AI Search
Getting your business named when someone asks an AI instead of typing a search.
“How do we get seen by AI?” is the question we get asked most often at the moment, usually by someone who has just watched an AI answer a question their business should have answered.
It’s a fair question and it has a real answer. It also has a great deal of nonsense attached to it, so here’s what we actually know.
What changed
Search used to send you a list. Increasingly it hands over an answer, with a handful of sources credited underneath. If your business isn’t among those sources, you aren’t in the conversation, no matter how well you rank on the page below.
The obvious assumption is that ranking well is enough to get cited. Until recently that was broadly true. It isn’t any more.
Ahrefs analysed roughly 4 million citations across 863,000 keywords and found that the proportion of AI Overview citations coming from pages ranking in the top 10 had fallen from 76% to 38% in about six months. More than a third now come from pages outside the top 100 entirely.
The reason is a mechanism called query fan-out. Instead of answering your question directly, Google quietly breaks it into a set of smaller related questions, searches all of them, and assembles the answer from what it finds across the lot. So the page that gets cited is often not the one ranking for the question you asked. It’s the one that best answered a sub-question you never typed.
That has a practical consequence, and it’s the whole basis of this service: covering a topic properly now beats optimising one page perfectly.
What we actually do
Fix the foundation first. Almost everything an AI needs is what a search engine has always needed: pages it can crawl, structure it can parse, content it can understand. Sites that are invisible to AI are usually just invisible.
Get the crawler rules right. This is where we most often find damage already done. Training crawlers and answer crawlers are different things with different names, and blocking the wrong one quietly removes you from ChatGPT or Perplexity answers while achieving nothing you intended. We audit what you’re currently allowing and write rules that match what you actually want.
Write for the fan-out. If citations come from sub-questions, then the work is mapping the questions around your topic and answering each one properly, in its own right, in language that stands up when it’s lifted out of context. This is why our pages carry proper question and answer sections rather than a marketing paragraph with a question mark on it.
Make your business legible as an entity. Models need to know what you are, not just what your page says. Consistent business details everywhere they appear, structured data that connects the pieces, and presence in the sources that get quoted about your industry.
Measure it. Referral traffic from AI tools, and regular checks on whether you’re named for the questions that matter.
What we won’t tell you
We won’t tell you we can get you into ChatGPT’s answers on a schedule. Nobody can.
We won’t sell you a file that fixes it. We publish an llms.txt because it’s free, but the evidence says no major provider currently reads it, and we’d rather say so.
We won’t pretend this replaces SEO. Every credible study points the same way: the sites that get cited are largely the sites that were already worth citing. AI search raises the value of being genuinely good and lowers the value of being merely optimised.
Who this is for
Businesses that get asked for by name and want to be found when someone checks. Businesses in considered-purchase categories, where buyers research before they call and increasingly do that research by asking rather than searching. And anyone who has typed their own category into ChatGPT, seen four competitors named, and not seen themselves.
Common questions
Is this just SEO with a new name?
Partly, and anyone who tells you otherwise is selling something. AI answers are assembled from the web, so crawlability, clear content and genuine authority still do most of the work. What's genuinely new is that ranking well is no longer enough on its own. Ahrefs found that the share of AI Overview citations coming from pages in the top 10 fell from 76% to 38% in about six months, because Google now breaks a question into sub-questions and pulls sources from all of them. That changes what you write, not just how you rank.
Can you guarantee ChatGPT will recommend us?
No, and be careful with anyone who says they can. There's no submission form, no paid placement and no algorithm anybody outside those companies can see. Answers also vary between users and change week to week. What we can do is make sure you're findable, quotable and described consistently enough that you're a likely source, then measure whether it's happening.
Should I block AI crawlers from my site?
Usually not, and it's easy to block the wrong thing. Training crawlers and retrieval crawlers are separate: GPTBot gathers training data while OAI-SearchBot fetches pages to answer live questions. Block the second and you remove yourself from ChatGPT's answers entirely. The same split applies to Anthropic and Perplexity. If you want out of model training but still want the citations, that's a specific set of rules and we'll write it for you.
Does an llms.txt file help?
Not yet, and we'd rather tell you that than sell it. It's a proposed standard for giving models a clean map of your site, but no major provider has committed to reading it. Google has explicitly said it doesn't use it, and independent analysis across roughly 300,000 domains found no measurable effect on citations. It costs nothing to publish and it may matter later, so we add one. We just won't pretend it's doing something it isn't.
Can I opt out of Google's AI Overviews?
Not without leaving Google Search, which is almost never worth it. AI Overviews are built from the normal search index using Googlebot, so there's no separate crawler to block. Google-Extended, which lots of people block thinking it covers this, only governs Gemini model training. If there's a specific passage you don't want appearing in an AI answer, the right tool is a data-nosnippet attribute on that section, which leaves the rest of the page working normally.
How would I know if it's working?
Referrals from chatgpt.com, perplexity.ai and similar sources show up in your analytics as ordinary referral traffic, and they tend to convert well because the person arrives already half-convinced. Beyond that we track whether you're actually being named for the questions that matter to your business, and check it on a schedule rather than once. The measurement side of this is younger and rougher than the rest of SEO, and we'll be honest with you about the error bars.
Go deeper
Worth reading on this
What's search engine optimisation?
A plain-English explanation of what SEO actually is, how organic results differ from ads, and what has genuinely changed now that AI answers sit above everything.
Stop building links. Build a website worth linking to.
The only durable SEO strategy is being genuinely useful. Why that advice has aged better than anything else we have published, and what it means in practice.
Technical SEO: making sure search engines can actually read your site
The technical layer beneath your content: crawling, indexation, robots directives, sitemaps, canonicals and Core Web Vitals, and how to tell which is broken.
Other services
Ready to stop guessing and start getting enquiries?
Tell us about your site and we'll show you, in plain terms, where you're losing visibility. No obligation, no jargon.
Get your free visibility audit