Organic AI Growth Kit, the study
Your buyers are asking AI about you. What is it telling them?
You have read what the AI assistants say about you.
Now find out why they say it, and what to do about it.
A few questions about your brand, then we write to you to set up the call. About two minutes. Not there yet? Start with the free report.
Imagine being named, ranked high, and recommended by every AI assistant.
One answer per square. This grid is the depth we see, and what we analyse.
01The problem
Buyers now ask before they search. The answer is a recommendation, not a list of links.
A buyer asks an assistant which brand to choose and gets three names, each with a reason, and a warning about a fourth. They call one of the three. Your search ranking never came up, and nothing about it reached your analytics. This is already how buyers in researched categories decide, and the answer hardens around whoever the assistants already cite.
No trace
No impression, no referrer, no query report. The conversation never reaches your analytics.
Trusted
The buyer takes the answer as advice from an independent advisor with nothing to sell, and acts on it.
Invented
Where the assistant lacks a fact about you, it fills the gap with what a typical brand in your category looks like.
Settling
Every month the answer is left alone, it hardens around whoever the assistants already cite.
We act like your buyer, and we get into those rooms, thousands of times, to build a clear picture of what is happening behind those closed doors.
02What you unlock
What is said about you, why, and what to do about it.
In one kit.
Every study delivers the same kit: how ChatGPT, Gemini, Claude, Perplexity and Google's AI answers perceive your brand, your products and your services, why, and a ranked plan to change it. The six main pieces, in one line each. All twelve are further down.
The baseline report
Where you stand today across ChatGPT, Gemini, Claude, Perplexity and Google's AI answers: how often you are named, how often you are recommended, and how you are described when a buyer asks about you by name. The numbers every later result is measured against.
The factual error register
Every wrong statement the assistants make about you, listed, marked as helping or hurting you, with a numbered fix for each.
The real competitive set
The brands the assistants recommend instead of you, in the order they rank them, and the reason each one is given. Usually not the list you track.
The positioning gap map
The attribute you want to be known for, who the assistants say owns it today, and what they believe you stand for instead.
The citation map
The pages and kinds of page the assistants cite when they answer about your category, and whether yours are among them. Your placement list, ready made.
The action kit
Every change ranked by the effect we expect, written as a brief your team or a contractor can execute, with the effort each one takes.
Everything above is built from the same evidence: thousands of real answers to the questions your buyers ask, read one by one.
See all twelve pieces of the kit
The baseline
Everything the AI assistants currently say about you, who appears, who wins, what is invented, and which sources keep appearing. Seven pieces.
How often you are named and how well you are spoken of, across every assistant your buyers use, chaptered and evidenced. The baseline every later decision, and every later result, is measured against.
Invented founders, credentials you never claimed, prices that are wrong, guarantees you never offered. Each one becomes a numbered fix on a list you control.
The competitors the AI assistants put beside you, in the order they rank them, and how each one is described. Often a different list from yours, and always the one that matters.
What you meant to be known for, against what the AI assistants say you are known for, and who took the attribute you wanted. So you know where to fight and where to walk away.
The sources the AI assistants cite when answering about your category, and whether yours are among them. Your PR and placement target list, ready made.
Whether the AI assistants can read you at all: structured data and schema markup, entity clarity (do the assistants know which company, product and place you are), knowledge graph signals, name, address and product data consistency across the pages they cite, and crawlability. The half of the problem no copywriter can fix, specified so a developer can.
Every place your own account of the business disagreed with what buyers and AI assistants actually see. Often the most valuable page in the whole study.
The action plan
The changes most likely to move the numbers, ranked, each written as a brief anyone can implement. Four pieces.
Every action written as an end state anyone can check, with the effort it takes and who does it. Your team can pick it up on Monday, and you can tell on Friday whether it was done.
The category's vocabulary as the AI assistants use it, your words that buyers never use, and the words they expect and do not find. Usable in your ads and pages this week.
Your buyers' real questions in their own phrasing, each one marked evidenced, claimed only, or never addressed. The list of proof you owe them.
An operating file for whoever writes your ads and pages, human or AI: the rules, the facts copy may never contradict, the buyer's words. So every new page makes the AI assistants' opinion of you better, and never worse.
The monitoring
Readings after your baseline that separate the models drifting from the effect of your own actions. One piece.
