# AI visibility platform RFP template

Vendor-neutral requirements checklist for evaluating AI visibility (GEO / AEO) platforms. Version 2026-10-18. Free to copy, edit and share (CC BY 4.0).

## How to use it

1. Delete requirements you don't need and change priorities to match your policy.
2. Send the same file to every vendor and ask them to fill in the response column.
3. Run a pilot (below), then score.

**Scoring:** Score each requirement 0 (not met), 1 (partly, or planned with a date), 2 (met) or 3 (met with strong evidence). A Must scored 0 disqualifies the vendor.

## Measurement quality (weight 25)

AI answers vary from one run to the next, so the method behind each number matters more than the dashboard.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| M1 | Must | Repeated sampling per prompt | How many times is each prompt asked per engine, market and period? | Runs per prompt and period, stated in documentation | | |
| M2 | Must | Sample size on every figure | Is the number of answers behind each figure shown in the product and exports? | Screenshot or demo account showing n per metric | | |
| M3 | Must | Uncertainty shown | Do you show a confidence interval or margin of error, and how is it calculated? | Method document naming the interval method | | |
| M4 | Must | Change detection | How do you decide a change is real rather than normal variation? Do alerts use the same rule? | Written rule and an example alert | | |
| M5 | Should | Collection lane disclosed | Is each answer collected from the consumer app or a developer API, and is it labeled? | Per-answer lane label | | |
| M6 | Should | Public methodology | Is your methodology published, versioned and dated? | Public URL and change log | | |
| M7 | Should | Reproducibility | Can we recompute a reported figure from the raw answers you store? | Raw answer export and a worked example | | |
| M8 | Should | Extraction accuracy | How do you check that brand mentions and citations are extracted correctly, and what is the measured error rate? | Accuracy audit with sample size | | |
| M9 | Could | Intent coverage | Can prompts be tagged by intent type (for example informational, commercial, navigational, transactional) and reported separately? | Prompt tagging in the product | | |

## Engines, markets and languages (weight 15)

Check what the plan you would buy includes, not what the vendor supports somewhere.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| C1 | Must | Engines included | Which AI engines are included on the proposed plan, and which cost extra? | Engine list per plan, with add-on prices | | |
| C2 | Must | Markets included | Which countries can we measure on the proposed plan, and is there a per-country fee? | Country list and pricing | | |
| C3 | Should | Native-language prompts | Are prompts written natively per language, or translated? | Example prompt set in two languages | | |
| C4 | Should | Unavailable engines handled honestly | When an engine isn't available in a market, is it reported as not checked rather than zero? | Product screenshot | | |
| C5 | Could | Local and multi-location | Can we measure city-level or location-level questions? | Demo of a local project | | |

## Insight and action (weight 15)

A score is useful only if it says what to change, and whether the change worked.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| A1 | Must | Answer receipts | Can we read the full answer text behind every number? | Answer viewer in a demo account | | |
| A2 | Must | Citations and sources | Do you show which sources each engine cited, at URL level? | Source report export | | |
| A3 | Must | Competitor share of voice | Can we compare our share of answers with named competitors by engine, market and topic? | Share-of-voice report | | |
| A4 | Should | Recommended fixes | Do you recommend specific actions, with the evidence behind each? | Example recommendation | | |
| A5 | Should | Re-test after a fix | Do you re-measure the affected questions after a change and report the result with its range? | Example before/after report | | |
| A6 | Should | Accuracy monitoring | Do you detect wrong or outdated claims about our brand, and how often? | Example accuracy report with sample size | | |
| A7 | Could | Shopping and product visibility | Do you track product-level visibility or shopping results, and in which markets? | Coverage list and method | | |

## Reporting and integration (weight 10)

Data should reach the tools your teams already use.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| I1 | Must | Exports | Can we export prompts, answers, citations and metrics in an open format? | CSV or API sample | | |
| I2 | Should | API | Is there an API, which plans include it, and what are the limits? | API documentation | | |
| I3 | Should | Alerts and digests | Which channels do alerts support (email, Slack, Teams) and can we set thresholds? | Alert settings | | |
| I4 | Should | Analytics connection | Can you connect to our web analytics to report AI referral traffic and conversions? | Integration documentation | | |
| I5 | Could | MCP or agent access | Can AI agents or assistants query our data through MCP or a similar interface? | Documentation | | |
| I6 | Could | Client-ready reports | Can reports be shared as links or branded PDFs? | Example report | | |

## Security, privacy and legal (weight 20)

Ask for documents at the start of the evaluation, not after the pilot.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| S1 | Must | Single sign-on | Do you support SAML or OIDC SSO, on which plans, and is setup self-serve? | SSO documentation | | |
| S2 | Must | Independent assurance | Do you hold a SOC 2 Type II report or ISO 27001 certificate? If not, what is the plan and date? | Report under NDA, or a dated plan | | |
| S3 | Must | Data processing agreement | Do you offer a DPA, and is it click-through or negotiated? | DPA document | | |
| S4 | Must | Sub-processors | Which sub-processors handle our data, where, and how much notice do you give of changes? | Published sub-processor list | | |
| S5 | Must | Data location and retention | Where is data processed and stored, how long is it kept, and how is it deleted? | Retention policy | | |
| S6 | Should | Roles and audit log | Which roles exist, and is there an audit log we can export? | Role matrix and log export | | |
| S7 | Should | Provisioning | Do you support SCIM user provisioning? | Documentation or dated plan | | |
| S8 | Should | Vulnerability disclosure | How do we report a vulnerability, and do you publish a security.txt? | Policy URL | | |
| S9 | Could | Security questionnaire | Will you complete our security questionnaire (for example CAIQ or SIG Lite)? | Completed questionnaire | | |

## Commercial terms (weight 15)

Ask every vendor for the same breakdown so offers compare like for like.

| ID | Priority | Requirement | Question for the vendor | Evidence to request | Response | Score |
|---|---|---|---|---|---|---|
| P1 | Must | Price for our scope | What is the total annual price for our brands, markets, engines, prompts and seats? | Written quote | | |
| P2 | Must | Volume in answers | How many AI answers per month does that price buy (prompts × engines × markets × runs)? | Calculation in the quote | | |
| P3 | Must | Software separate from services | Which part of the price is software and which is onboarding or managed services? | Line items | | |
| P4 | Should | Seats | How many users are included, and what does an extra seat cost? | Price list | | |
| P5 | Should | Term and renewal | What are the minimum term, renewal and cancellation terms? | Order form | | |
| P6 | Should | Payment options | Can we pay by card, invoice or purchase order? | Billing terms | | |
| P7 | Should | Data on exit | Can we export all our data if we leave, and when is it deleted? | Terms clause | | |
| P8 | Could | Service levels | Is there an uptime commitment with service credits? | SLA document | | |

## Pilot plan

1. Write one prompt set: at least 50 questions across the intent types you care about, for your key products and markets.
2. Give every shortlisted vendor the same prompts, engines, markets and competitors.
3. Run for two to four weeks so each tool collects repeated answers.
4. Compare each vendor's figures with their ranges. Two tools can disagree and both be within normal variation.
5. Open ten answers in each tool and check the extraction by hand.
6. Score each requirement 0 to 3 and apply the section weights agreed before the pilot started.