks through each section in enough detail that you can lift the structure directly into your own procurement document.

The reason a purpose-built RFP matters in 2026 is that AI search procurement has outpaced procurement practice. With AI Overviews mature, ChatGPT search mainstream, and agentic browsing beginning to influence B2B buying, hundreds of agencies have rebranded around AEO and GEO, and a generic marketing-services RFP cannot distinguish genuine capability from repackaged SEO. The template below, which we call the seven-section AI Search RFP, is designed to make that distinction visible on paper.

One principle governs the whole document: every question should be answerable with evidence, not adjectives. Wherever a section invites vendors to describe their approach, pair it with a demand for artifacts, sample deliverables, anonymized reports, named tools, and dated results. Vendors with real practices find this easy; vendors without them reveal themselves in the gaps.

How Should You Frame Background, Objectives and Scope?

Section one, background and objectives, should give vendors the context that determines strategy: your category and primary competitors, current organic and AI referral baselines if known, the buying journeys that matter most, and two or three business objectives stated as outcomes, such as growing AI-sourced pipeline in North America or defending citation share against a named challenger. Vague objectives produce vague proposals; specific objectives let you compare how differently vendors think about the same problem.

Section two, scope of work, defines what you are actually buying. A complete AEO scope spans five workstreams: technical and schema optimization for machine readability, content creation and refactoring for citation extractability, authority and citation development across sources AI engines trust, ongoing visibility monitoring, and strategy governance. Specify which engines are in scope, most enterprise programs start with Google's AI surfaces and ChatGPT, then add Perplexity, Copilot, and Gemini, and state expected volumes where you can, for example 20-30 content interventions per month.

Be equally explicit about what is out of scope, such as paid media, high-volume net-new content production, or developer implementation your internal team will own. Then require vendors to respond to scope with a deliverables table mapping each workstream to named outputs, monthly hours, and the role responsible. That single table becomes your primary tool for comparing proposals that otherwise price identical language at figures ranging from $6,000 to $30,000 per month.

Two practical inclusions strengthen this section further. Attach your priority prompt list, even a draft of 50-100 queries, so vendors respond to your actual answer space rather than hypothetical examples, and state the internal constraints that shape delivery, such as CMS limitations, legal review cycles that add 5-10 business days to publishing, and developer availability. Vendors who visibly adjust their plan around these constraints in the written response are demonstrating exactly the operating discipline the engagement will require.

What Measurement and Reporting Requirements Should You Specify?

Section three, measurement, is where weak vendors fail fastest, so write it with precision. Require a baseline audit within the first 45 days covering citation share across a prompt set of 100-300 queries you approve, current AI referral traffic and conversions, and competitive share of voice in your answer space. Require monthly reporting against that baseline and quarterly business reviews that connect visibility movement to pipeline.

Name the metrics you will accept as evidence of progress: citation share by engine and prompt cluster, AI referral sessions and their conversion rate, and assisted pipeline attributed to AI surfaces. Ask each vendor to disclose their measurement stack, whether tooling is proprietary or licensed, and, critically, whether you receive direct access to raw prompt-tracking data and retain it after termination. Industry benchmarks suggest AI-sourced visitors convert at two to four times the rate of classic organic traffic, so attribution quality is where the ROI case is won or lost.

Also require vendors to state expected timelines honestly. Credible responses will indicate 90-120 days before meaningful citation movement and six to nine months before reliable pipeline impact. Treat any promise of guaranteed citations within 30 days as a disqualifying answer, and say so in the RFP; it deters opportunists from bidding at all.

How Should You Weight Evaluation Criteria and Score Responses?

Publish your scoring model inside the RFP itself. A weighting that has served enterprise buyers well: methodology and approach 25 percent, measurement and reporting capability 20 percent, team and staffing 15 percent, relevant evidence and references 15 percent, price and commercial terms 15 percent, and contract flexibility including exit and data portability 10 percent. Publishing the weights channels vendor effort toward what you actually value and gives procurement a defensible audit trail.

Score each criterion on a 1-5 scale with written anchors agreed before responses arrive, so a 4 in methodology means the same thing to every reviewer. Use three to five reviewers spanning marketing, procurement, and whoever owns web engineering, and have each score independently before discussing. In our evaluation work we typically see independent scoring surface 20-30 percent divergence between reviewers on the same response, and those divergences are usually the most informative conversations in the process.

