A while back I noticed something in the numbers of a few sites I follow. Traffic arriving via an AI answer, meaning someone asked ChatGPT or Perplexity something and got a link back, behaved very differently from regular search traffic. Those visitors bought more often, signed up more often, got in touch more often.
Searchable.com put a number on it that stuck with me: visitors arriving via an AI answer convert at around 14 percent, compared to just under 3 percent for regular organic search traffic. You can debate how solid that figure is. It comes from one source and your situation may differ. But the direction matches what I see. Someone who asks an AI a question is closer to a decision than someone still clicking around in Google.
And that is exactly the tricky part. Because you can be perfectly findable the traditional way, ranking nicely on your keywords, and at the same time be completely absent at the moment an AI is assembling an answer for someone who is almost ready to decide. You exist somewhere online, but you are not there when it counts.
In this piece I walk through what it takes to actually get cited. I follow roughly the three layers Searchable describes, but I add what I see working in practice, and I am honest about what you cannot force.
First: what does an AI actually do when it builds an answer
Imagine someone asks Perplexity: “what is the best way to handle X.” The AI pulls together a handful of sources, extracts pieces from them, and assembles those into an answer with citations below. You want to be one of those sources.
That only works if the AI can find a passage on your page that answers the question on its own. Not your whole article, one block. An AI does not cite pages, it cites passages. The easier you make those passages to extract, the better your chances of making it in.
Almost everything that follows comes back to that.
Layer one: the foundation
This is the work that is not exciting but does make the difference. Three things.
Write in blocks that stand on their own. Sections of 120 to 180 words under a clear heading demonstrably produce more citations. The mechanism is straightforward: a block that size is large enough to give a complete answer and small enough to be picked up whole. A heading that poses a real question helps, because that is often literally what the user typed.
A practical check: take a paragraph, read it in isolation, and see if it still makes sense. If it leans entirely on the sentence before it, it is hard to cite.
Add schema markup to your pages. That is the structured data that tells a machine what it is looking at: an article, a frequently asked questions block, an organization. Invisible to your reader, readable by the crawler. Article, FAQ and Organization are the three to start with.
Show who wrote it. A name on the article, a short bio that explains why that person has something useful to say, a reference or a review. This falls under what Google calls E-E-A-T: experience, expertise, authority and trustworthiness. AI models weigh similar signals. An anonymous text with no attribution gets trusted less. That is just logical.
Layer two: building momentum
Once the foundation is in place, you go broader.
Be present in more places than just your own site. AI models pull a lot from Reddit, Quora, YouTube and LinkedIn, because those are places where people ask and answer questions in plain language. If your name or your insight shows up there too, in a genuine way rather than as spam, you come up more often as a source. You do not need to be everywhere. Pick the two places where your audience actually is.
Keep your content fresh. A page updated within the last three months gets cited on average 1.67 times more often. That does not have to mean a full rewrite. A new figure, an added example, a date that is current. An AI would rather reference a page that is maintained than one that has been sitting unchanged since 2022.
Build coherence across your pages. If you have three or four pieces on a topic that link to each other and form a whole, you read to a model as a source that genuinely understands the subject. Isolated articles with no connection between them miss that signal.
Layer three: an edge others cannot copy
The first two layers can be copied by anyone. This layer is harder, and that is exactly what makes it valuable.
Make something only you have. Your own numbers from your own practice, a small survey among your clients, a way of looking at something that you developed yourself and gave a name. An AI that encounters a unique data point that exists nowhere else has little choice but to cite you as the source, because there is no alternative.
A warning I will add immediately: do not make those numbers up. A fabricated statistic gets exposed eventually, and then you lose your credibility in exactly the channel that runs on trust. Better to start small with something real than big with something invented.
How to start today
You do not need to turn this into a project. Take your five most important search questions, the things you want to be found for. Ask them to ChatGPT and to Perplexity. See who gets cited. If you are not there, look at who is, and read how their page is built. You usually see it immediately: short blocks, clear headings, a recognizable author.
Hold that up against your own page and you will know where the work is.
Where Findori comes in
Up to here this is manual work, and you can do all of it by hand. But most of it is exactly the kind of work you easily forget or leave half-finished, and that you have to redo on every single page.
That is what we built Findori for, a plugin for WordPress. The free version handles a significant part of the foundation without you having to think about it: schema markup for your articles and FAQs, author and E-E-A-T signals properly in the code, a check that looks at whether your text is divided into citable blocks, and control over which AI crawlers can visit your site plus a log showing who came by.
The Pro version covers the work of layers two and three: a citability check per page, an assistant that writes alongside your content, help with internal links, and a freshness monitor that reminds you which pages are due for an update.
And then the part I believe in most, because it answers the question every other tool leaves unanswered: are you actually getting cited? The AI Visibility Monitor takes a topic or a domain and shows whether your site appears in the answers from ChatGPT and Perplexity, and who is mentioned in your place when you are absent. That is not guesswork anymore, that is measurement. You see whether the work you are doing is actually landing.
Being early in a channel that converts well is rarely a bad idea.
One last practical note
Much of this work is invisible to your reader and readable by a machine. An llms.txt file, a kind of table of contents for AI models, or the schema markup beneath your pages. You do not see much of it yourself, but it is exactly the kind of signal that tools measuring your findability look for. Google’s own PageSpeed Insights now has a category for it, “agentic browsing,” which checks among other things whether your llms.txt is in order. Worth having this right, even if you only notice the effect later.
Start with your five questions. See what the AIs answer now. That will tell you whether there is work to do here.
This article is inspired by the three-layer framework and figures from How to increase AI search citations by Searchable. The statistics mentioned come from that piece.
