How to Evaluate a Generative AI Search Optimization Proposal

To evaluate a generative AI search optimization proposal, check six things: whether it starts with a baseline of your current AI visibility, what work is included, how results are reported, whether the claims are honest, what it costs, and how easy it is to leave. Score each proposal on the same points so you compare like with like.

The six-part scorecard

Give each section a score from 0 to 3. A proposal that scores below 10 out of 18 needs serious questions before you sign.

1. Baseline

Does the proposal show what AI assistants say about you today? A strong one includes test results for several real buyer questions across more than one assistant. A weak one starts work without measuring anything.

2. Scope of work

Look for a clear list of deliverables:

• Technical review of crawler access and structured data.
• A content plan tied to specific prompts.
• Cleanup of business facts across listings.
• Outreach to earn genuine third-party mentions.

Vague phrases like “AI optimisation package” without detail score low.

3. Reporting

The proposal should say what you will see each month: prompts tracked, mention rate per assistant, competitor comparison and accuracy of descriptions. If the only metric is website traffic, it is a classic SEO proposal with new wording. Our guide to measuring AI visibility lists useful metrics.

4. Honesty of claims

Mark down any proposal that:

• Guarantees a mention, a ranking or a traffic increase by a date.
• Says the agency is an official or certified partner of an AI company.
• Offers to put your brand into AI training data.
• Includes review generation, forum posting under fake names or paid links.

Honest proposals talk about improving your chances and tracking trends.

5. Pricing clarity

You should understand exactly what each fee covers. Separate one-off setup from monthly work. Check how many prompts, pages or outreach attempts are included. See our AI search optimization pricing guide for typical models.

6. Exit terms

Short notice periods and full ownership of content and data score well. Twelve-month lock-ins with no performance review score poorly.

Questions to send back to every agency

1. Which assistants will you test, and how many times per prompt?
2. Who on your team writes the content, and how do they learn our business?
3. Which publications or directories do you expect to target for our industry?
4. What will you need from us each month?
5. What would make you recommend stopping the project?

The last question is telling. A confident provider can describe what failure looks like.

Comparing a proposal with a tool

Sometimes the right comparison is not agency against agency, but agency against software plus your own time. If a proposal mainly offers monitoring and a few articles a month, a platform may cover much of it for less. If you need heavy outreach and a large content team, an agency may be worth it.

A worked example

Say Agency A offers “monthly AI optimisation” for a flat fee with traffic reports only. Agency B offers a baseline test of 40 prompts, four rewritten pages a month, listing cleanup and a monthly mention report, with 30 days’ notice. Agency B will likely score higher on baseline, scope, reporting and exit terms, even if it costs more. Use the scorecard to justify your choice internally.

How Keboost helps

Keboost can be your benchmark when you read proposals. It tracks your prompts across ChatGPT, Gemini, Claude, Perplexity and Copilot, shows where you and competitors appear, and creates and distributes focused content, at €29 or €99 per month with no lock-in. Run a free AI visibility check to get your own baseline before you compare offers.