Ai Search
How to Rank in AI Search
To rank in AI search means to become one of the few sources an assistant names inside its written answer. There is no column of ten blue links to climb. A system like ChatGPT, Perplexity, or Google's AI Overviews reads the web, picks the sources it judges most relevant and trustworthy, and quotes them. You get there by being retrievable to AI crawlers, legible as a specific business, and structured so a model can lift a clean answer from your page.
That reframes the work. You are not chasing a single position number against everyone else on the query. You are making one assistant, on one question, reach for you instead of a competitor. The levers are concrete, and most businesses have never touched them.
Ranking versus retrieval: what the model is actually doing
Most AI answers are built with retrieval. The model runs a search, pulls back a set of pages, reads passages from them, and composes a reply grounded in what it found. This is often called retrieval augmented generation, or grounding. The passage that gets quoted is the one that most cleanly answered the question, not the one that paid the most or stuffed in the most keywords.
It helps to separate a citation from a mention. A citation is a linked source the answer points to, usually a footnote or a card you can click. A mention is the model naming your business in the prose without a link. Both are worth winning. A citation earns a visit, while a mention strengthens the association between your name and what you do, which is what pulls you into the next answer.
The levers that decide whether AI search cites you
A handful of factors do most of the work. None of them are secrets, and none of them are tricks.
Crawler access
Entity clarity
Answer-shaped content
Corroboration off your own site
Structured data
How we build for AI search
Ninetyfifth Marketing performs AI search optimization on the websites we build, because those levers live in the markup, the content structure, and the crawl rules, not in a panel you can bolt on afterward. The custom site and the optimization are one piece of work, which is why we do not retrofit sites we did not make.
The method is concrete. We audit how assistants describe your business today and where they send people instead. We make the site fully crawlable to the major AI agents and publish an llms.txt file that points them at your most important pages. We rewrite priority pages into clear question and answer blocks, mark up the facts with schema, and tighten your entity so your name, location, and services read the same everywhere they appear. Then we monitor which prompts surface you and which still surface a competitor, and we keep working the gaps.
The deliverables are things you can see: a crawl and visibility audit, restructured pages, schema, an llms.txt file, and ongoing monitoring. We work on a published monthly retainer with a ninety day initial term, and we take one SEO client per industry per city, so we are never optimizing two rivals against each other. See /services/seo/ for how the engagement runs and /pricing/seo/ for the model.
Why you cannot simply buy a ranking
There is no paid placement inside an AI answer today. You cannot bid for a citation the way you bid for a Google ad slot. That is good news if you are willing to do the work and bad news if you were hoping to shortcut it. Visibility in AI search is earned by being the clearest, best-structured, and best-corroborated source on the question, which is a function of the site and the content, not a media budget.
It also compounds. Once a model consistently associates your business with a service in your city, it tends to keep reaching for you. The businesses that start structuring for retrieval now are building a position that is hard for a latecomer to dislodge.
Where to start
Before changing anything, find out what the assistants already say. Ask ChatGPT, Perplexity, and Gemini the questions your customers would ask, and note whether you appear, whether the details are right, and who shows up instead. That baseline tells you whether your real problem is access, entity confusion, weak content, or thin corroboration, and it tells you which lever to pull first.
Questions
Is ranking in AI search the same as SEO?
They overlap but are not identical. Classic SEO aims for a position in a list of links, while AI search aims to be a source an assistant quotes inside a written answer. The technical groundwork is similar, crawlability, clear entities, and good structure, but the target is a citation rather than a rank.
How long does it take to show up in AI answers?
It depends on how often the assistants recrawl and how much corroboration you already have. Access and structure changes can register within a crawl cycle, while building the entity signals and third-party mentions that make citations consistent takes longer. No one can honestly promise a date.
Can I do this on my existing website?
Some of it, yes, if you control the markup, the content, and the robots rules. Ninetyfifth Marketing does this work on sites we build, because the levers live in the structure of the site itself rather than in an add-on. If you want to try it yourself first, start by auditing what the assistants say about you today.
Do I need an llms.txt file to rank in AI search?
It helps but it is not a magic switch. An llms.txt file points AI systems at your most important pages in plain language, which makes your site easier to navigate. It works best alongside crawler access, schema, and answer-shaped content, not as a standalone fix.
