A VP of digital marketing has three proposals for AI search work. One says the bank will be recommended in ChatGPT within a quarter. One prices a monthly "AI visibility" retainer without saying what is measured. The third comes from the bank's existing SEO agency, which has added generative engine optimization to its scope. Procurement wants a scoring sheet, and compliance wants to know what each vendor will do with the bank's data.
This guide gives a neutral way to make that decision. It covers whether to hire at all, what drives the cost, what an RFP should contain and how to score the answers. CB/I Digital is one of the agencies a bank might consider, and the same questions apply to us.
A defined split often works best. The bank retains strategy, product facts and sign-off. Outside help covers work its teams cannot staff or do not handle often enough to develop the necessary skills.
Survey data suggests most organizations start in-house. Minuttia's 2026 State of AEO report, a survey of 599 people in SEO, content and growth roles published by an agency, found that 52.8% of respondents' existing SEO teams manage AI search, 19.4% treat it as a cross-functional effort and 9.2% have a dedicated specialist or team. Only 2.7% outsource it to agencies. The sample leans toward practitioners and includes few financial services respondents, so it describes the wider market rather than banks. The same survey found the most common challenge was a lack of measurement and attribution tools, cited by 40.6% of respondents.
Skills are part of the decision. In an analysis of 1,175 US SEO job listings by the recruiter Search for Hire, 54.9% of descriptions mentioned AI-related skills such as GEO, AEO or large language models, while only 11% of job titles referred to AI, which suggests employers are adding the work to existing roles. Building those skills in-house takes time, and an agency also needs time to learn the bank's products and review process.
The table below compares the common arrangements.
| Arrangement | Works well when | Watch for |
|---|---|---|
| In-house only | The bank already has a strong SEO and content team, and the scope covers a few products | Measurement across several AI platforms takes tools and time the team may not have |
| Existing SEO agency extends its scope | The agency knows the bank's site and has already passed vendor review | Whether its AI search method goes beyond relabeling existing SEO work |
| Specialist AI search agency | The bank needs cross-platform measurement, passage-level content work and outside source alignment quickly | Onboarding to the bank's products, compliance workflow and release calendar |
| In-house owner with an agency for defined parts | The bank wants to keep strategy and approvals inside and buy capacity for measurement, drafting or technical recommendations | A clear written split of who does what, so nothing falls between the two |
Our article on strategy, ownership and operating model covers how to divide that work and who inside the bank should own the result.
There is no reliable published benchmark. The price surveys that exist cover general SEO and mostly small businesses, and AI search pricing claims on agency blogs come without a stated method. A bank can still assess what drives the cost and compare proposals on that basis.
Budget data is a little more useful than price data. In the Minuttia survey, almost 40% of respondents had yet to give AI search its own budget, 30.2% had assigned it between 11% and 40% of their SEO budget, 22.9% had assigned 10% or less and 7% more than 40%. In our reading, that supports treating AI search as part of a bank's search budget rather than a new line, for the reasons our business case article sets out.
| Cost driver | Why it changes the price | What to ask in the RFP |
|---|---|---|
| Products and categories in scope | Each product needs its own questions, page work and source mapping | Price the first phase for a named set of products |
| Markets and branch footprint | Local questions bring in branch pages, local listings and more question variants | Which markets are covered, and whether branch pages are in scope |
| Platforms and run frequency | Each platform tracked and each repeated run adds measurement work | Which platforms, how often, and how many runs per question |
| Tools | Tracking tools carry their own license fees | Whether tool costs are included, passed through or paid by the bank directly |
| Content and page production | Passage-level rewrites, new supporting articles and tables take writing and editing time | How many pages per month, and what counts as a page |
| Compliance review rounds | Each round of comments adds revision time on both sides | How many revision rounds are included, and how review delays are handled |
| Outside source work | Correcting comparison listings and publisher descriptions takes outreach and follow-up | Who contacts publishers, and whether paid placements are excluded |
| Technical implementation | Recommendations are cheaper than changes made on the bank's systems | Whether the agency recommends changes or implements them, and with which access |
Pricing models vary. A fixed fee suits a defined audit or first phase, a monthly retainer suits ongoing measurement and fixes, and many proposals combine the two. Fees tied directly to citation counts deserve caution. The interagency guidance on third-party relationships warns against performance measures that encourage imprudent behavior, and paying per citation rewards volume over accuracy on topics where accuracy is regulated.
The bank's own costs belong in the comparison too. Compliance reviewer time, web release slots and product team input are needed whichever vendor wins, and a proposal that assumes none of them is understating the work.
