How do I choose a generative engine optimization agency?

Choosing a generative engine optimization agency is like choosing any specialized writer: look at their method, their samples, and what they refuse to promise. The right agency will show you the exact questions they run, how they score the answers, and what pages you get. The wrong one will sell placement inside ChatGPT as if it were ad inventory.

Five questions to ask

1. What is the prompt list? A real agency has one. It should be based on your customer questions, not a generic keyword list.
2. How do you score the answers? Look for a 0–2 scale: not mentioned, cited, or named. Anything fuzzier is not measurement.
3. What do I actually receive? You want published pages with schema, not just a PDF report.
4. How often do you rerun the prompts? Monthly is standard. One-time snapshots are incomplete.
5. What happens if nothing moves? The honest answer is that some questions do not move, especially broad ones dominated by national brands.

Red flags

• — Guaranteed mentions in ChatGPT, Claude, Perplexity, or AI Overviews.
• — "AI recommended" badges or "as seen in" claims.
• — No visible author or date on their own guides.
• — A platform seat fee with no clear page count.
• — Refusal to share the prompt list or the scoring method.

What a good proposal looks like

A good proposal names the deliverables, the measurement, and the limits. For example: 20–30 commercial-intent questions across multiple systems, a documented baseline, a defined set of priority fixes, one revision, and regular reruns if monitoring is included. It also states explicitly that placement is not guaranteed and that the work is the improvements and the observed snapshot.

Who this is for

• — Owners who have been pitched "AI SEO" but cannot tell what is actually being bought.
• — Marketing teams that need an outside pair of eyes on what AI engines say about the brand.
• — Small businesses that want pages built once and then maintained, not a dashboard subscription.

Sources

Claims about engine behavior should be verified against OpenAI, xAI, Perplexity, and Google Search Central documentation. Claims about agency pricing should come from written proposals, not industry averages.