Questions to Ask an AI SEO Agency: Twelve, and What the Answers Should Sound Like

How to use these twelve questions

You do not need to be an expert to interview an agency. You need twelve plain questions and the patience to wait for a real answer. Each one below comes with what a good answer sounds like and what a weak one sounds like.

Take the list to the call, or send it ahead and ask for written replies. Written answers are often more honest, because the agency has to commit to wording. If an agency declines to answer in writing, notice that.

Before you start, it helps to know what the work is. Our guide to what an AI search provider actually does explains it in about ten minutes, and it makes the answers easier to judge. If you want to compare several suppliers in a structured way, see how to choose an AI search optimization agency.

The questions run in four stages, shown in the diagram. Start with proof, because a vague answer there colors everything that follows.

Questions 1 to 4: Proof

1. What questions do you measure, and who writes the list?

Good: "We write it with you, from your customers' real language, usually twenty to fifty questions, and we keep it fixed." Weak: "Our system generates it from your website," or "It is proprietary."

2. Can I see a baseline from a past client?

Good: They show one with the name removed, including the full answers and the competitors named. Weak: They show a slide with percentages and no source answers.

3. How do you deal with answers changing from run to run?

Good: "We run each question several times, across assistants, and report a rate." Weak: A blank look, or "the AI is consistent."

4. Show me a question where a client went from absent to named, and what you did.

Good: A specific example with dates, the work done and the before and after answers. They also mention a question that did not move. Weak: "All our clients see improvement."

If you want to see what a baseline involves before the call, our AI visibility checker guide shows how to run a rough one yourself.

Questions 5 to 8: The work

5. Who writes the pages, and does a person review each one?

Good: A clear description of the writers, the reviewers and what the review checks, such as accuracy about your products and policies. Weak: "Our AI handles it" with no mention of a person reading the result.

6. What will you do outside my own website?

Good: Names the kinds of source that matter in your category, such as review sites, directories, forums, trade or local press, and says how they would check and fix your presence. Weak: "We build authority" and nothing else.

7. What happens in the first thirty days?

Good: A crawler access check in week one, a question list agreed with you and a baseline, then first pages. Weak: A strategy document, with measurement starting in month three.

8. How do you decide what to write first?

Good: "From the gaps in real answers: questions where a competitor is named and you are not." Weak: "From keyword research."

Question 6 deserves an extra moment. Assistants lean on third-party sources more than on what a business says about itself. Our page on what AI engines cite explains the idea, and an agency that has not thought about it will show it here.

Questions 9 to 11: Terms

9. What exactly does the fee include, and what costs extra?

Good: A written list of included items, with setup, developer time, extra questions and extra assistants called out. Weak: A single figure and "it depends."

10. What do I own if I stop?

Good: "The pages are on your site. The question list and the measurement history are yours to export." Weak: The history lives behind a login that closes when the contract ends.

11. How long is the commitment, and how do I leave?

Good: A short first term, a clear notice period and a way out if the baseline shows the plan is wrong. Weak: Twelve months up front for work whose results show slowly.

For what drives the price behind question 9, see our page on how much AI SEO costs. It explains why quotes for the same label can differ so widely, and gives you a way to compare them.

Question 12: Limits

12. What will you not promise?

This is the best question on the list, because it is hard to rehearse.

Good: A short list that comes quickly. "We will not guarantee a placement. We will not give you a single timeline for all assistants. We will not buy mentions in bulk. We will not claim revenue from AI answers with false precision." Weak: A pause, then "we are confident in our results."

An agency that has been in this work for a while has been burned by overpromising, or has watched others get burned, and holds its limits ready. One that has never thought about limits has probably never had to defend its claims.

If the answers to the other eleven were good and this one is weak, ask a follow-up. "What would make you tell a client the plan is not working?" A good agency has a straight answer, and a willingness to say it.

How to score the answers

Do not tally points mechanically. A few answers matter far more than the rest.

Treat answers to questions 1, 2 and 10 as near-deal-breakers. They cover who writes the list, whether a baseline exists and whether you own the history. A weak answer on any of them makes everything else harder to trust.

Treat answers to 5, 6 and 7 as strong signals about quality of work. Treat 9 and 11 as about money and risk. Treat 3, 4, 8 and 12 as about thinking, which is the hardest thing to fake.

Write the answers down as you hear them, in the agency's own words. Memory flatters. Two days later, a fluent salesperson feels better than they sounded. Notes let you compare calls fairly.

If you are talking to several agencies, ask each the same twelve, in the same order. Differences between answers then reflect differences between agencies and not between conversations.