Once your baseline is set, we keep measuring. Competitors act, AI models update, citation sources shift, opinions move. The readings separate the models drifting from the impact of your own actions, so you always know where you stand, what moved it, and how to hold and improve your position in an environment that never sits still.
03In practice
What clients found, and what they are doing about it.
Every business believes something about itself that the AI assistants do not. The study is where the two meet.
They were watching the wrong competitor.
The brand was sure it knew who was taking its customers. The AI assistants disagreed. When buyers asked for a recommendation without naming anyone, the assistants never mentioned the brand, and handed the answer to four competitors, two of which the brand had never heard of.
Asked about the brand by name, the assistants knew it well and argued against it. They quoted the brand's own strongest marketing claim and rebutted it in the same breath, they credited it with certifications it did not hold, and they invented a co-founder it never had. Most of the argument against the brand traced back to two causes, both fixable in days.
The brand now has a ranked action plan, with those two causes at the top.
The AI assistants were sending buyers to the wrong city.
The founder assumed the AI assistants had never heard of her business. They had, and they were confidently getting the basics wrong, including where it was.
The study also found something she had not seen. The business was serving two different kinds of buyer, who used different words and compared her against different competitors, and every page had been speaking to both as one. It also surfaced buyers with real commercial intent for what the brand offered, in a market the founder did not know existed.
She has since rebuilt her advertising around the two buyer groups, is preparing to tap into the newly discovered high potential market, and is correcting the facts the assistants got wrong.
We thought we knew who was taking our customers. The study showed the AI assistants recommending four brands when buyers asked about our category, two of which we had never heard of, and quoting our own marketing claim back as a reason against us. Uncomfortable to read, and exactly what we needed to know. The Siavak team learned our business fast, asked the questions we had not thought to ask, and gave us a report the whole marketing team could work from.
CMO, a consumer brand in the United StatesI thought the AI assistants did not know my business at all. They did. And they got the basic facts wrong. The study also found things I did not see myself: I have two very different kinds of buyer, and I was talking to both in the same way. And there is a whole market I did not know about. I changed my advertising because of it. I came with a question about AI. I left with a much clearer business. The report told me what the assistants say about me, why, and what to change, step by step, in an order I could follow on my own.
Founder, a design business in Europe04How a study runs
Four weeks to real insight. Eight weeks to real impact.
It starts with the free report, and every step after it ends with something you can use.
The free report
We look you up the way a buyer would and send you what the assistants say about you today, in their own words. It shows what is happening; the study shows why.
Discovery
We learn what you sell, to whom, and against whom, in a few short rounds of questions, checked against what buyers and the assistants already say about you.
Study design
We set the questions your buyers ask, the assistants that matter in your market and the competitors to track, and you see the scope and the quote in writing before you decide anything.
Research
We ask every assistant thousands of times, in your buyers' words, read every answer ourselves, and trace each claim to the sources behind it.
Report and action kit
Your baseline across the assistants, why they say what they say, and a ranked plan to change it, walked through with you in a working session.
Monitoring
The same questions, the same assistants, after your changes land, so you see what moved and what moved it, apart from the models drifting on their own.
05The measures inside the report
What the report measures, and why the assistants say what they say.
Two dimensions. You have to uncover both, and optimize for both.
How often you appear, and how well you are spoken of when you do. Being talked about and being talked about well are separate problems with separate fixes.
Every study answers the same two questions about your brand, your products and your services. What the AI assistants are saying. And why they say it, so you can change it. You will also know which of what is said about you can be changed in weeks, and which will take months.
You can be perfectly visible and still lose on the vertical axis, because the AI assistant names you and then argues against you. A sentiment score will not tell you that. Sentiment averages a paragraph into a mood. It cannot separate a true criticism from an invented one, a recommendation from a passing mention, or tell you which competitor took the argument. You need the claims themselves, one by one, and that is what a study gives you.
A buyer asked for your category. Three brands were named. Was yours?
Share of organic mentions and recommendation rate, across every major AI assistant.
A buyer asked about you by name. What did the assistant say?
Claim balance: every statement for or against you, one by one, checked as true or invented.
A buyer never arrived, and your analytics saw nothing.
Steering rate: how often the assistant sends a buyer towards you, away from you, or to somebody else.
A fact the model got wrong stays wrong until somebody fixes it.
Accuracy rate, and the fix list: what can be changed in weeks, and what will take months.
Share of organic mentions, the shelf
How often you appear when a buyer asks without naming a brand. For a brand that was never built for this, the number often starts at zero, and it is where customers are won or never met.
