FIELD NOTE
If a pitch deck says “we optimize for ChatGPT, Perplexity, and AI Overviews,” pause. That sentence is easy. The work is harder. Before you pay an agency to “fix your AI search,” ask questions that force them to show their process. You are not looking for a slogan. You are looking for a plan you can audit.
1. “What does success look like in 90 days?”
A weak answer: “You’ll appear in AI answers.” A strong answer includes a baseline, a query set, a content inventory, a technical checklist, and a reporting cadence. Ask them to define success in business terms: qualified leads, product-page traffic, comparison-page views, or support-ticket reduction. If they cannot name a baseline, they cannot measure improvement.
2. “Which queries should we prioritize, and how do you know?”
They should segment queries by intent: informational, commercial, transactional, local, comparison, technical, and support. They should also show the evidence: search-console data, analytics, customer language, product specs, review themes, and AI answer samples. If the answer is “we’ll target high-volume keywords,” that is a legacy SEO script. AI search work starts with the questions your customers actually ask, not just the keywords they type.
3. “How do you distinguish a mention, a citation, and a recommendation?”
This matters. A brand mentioned in a paragraph is not the same as a product recommended for a buyer’s need. A citation is not the same as a product comparison. Ask how they classify AI output. A competent agency will track answer type, source attribution, brand position, sentiment, and whether the answer is factual. If they only count “appears in AI search,” they are selling a vanity metric.
4. “What is your AI answer monitoring process?”
Ask for the workflow: which tools, which prompts, how often, who reviews, and how they handle model drift. They should show you a sample audit. The sample should include the exact prompt, the AI response, the sources used, the brand’s position, and a recommended action. If they cannot produce a sample, ask for a paid pilot with a defined audit deliverable.
5. “What technical changes will you make in the first 30 days?”
They should be able to list concrete items: crawlability, index coverage, structured data, canonical issues, page-speed blockers, entity clarity, schema validation, internal linking, and content accessibility. Ask how they verify that AI systems can parse the page. If the answer is “we’ll add schema and hope,” that is not a technical plan. They should show before-and-after validation.
6. “How do you make content quotable without making it sound like a template?”
AI systems often reward clear structure: direct answers, comparison tables, specs, examples, caveats, and source-backed claims. But generic content is easy to ignore. Ask how they turn product data, case studies, and customer language into content that is specific, useful, and citable. A good agency will show a before-and-after rewrite. A bad one will promise “AI-friendly content” without showing the method.
7. “What do you do when the AI answer is wrong?”
This is the real test. They should have a correction process: document the error, identify the likely source, update the page, add structured evidence, strengthen third-party references, and re-test the prompt. If they say “we’ll just make the content better,” that is too vague. Wrong AI answers are a reputational risk. You need a process, not a shrug.
8. “How do you separate SEO, AEO, and AI search work?”
They should define the scopes. SEO covers crawl, index, technical health, and traditional rankings. AEO covers answer boxes, featured snippets, and structured answer formats. AI search covers LLM recommendations, citations, and conversational answers. If they blur the lines, ask how they will report each stream separately. If they cannot, you will get a mixed report that proves nothing.
9. “What will you stop doing that no longer moves the needle?”
A credible agency will tell you what to stop. Maybe it is low-value link building, keyword-stuffed pages, or chasing rankings for queries that do not convert. Maybe it is over-investing in content that no buyer reads. If they refuse to say “we will stop this,” they are not thinking about your budget. They are thinking about their invoice.
10. “What do you not know yet?”
This is the most revealing question. A strong agency will name unknowns: model behavior, citation transparency, client-specific query volume, or the time required to see AI search impact. They will propose a pilot, a test set, and a review date. A weak agency will claim certainty. If they promise guaranteed AI rankings, walk away. The models change. The search landscape changes. Your agency should be honest about uncertainty and rigorous about testing.
Takeaway: Before signing an AI search engagement, make the agency prove it can measure, test, and report. Ask for a baseline, a query set, a technical checklist, a sample audit, and a 90-day plan. If they can only talk about “AI visibility,” you are buying a buzzword. If they can show you how they will find the gap, fix the page, verify the answer, and report the result, you have an agency that understands the work.