Follow-ups that work on almost any answer

When an answer is vague, three follow-ups usually cut through.

"Can you show me an example?" Real experience produces examples quickly. Rehearsed claims produce more claims.

"What would that look like in my case?" It forces the answer onto your business, where general principles stop working and real planning begins.

"How would I know if it was not working?" A supplier who can describe failure clearly has probably seen it, and probably measures for it.

Be friendly and curious, not combative. Most agencies, and certainly the good ones, respond well to specific questions. The aim is to find out how they work, not to catch them out. If a supplier gets defensive about plain questions on a first call, imagine the fifth month of the engagement.

Questions to ask yourself first

Before the agency calls, answer four questions of your own.

Who on your side will work with the agency each week, and how many hours do they have? What would count as success in ninety days, in terms you could check? Which twenty questions would you most want an assistant to answer with your name? And what is the most you would spend before you had seen evidence?

Having answers ready changes the call. You are no longer being shown a menu. You are describing a job and watching how the agency responds. Many agencies, asked to price your specific job, will propose something smaller and more honest than their standard package.

A worked example, labeled as an example

Here is an invented exchange, to show how the same question lands with two different agencies. The business and the answers are made up for illustration.

A flooring store owner asks question 4: "Show me a question where a client went from absent to named, and what you did."

Agency one answers: "A client in home improvement asked us why a competitor was named for 'best flooring for a house with dogs.' Our baseline showed the competitor had a page that answered exactly that, with a list of scratch-resistant options. The client had a general flooring page and nothing on pets. We wrote a page that answered it in the first paragraph, with specifics on the products they actually stock. About six weeks later, in repeated runs, the client began appearing in some of the answers. Not all of them. A second question on stairs did not move at all, and we are still working on that one."

Agency two answers: "We have a lot of clients who are seeing great results across all their key questions. Our approach is proven and our clients are very happy."

The first answer has a specific question, a specific gap, a specific fix, a time frame, a partial result and an honest miss. The second has nothing you could check. You do not need to know anything about AI search to hear the difference. That is the quality you are listening for in every answer on the list.

What to listen for besides the words

How an agency answers can tell you as much as what it says.

Speed and ease suggest practice. A good agency has answered these before and does not need to improvise. Long silences are fine when the question is hard. Fluent filler is the thing to be wary of.

Who answers matters. If a salesperson answers every question about the work, ask to speak with the person who would actually do it for a few minutes. Their answers to questions 5, 7 and 8 will be more reliable.

Listen for the word "we" and the word "you." Good agencies talk about what you will be able to see and check. Weak ones talk about what they are great at.

Notice what they ask you. An agency that asks how your customers phrase their worries, where they read before buying and what they compare you against is doing real work. One that asks only for your analytics login and your keyword list is running a standard process.

Finally, watch for honest disagreement. If you suggest something unwise, a good agency will say so politely. An agency that agrees with everything is selling, not advising.

A short message to send before the call

You can save time by sending a short note ahead. Something like this works.

"We are looking at AI search help. Before we talk, could you send a sample baseline with the client name removed, an example of a question where a client went from absent to named, and a list of what the monthly fee includes and what costs extra? We would also like to know who would write our pages and whether a person reviews them."

Notice what comes back. A prompt, clear reply with attachments is a good sign. A reply that says only "let us hop on a call" is fine for some agencies and a flag for others, since the point of the note was to avoid a pitch.

Whatever arrives, compare it across agencies using the same twelve questions. Our page of AI SEO agency red flags lists the patterns that tend to show up in those replies.

How RunOctopus answers these questions

RunOctopus is a done-for-you service. Its AI, Ollie, researches your business, then writes and publishes pages directly on your own website, while a team plans, writes, reviews and publishes alongside it. It works with Shopify, WordPress, WooCommerce and Wix.

On terms, you install nothing, approve nothing, upload nothing and schedule nothing, and pages go live on their own. The first week is 8 long-form articles, 6 pages and a custom interactive tool for your niche, then 175+ pieces of content a month.

On proof, it reports on where ChatGPT, Perplexity, Claude and Gemini name your business. Its 90-day promise is more visibility on Google and in AI answers, measured against the 90 days before you joined, or every dollar back.

Florida Realtor Careers went from 8 to 116 Google clicks a month since launching in June (Google Search Console, 28 days to September 29, 2026).

Ask us the twelve if you like. Or start with the free site read: paste your address, Ollie reads the site and shows the plan. No account, no call.

The twelve in one line each

Measured questions, a baseline, repeated runs, a real example, who writes, outside work, the first thirty days, how topics are chosen, what the fee covers, what you own, how you leave and what they will not promise.