Recommendation rate
How often you are in the recommendation list, how often you are the only one in it, and who is there when you are absent. Only the recommendation sends you a buyer.
Claim balance
The arguments for or against you, when someone checks you by name. The score the AI assistant is keeping on you.
Accuracy rate
How much of what is said about you is actually true. Every invented fact, in your favour or against you, so you can fix the negative and protect the positive.
Steering rate
How often the AI assistant pushes a buyer towards you, away from you, or towards somebody else. The salesperson and the bouncer.
Who else is on the shelf
Who the AI assistants actually put next to you, in what order, and how far that drifts from the competitors you know of and track. Often a different list from yours.
Why it is being said
The citation sources
When an AI assistant shows its sources, we capture them. You see which pages and kinds of page keep appearing around your brand and your category, and whether yours are among them. The answers are associated with somebody's pages. You will know whose.
The arguments
What is said for and against you, and for and against each competitor, claim by claim. You will see exactly which sentence is doing the damage, and which one is doing the selling.
The gaps
What your buyers ask for that nobody in your category proves, which competitor fills each gap today, and on what evidence. So you can act to fill it, with the argument and the proof the AI assistants are already looking for.
The blue ocean
The claims nobody owns yet. The attributes your buyers care about that no competitor has taken. The places you can lead with the least effort, because nobody is arguing back.
Asking your ChatGPT once is holding a single piece of a puzzle. We assemble the whole puzzle, thousands of pieces, to show you the clear picture.
Competitors you have never heard of, and the ones you track that never appear
The organic shelf: who the AI assistants recommend when a buyer asks without naming a brand.
Scroll sideways for the whole chart
Asked by name, the AI assistants argue against you, or act as your salesperson
The claim balance: every statement made about you when a buyer checks you by name, sorted by topic and by direction.
Scroll sideways for the whole chart
What shapes the AI assistant's opinion of you?
The citation map: the AI assistants, and the kinds of page they cite when answering about you.
Scroll sideways for the whole chart
06Why a study, not a tool
A number to peek at,
vs a plan for success.
Monitoring tools count mentions and attach a sentiment score, and because they monitor a whole market at once they have to ask generic category questions. We uncover what the AI assistants actually say about your brand and your products, claim by claim, against the questions your own buyers ask. A tool that only counts mentions can only see one dimension.
07Fit
Who this is for,
and who it is not for.
This is for you if
Your buyers research before they buy.
Products and services that cost more, take longer to decide on, are technical enough to need understanding first, or carry enough risk that the buyer asks around. Industrial and business purchases, such as machinery, components, lab equipment, or the software a company runs on. Professional services, such as law firms, accountants, clinics, architects, or agencies. Complex consumer purchases, such as a mattress, an e-bike, a kitchen, a course, an insurance policy, or a supplement someone takes every day. In those categories the AI assistants are already in the buyer's research, and what they say decides who gets the call.
This is not for you if
Your buyers decide on impulse.
A low ticket purchase, made in seconds, where the buyer does not research and the AI assistant is never asked. Nobody is talking about you behind closed doors, so there is nothing for a study to measure yet.
08Who does the work
Siavash Kanani
Founder and head of analytics
More than ten years in digital analytics, and Siavak Analytics since 2017. Siavash analyzes, advises, designs and strategizes, and reviews implementation, for brands and for the agencies that serve them.
Seven years of teaching Looker Studio at lookerstudiomasterclass.com. A speaker at every MeasureSummit since it started in 2020, at several MeasureCamps, and at five Superweeks.
Siavash on LinkedIn09Before you ask
The questions everybody asks.
Is this "AI visibility", "GEO", or "AEO"?
It goes by all of those names, and they all describe the same goal: being in the answer when a buyer asks an AI assistant, and being described well when you are. Most of what is sold under those names is a tool that counts mentions, or an agency that rewrites your pages. This is the research that has to come first: what is actually being said about you, why, and what will change it. Without that, the tool counts the wrong thing and the agency rewrites the wrong page.
How does this relate to the SEO we already do?
The AI assistants read many of the same pages your search engine does, so good SEO is a foundation, and nothing in the study asks you to undo it. But ranking and being cited are different. A page can rank first and never be quoted by an assistant, and a brand can be absent from search and still be the one recommended, because of what a review site, a forum thread or a comparison page says about it. The study shows you which pages are shaping the assistants' opinion of you, and whether yours are among them. Much of what it asks you to change lifts your search performance as well, because it is the same evidence, put where both can read it.
