AI Search Optimization Agency: How to Tell a Real One From a Renamed One

What you are hiring an AI search optimization agency to do

You are hiring someone to change which businesses an AI assistant names when your customers ask for help. That is a narrower and stranger job than ranking on Google, and it is worth being clear about it before you read a single proposal.

When a person asks ChatGPT, Perplexity, Claude, Gemini or Google's AI answers for a recommendation, the assistant writes a short reply and usually names a few businesses. An agency that does this work well is trying to get you into that short list, for the questions your buyers really ask, and to show you evidence that it happened.

This page is about judging the supplier. If you want to understand the work itself, what is fast, what is slow and what cannot be promised, read our guide to what the work actually involves first. It takes ten minutes and it makes every conversation with an agency shorter, because you will already know which of their claims are worth testing.

One caution before the checklist. Almost every agency that sold search work last year now sells AI search work this year. That is not a scandal. Much of the underlying craft carries over, and honest agencies say so. The problem is that you cannot tell the honest ones from the rest by reading their websites, because the websites all use the same words.

Two kinds of agency use the same label

Sit through enough sales calls and you will notice that suppliers fall into two groups, whatever they call themselves.

The renamed menu

This agency already sold search engine optimization, content and link building. It added a line for AI search, usually by relabeling existing services. The audit now mentions AI. The monthly report has a new page with a screenshot from ChatGPT. The work you receive is the work you would have received last year.

That is not automatically bad. If your site has real problems with crawling, speed or thin pages, fixing them helps in both worlds. The issue is paying a premium for a label. If the deliverables match the old menu, the price should match the old menu too.

The agency built around answers

This supplier starts from the question, not the keyword. It can tell you which questions your buyers ask, which assistants answer them, what each assistant said last month and which businesses it named. Its work plan is built around the gaps in those answers.

You can tell the second kind by what it asks you. A good agency asks how your customers phrase their worries, what they compare you against and where they read before they buy. A renamed menu asks for access to your analytics and your list of target keywords.

Ask for evidence, and know what counts

The whole decision comes down to evidence, so it helps to know what counts and what does not.

A single screenshot proves very little. These assistants vary. Ask the same question twice and the answer can change. An agency that shows you one screenshot of one answer naming a client is showing you a lucky draw, whether or not it meant to.

A repeated measurement proves something. The same set of questions, run on a schedule, against the same assistants, with the full answers saved. Over a few months you can see whether a client's name appears more often. Ask to see one real series, with the client's name removed if needed.

Before and after pairs are stronger still. Ask for a question where the client was absent, the specific work done in response and the same question a few weeks later. A good agency has several. A weak one has a story about general improvement.

Rankings on Google do not count here. A client can rank first for a keyword and never be named in an assistant's answer, and the reverse also happens. If the agency's proof is a ranking chart, it is proving the old service.

Our page on comparing generative engine optimization companies goes deeper on how to read this evidence side by side when you have several proposals on the table.

Find out who actually does the work

Agency pitches are often delivered by senior people and executed by junior ones, or by software. Neither is automatically a problem, but you should know which one you are buying.

Ask these plainly. Who writes the pages? Who decides which questions to target? Who reads the assistants' answers each month and writes up what changed? If the answer to the last one is a dashboard, then nobody is reading the answers, and the report will say what the software says.

Ask whether a human reviews every page before it goes live, and what that review checks. A page can be well written and still be wrong about your product, your return policy or your service area. In this work, a confident wrong sentence is worse than silence, because assistants repeat what they find.

Ask what happens when the person you met leaves. In small agencies the relationship often is the product. In larger ones the process is, and the process should be written down somewhere you can read it.

None of this needs a long interview. Three direct questions and the quality of the answers will tell you most of what you need. If you want a full list to work from, our page of questions to ask an AI SEO agency has twelve, each with what a good answer sounds like.

Look for work outside your own website

A proposal that is entirely about your own pages has left out half the job.

Assistants do not rely on what a business says about itself. They lean on third-party sources: review sites, forums, local and trade press, directories, comparison articles. When the assistant has to choose between businesses, it tends to prefer the one that several independent sources describe in the same way. Your own site is a lead. It is rarely the final word.

So a good agency will ask where your category gets discussed. Which review sites matter for you? Which forums? Which local or trade publications? Then it will build a plan for being present and accurate in those places. That work is slower, it cannot be bought in bulk and it is awkward to put in a monthly deliverable list, which is exactly why weak proposals leave it out.

If you want to see what this looks like in practice, our page on what AI engines cite explains the kinds of sources assistants lean on. An agency that cannot talk about those sources in specific terms for your category has not looked.

Beware the opposite failure too. Any offer to buy mentions, reviews or placements in bulk is a liability, not a shortcut. Assistants and the sites they read have both become good at ignoring paid noise, and a pile of thin placements can leave you worse off than before.