The RFP should define the scope precisely and ask how the agency would carry out and measure the work. It should also ask how the agency would fit the bank's review process and handle its data. The checklist below covers the sections most marketing RFPs need for this work, including the items a bank's vendor management team will look for.
| Section | What to ask for |
|---|---|
| Scope | The products, markets and AI platforms in the first phase, and the decision point at its end |
| Method | How the agency finds gaps, and whether its advice is aligned with Google's published guidance on generative AI features, a question Google itself suggests asking |
| Measurement | The question set, indicator definitions and sample readout described in our guide to measuring a bank's AI search visibility |
| Content and compliance | How drafts reach the bank's compliance reviewers, whether it can work from a library of approved wording, and who signs off each piece |
| Technical | What it would ask the web team to change, including crawler access and page structure, and how those requests fit the release calendar |
| Outside sources | How it works with comparison sites and publishers, and a statement that it will not pursue paid or inauthentic mentions |
| Data and security | Which bank systems and data it would access, including any customer information, which AI tools it would use on bank material, whether bank data could be used to train models, its security reports such as a SOC report, and how it reports a breach |
| Subcontractors | Which tools and contractors it relies on, including AI tool vendors, and how the bank is told about changes |
| Ownership | That question sets, tracking data, content and reports created for the bank belong to the bank |
| Exit | How the bank's data and materials are returned or destroyed at the end of the contract |
| Regulatory access | Acknowledgment that the work is subject to examination and oversight by the bank's regulators |
| Team and references | Who would do the work, their experience with regulated industries, and references the bank can call |
| Pricing | A price for the defined scope, with assumptions stated, so proposals can be compared line by line |
The data, subcontractor, ownership, exit and regulatory access items follow the 2023 interagency guidance on third-party relationships, which lists them among the topics a bank typically considers in due diligence and contracts. On 15 September 2026 the OCC, Federal Reserve, FDIC and NCUA proposed to rescind and replace that guidance, and the proposal says there are no generally applicable expected contract terms, even for higher-risk relationships. Comments are due by 16 November 2026. Until a final version is issued, the 2023 guidance applies, and the bank's vendor management team decides which items apply to a marketing vendor.
Score each agency on how it would do the work. Strong answers explain what the agency controls, provide evidence from comparable work and show how its process fits the bank's review. Promises of outcomes beyond the agency's control are a warning sign.
Google's guidance on hiring an SEO is a useful baseline because it now addresses AI search directly. It says no one can guarantee a first-place ranking and advises finding another vendor if one makes that promise. It also says a bank should ask whether a vendor's advice on AEO or GEO aligns with Google's own guidance on generative AI features. It also advises granting read access to Search Console at first rather than write access. Google's guide to its generative AI features adds a warning about third-party tools that claim to use internal Google metrics, since no third-party tool has access to its ranking or AI systems.
| Criterion | A strong answer | A red flag |
|---|---|---|
| Promises | Explains which levers the agency controls and how it will report progress | Guarantees placement in ChatGPT, AI Overviews or a first position |
| Evidence | Results from comparable regulated work, with the method behind them and permission to share | Screenshots of single AI answers presented as proof |
| Platform guidance | Cites the platforms' own documentation for its recommendations | Sells AI text files or content chunking as a Google win, or claims a special relationship with a platform |
| Measurement | A defined question set, repeated runs, accuracy checks and written definitions for each indicator | One "AI visibility score" with no definition, or a tool said to use internal platform data |
| Compliance fit | A drafting and review process built around the bank's compliance step | Treats compliance as a final approval on finished pages |
| Transparency | Shares every change it recommends and the reasoning behind it | Vague about what it will do or which tools it will use |
| Access | Starts with read-only access and asks for more only when a task needs it | Asks for broad system access before the scope is agreed |
Our guide to measuring a bank's AI search visibility explains how to judge the measurement section in more detail. A short paid first phase covering a few products can test the agency's method on the bank's own products. It should end with a baseline and a fix list.
If CB/I Digital is on your shortlist, hold us to the same scoring sheet. We will answer every section, including the data, tooling and subcontractor questions, and price against the scope your RFP defines.
Our SEO & AI Search services describe what that work covers. When your RFP is ready, send it to us and we will answer it in the same format as every other vendor.
Should the bank's current SEO agency take on AI search?
It can, if it answers the same RFP questions well. The advantage is that it already knows the bank's site and has cleared vendor review. The test is whether its AI search method adds cross-platform measurement, passage-level content work and outside source alignment, or relabels the existing scope.
Does an AI search agency need full vendor due diligence?
That is for the bank's vendor management and compliance teams to decide. Both the 2023 guidance and the September 2026 proposal scale due diligence to the risk of the relationship. In our reading, an agency that only reviews public pages carries less risk than one with access to analytics, customer data or the content management system, but that judgment belongs to the bank. The access the agency needs should be settled before the review starts.
Who should sit on the evaluation panel?
The marketing owner of AI search, someone from the web team, a compliance reviewer and a vendor management or procurement representative. Each sees a different risk in the proposals, and the compliance reviewer in particular can tell early whether a vendor's drafting process will fit the bank's review.
Copyright © 2026 CB/I DIGITAL INC. All rights reserved | Privacy Policy