We rank well on Google. Do we still need this?
Ranking on Google gets you discovered. That is where the buyer's work starts, not where it ends. After discovery the buyer researches you, and that is where the AI assistants are deeply involved: the buyer asks about you by name, compares you with the names the assistant offers, and takes the answer as the opinion of an independent, unbiased advisor. That opinion makes or breaks the actual sale. Search decides whether they find you. The assistant's answer decides whether they become a paying customer.
Strong search performance makes the fixes faster, because the assistants already know where to find you. It does not tell you what they say once they do.
Which AI assistants do you cover?
Every major assistant your buyers use, including Google's own AI answers, and the set is chosen for your market during discovery, because it is not the same for a consumer brand in the United States as for an industrial supplier in Germany. Each one is asked the same questions, and each one is read on its own, because they do not agree with each other.
How do I know the AI assistants matter for my category yet?
Ask one. Type the question a buyer in your category would type, without naming a brand, and see whether it names anyone. If it does, the conversation exists and somebody is already winning it. The free research report shows you what is said about you today, and discovery then measures how much of your buying decision runs through it. If the answer is very little, we tell you before you commit to a full study.
How is this different from an AI monitoring tool?
Monitoring tools count mentions and attach a sentiment score, and because they monitor a whole market at once they have to ask generic category questions, never your buyers' questions about your brand. We uncover what the AI assistants actually say about your brand and your products, claim by claim: whether each claim is true, which competitor it favours, where the assistant read it, and what it takes to change it. A tool works from one kind of signal. A study combines several, in one controlled reading of your brand. A tool gives you a number to watch. A study gives you the reasons, and a plan.
The tools are built to scale across thousands of brands, which is what keeps them shallow. The study we do together goes deep on yours.
Can I just ask ChatGPT about my brand myself?
Yes, and you should, because it is how most people first notice the problem. One conversation is like asking a friend, or a stranger on the street, what they think of you. It tells you what one person said once, and nothing about what everyone who is asked is being told. AI assistants answer differently every time, differently for every phrasing, and differently from each other.
A study collects thousands of real answers, across all of them, and reads them beside data a chat window cannot give you, to form a clear picture. Asking your ChatGPT once is holding a single piece of a puzzle. We assemble the whole puzzle, thousands of pieces, to show you the clear picture.
What is in the free research report, and what is not?
What the AI assistants say about your brand today, from memory and after looking you up, in their own words, for the one product or service you name. It shows you whether you are named when a buyer asks, and what is said when a buyer asks about you by name. It does not show why, or what to change; that is the research. There is nothing to commit to, and nothing follows it unless you ask.
What does a study cost?
Every study is designed for one brand after discovery, and the quote follows the scope: how many products or services, how many markets, how deep the competitor set. You see the scope and the quote in writing before you decide, and you do not decide anything before that.
How do you measure results?
Against your own baseline. The same questions, the same AI assistants, after the study, so what moved is measured the way it was first measured. The readings separate three things that move at once: the models drifting on their own, your competitors acting, and your own actions landing. You see the share of organic mentions, the recommendation rate, the claim balance, the accuracy rate and the steering rate, each beside its baseline.
How long does it take, and what do I have to do?
About four weeks from the start of discovery to your baseline and action plan. Your part is a few short rounds of questions at the start, and a working session at the end of the research, around week four. Every action comes with a brief clear enough for your own team or a third party contractor to pick up and implement.
Do you implement the recommendations?
No. We uncover, we explain, and we prepare a detailed and prioritized action plan so anyone can execute it, whether someone in house or an expert you hire for it. Keeping the implementation outside the study is what makes the later measurement worth trusting: whoever writes your content and your pages should not be the one grading whether it worked. Where you need hands, we can point you to people who have done it.
What happens after the research?
By then you know what moved and what moved it. The AI assistants will keep changing, and so will your competitors. You might keep the monitoring running, or pause and come back for another study when something significant has happened in the market or in your product or service line. That conversation happens when both of us can see what it should be.
10The next step
Tell us about your brand,
and we will come back
with what your study has to cover.
Each study is custom, designed after discovery. The call is to learn about you, and to answer any questions you have about the process, the timing, what you get and what to expect. A few questions about your brand, then we write to you to set up the call. About two minutes.