Check that they start from buyer questions

Keyword lists were built for a search box. People type three or four words and scan ten links. They talk to an assistant differently. They write a full sentence, add their situation and ask for a recommendation.

"Plumber near me" becomes "my water heater is leaking and I rent, who should I call and what should it cost." "Running shoes" becomes "I have flat feet and run about fifteen miles a week, what should I buy." A question set built from real situations like these is the foundation of everything else.

Ask the agency to show you how it builds that set. A good answer involves your sales calls, your support inbox, your reviews and the questions people ask at the counter. A weak answer involves a keyword tool and a language model that guesses what people might ask.

Then ask who signs off on the final list. It should be you, because you know how your customers talk. If the agency refuses to share the list, or says it is proprietary, treat that as a stop sign. The list is the yardstick for the entire engagement, and you cannot judge results against a yardstick you are not allowed to see.

Read the contract for what it commits them to

Contracts in this field are young, and the weak ones are easy to spot once you know what to look for.

Activity promised, outcome not mentioned. "Twelve pages a month" tells you what the agency will do. It says nothing about whether any assistant will name you. Ask what outcome the agency is willing to be measured against, and whether any part of the fee depends on it.

No baseline clause. The contract should require a measurement before work starts. Without it, any later improvement can be claimed and any decline can be blamed on the assistants.

Long lock-ins. Twelve months up front for a service whose results show slowly is a lot to ask of a client who cannot yet judge quality. Look for a short first term with an exit, or a clear way out if the baseline shows the plan is wrong.

Ownership of the work. The pages, the question set and the measurement history should belong to you. Ask in writing what you keep if you leave. The measurement history is the most valuable piece and the one most often lost.

Our page on AI SEO agency red flags covers the warning signs that appear in the pitch itself, before a contract exists.

Run a small test before a long commitment

You do not have to decide on the strength of a pitch. A small paid test, even a few weeks long, tells you more than any number of calls.

Agree on ten to twenty questions that matter to your business. Have the agency, or you, run them across the assistants you care about and save the full answers. That is the baseline. It will probably look discouraging. That is normal and useful.

Then choose two or three questions where the answer is plausibly fixable. Maybe the assistant names a competitor because it found a clear comparison page and you have none. Maybe it gets a fact about you wrong. Ask the agency what it would do about each, and have it do just those things.

At the end of the test, run the same questions again. You are not looking for a miracle. You are looking for three things. Did the agency explain what it did and why? Did the measurement work and the answers get saved properly? Did anything move in the direction the agency predicted? An agency that behaves like a careful scientist during the test will probably behave the same way during the engagement.

If the agency declines any kind of small test, ask why. Sometimes there is a real reason, such as a minimum scale below which the work is not worth doing. Sometimes it is a sign the proposal depends on you not looking closely.

Match the size of the agency to the size of the problem

A one-person consultancy, a mid-size agency and a done-for-you service are different tools. Matching the tool to the job saves money.

If you have a developer, an editor and time, a good consultant can set the direction while your team does the building. That is cheaper and keeps the knowledge in house. We cover that choice in our page on when an AI SEO consultant is enough.

If you have no one to write and publish, advice alone will leave you with a plan and no pages. A consultant's report that nobody acts on is the most common way this money is wasted. Then you need people who publish, which means an agency or a done-for-you service.

Scale matters too. A store with a few hundred products and a service business with one location have very different needs. Ask the agency about its smallest client and its largest. An agency that has only served big brands may price you out or hand you a process built for a team you do not have.

The company's own history is a fair question too. How long has it done this specific work, as opposed to search work generally? Anyone can be honest about being new. What you want to avoid is an agency that implies a long track record at something that has only existed for a short time.

How RunOctopus handles this

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

There is nothing to install, nothing to approve, nothing to upload and nothing to schedule. You do not approve each page. Pages go live on their own. In the first week that means 8 long-form articles, 6 pages and a custom interactive tool built for your niche. After that it is 175+ pieces of content a month, written, published and linked.

It also reports on where ChatGPT, Perplexity, Claude and Gemini name your business. The promise is more visibility on Google and in AI answers within 90 days, measured against the 90 days before you joined, or every dollar back.

Results from Google Search Console, 28 days to September 29, 2026. Sterman, a real estate business in Los Angeles, went from 44 to 210 Google clicks a month since April. All Angles Creatures, the reptile supply store where RunOctopus started, went from 0 to 4,500+ in five months.

You can see what Ollie would do for your site, free, with no account and no call: paste your address for the free site read.

A short way to decide

If you only remember five things, make them these.

One, ask for a repeated measurement, not a screenshot. Two, ask who writes the pages and who reads the answers. Three, check that the plan reaches beyond your own website. Four, read the contract for outcomes and for who owns the history. Five, run a small test first.

An agency that welcomes all five is probably worth talking to, whatever it charges. One that resists most of them is telling you how it expects to be judged.