Challenge · AI search visibility
One question, three gates, and the fix at each one
A buyer asks one thing. Between that question and your name appearing in the answer there are three separate checks, they happen in order, and each one fails differently. Most reporting tells you only that you are missing, which is the easy half.
By Priyadharshini Krishnan, Co-Founder & Chief AI Officer · 17 September 2026 · 8 minute read
SEO is work most companies already do. The rest is new, and AEO and GEO are the names people have put on it. All three are gates on the same question, in order: the engine has to be able to read you at all, your page has to be what the answer gets built from, and other people have to have written enough about you for you to be recommended. Fail the first and the other two never happen. That is why which gate you fell at is the whole diagnosis, because the fix is a different piece of work at each one, done by a different person, on a different timescale. A tool that only tells you that you are missing has answered the easy half.
Gate one. Can the engine read you?
Before an assistant can say anything about you, something has to exist, be readable and rank: your own pages, and the pages that mention you. This is the demand your team already works, and it is what an engine reads before it answers anything.
Missing here looks like being absent from the ten links too. If the query behind the question returns nothing of yours, there is no passage to quote and nothing for a recommendation to attach itself to.
The fix lives on your own site and in the links pointing at it, and it is the piece your SEO team or agency has been doing for years. Nothing about AI search replaces it.
Gate two. Are you the answer?
The buyer asks something specific and gets one direct answer. Being that answer means being the page it was built from, and that only happens if your page answers the question on its own, away from everything around it. A page that ranks well and takes four paragraphs to reach the point gets read and skipped.
Missing here looks like ranking while somebody else is the answer, which is the most frustrating version because every dashboard you own says you are winning.
The fix is still on your own pages, but it is a different job from gate one: answers written to survive being taken out of their page, with the structure that lets a model extract and attribute them. Your team can do it in an afternoon per page once somebody has said which pages.
Gate three. Does it recommend you?
The last step names two or three vendors with reasons attached. The reasons come from what other people published about you: review sites, community threads, comparison articles. Your own pages carry less weight at this gate than at either of the others.
Missing here looks like being quoted as a fact and left out of the shortlist. You are in the answer, and you are not one of the names the buyer writes down.
The fix sits on somebody else's property, which is why it is the slow one. Find which sources the answers actually draw on for your questions, and get your name into them honestly. This is outreach, review programs and being genuinely useful in the places buyers already read.
Why the order matters
The gates are sequential, so effort spent at gate three while gate one is failing buys you nothing. The single most common expensive mistake in this category is a brand commissioning citation work while its own pages are not readable or not ranking for the query underneath.
Which module does what
Four things run underneath all three gates, and each one answers a question the gates cannot answer for themselves.
- Brand DNA decides what the drafts say at every gate, so a fix still sounds like you. It is also what an answer is compared against when an assistant describes you differently from how you describe yourself.
- Competitor Signals holds the set of names you are measured against, so an absence has a shape. Being missing means nothing until you know who was there instead.
- Social Signals holds the places the third gate reads from, and what they say about you there. Most of a recommendation is assembled from property you do not own.
- Citability audit checks whether your own pages can be ranked, quoted or cited at all, which is the same question whichever gate you are working on.
A question failing at each gate in turn
Take one question a buyer genuinely asks: which contract management tools work for a legal team of five. Run it, read the answer, and the diagnosis is usually one of three things.
- You are not in the answer, and you are not in the ten links either. Gate one. Nothing you publish about being quotable will help until the page exists and ranks.
- You rank for the query, and the answer quotes a competitor's documentation. Gate two. Your page has the information and cannot be lifted out of it cleanly.
- You are mentioned in passing, and the two names recommended are competitors with reasons attached. Gate three. The review sites and comparison articles the answer leaned on do not carry you.
Three different diagnoses, three different pieces of work, three different timescales. One score, however carefully calculated, cannot tell you which of them you are looking at.
The labels matter less than knowing which step broke, because the fix is a different piece of work at each one.
What to do this week
- Write down ten questions a buyer would really type before choosing in your category.
- Run each one and record three things: were you named, were you quoted, were you recommended.
- For each question where you were missing, work out which gate it failed at before deciding on any work.
- Group the failures by gate. The group with the most questions in it is where your next quarter goes.
If you are an agency
This is the frame that makes the conversation with a client concrete. You already own gate one and have for years, which is worth saying out loud, because otherwise AI search sounds like a replacement for the work. Gates two and three are the new scope, they are separable, and they price differently because one sits on the client's own pages and the other sits on other people's. A client who understands which gate they are failing at stops asking for everything at once.
How we do this ourselves
AI Search Visibility runs your tracked questions on a schedule across the assistants your buyers use, keeps every answer with the sources behind it, and says which of the gates the absence sits at. It then produces the brief, the draft and the structured data for that gate, and re-checks the question afterwards, so the loop closes on evidence. Brand DNA, Competitor Signals, Social Signals and the citability audit all feed the same picture.