Note that price carries only 15 percent. In a market where enterprise AEO retainers typically run $8,000-$25,000 per month, the spread between a mediocre and an excellent vendor shows up in outcomes worth far more than the fee difference. The scoring model should make it structurally difficult for the cheapest boilerplate bid to win.

Build the scoring workbook before the RFP goes out, not after responses arrive, and include a mandatory pass-fail gate ahead of weighted scoring: responses that guarantee specific citations, refuse data portability, or decline any required disclosure are eliminated regardless of score. Gating first keeps reviewers from being swayed by polished decks attached to disqualifying terms, and it shortens evaluation by removing weak responses before the committee invests hours in them.

Which Vendor Disclosures Should Be Mandatory?

Section four should compel disclosures that vendors rarely volunteer. Require each bidder to state what percentage of delivery is performed by subcontractors or offshore partners and who your named day-to-day contacts would be. Require disclosure of how AI-generated content is used in deliverables, what human review it receives, and how first-hand expertise is incorporated, since answer engines increasingly discount undifferentiated synthetic text.

Demand a written description of citation acquisition tactics. You want to know whether authority building relies on legitimate digital PR, expert contributions, and community participation, or on tactics that could be read as engine manipulation, such as parasite placements, fabricated review activity, or coordinated content farms. Require vendors to confirm which tactics they will never use and to accept contractual liability if undisclosed tactics trigger visibility penalties.

Finally, require disclosure of tool ownership and data rights, whether their tracking platform is proprietary or white-labeled and what happens to your historical data at exit, any current engagements with competitors in your category, and their policy on performance guarantees. A vendor who guarantees specific citations should explain the mechanism; since no one controls model outputs, the honest answers describe input commitments, deliverables and process, rather than output promises.

What Pricing Structure Should the RFP Request?

Section five should require itemized pricing rather than a single monthly figure. Ask for cost per workstream, technical, content, authority, measurement, and strategy, plus blended hourly rates and total monthly hours. Itemization is what lets you compare a $12,000 proposal to an $18,000 one on substance, and it exposes padding such as four-figure monthly charges for automated reporting.

Request pricing for three engagement shapes: a 90-day pilot, typically $20,000-$45,000 at enterprise scale, a twelve-month retainer, commonly $8,000-$25,000 per month depending on scope, and optional project work such as a standalone audit at $7,500-$25,000. Asking for all three reveals how the vendor thinks about risk sharing and lets you sequence commitment, pilot first, retainer on evidence.

Invite hybrid commercial structures but set boundaries: a base fee covering at least 70-80 percent of compensation, with any performance component tied to metrics defined in your measurement section rather than raw mention counts. Require all assumed tooling costs to be included, and ask vendors to state every circumstance under which fees could increase during the term.

Finally, require the assumptions behind the numbers: expected client-side turnaround times, the number of stakeholder review rounds included per deliverable, and what happens to unused hours at month end. Retainers fail as often on assumption mismatch as on capability, and forcing assumptions into the written proposal makes them contractual rather than conversational.

How Should You Run the RFP Process and Timeline?

Section seven, logistics, sets the calendar and rules. A realistic enterprise timeline runs five to seven weeks: one week for vendor questions with answers shared to all bidders, two to three weeks for written responses, one week for independent scoring, and one to two weeks for finalist presentations and a bounded sample task. Announce the decision date in the document and hold it; process discipline signals that measurement discipline will follow.

Ask finalists to bring the actual delivery team to presentations, not the sales lead, and to walk through one anonymized client engagement end to end: the baseline, the interventions, the citation and pipeline outcomes, and what failed along the way. The willingness to discuss failure candidly is one of the most reliable positive signals in the entire process, because AI search remains a discipline where index volatility humbles every practitioner.

Well-run buyers also arrive at the RFP with their own baseline already in hand, which shortens evaluation and sharpens negotiation. Lemniscate Growth's free GrowthGPT platform includes AEO Checkers and GEO Scorers that can establish that pre-RFP baseline, and its pipeline-first consulting practice reflects the same principle this template enforces: buy AI search services against measured outcomes, not enthusiasm. Issue the document, trust the weights, and let the evidence pick your vendor.