In this blog post I'm going to show you exactly how to get found by AI, using the best evidence I could ask for: a complete stranger who booked a sales call with me last week because Claude told him who I was.
This is Week 15 of my rebuild in public, the series where I tear my website back to the studs and put it back together with nothing airbrushed out. For the last nine of those weeks, a big lump of the work has been aimed at one slightly mad-sounding goal: getting an AI to recommend me, by name, when someone asks for help in my field.
Not a click. Not an impression on a dashboard I squint at every morning. A real person, at a real company, with a real budget, who emailed me cold because an AI engine described my business to him accurately enough that he went and found my inbox. This is the week I found out it works.
Quick result. On 24 June a growth lead at a US coaching marketplace emailed me cold. He already knew I sell one to one AI implementation calls and that my newsletter reaches 15,000 people. On our call, I asked how he found me. His answer: "Through Claude AI." That is what it means to get found by AI. And the system behind it is repeatable: an llms.txt file, answer-first content, entity schema, original data, and relentless internal linking. Here is the whole thing, warts and all.
The rebuild in public, week by week
This post is part of a series where I rebuild my business in the open, one experiment at a time. If you are new here, start at Week 1 and work forward. Here is the whole series so far:
- Week 1: I'm De-Indexing 1,300 Pages to Save My Website
- Week 2: My Open Rate Crashed to 11%
- Week 3: The Unglamorous SEO Work Nobody Talks About
- Week 4: Crawled, Not Indexed
- Week 5: I Took My Open Rate from 11% to 70%
- Week 6: Conquering My Inbox Instead of My Blog Posts
- Week 7: The Income Stream That Saved My Business Three Times
- Week 8: I Let Claude Loose on My Own Website
- Week 9: I Redesigned My Whole Newsletter
- Week 10: I Built a Tool to Automate My WordPress Publishing
- Week 11: I Dug 786 De-Indexed Pages Back Out
- Week 12: AI Inbox Management Gave Me My Life Back
- Week 13: The Month My Website Started Paying Me Again
- Week 14: My Rebuild in Public Numbers, the Month Traffic Crossed Half a Million
- This week, Week 15: How to Get Found by AI (you are here)
Why I bet nine weeks on getting found by AI
Because the old game stopped paying.
I've been doing SEO since before Google could spell. For twenty years the deal was simple: rank, get the click, get the enquiry. Then AI Overviews arrived and started answering questions right there on the results page. ChatGPT and Claude became the first place a chunk of my buyers ask for recommendations. The click never happens, so the ranking never matters.
Search didn't die. It moved house.
When I started rebuilding this site in public, I made a decision that felt risky at the time. I would optimise for the engines that answer, not just the engine that ranks. If someone asks an AI "who can help me put AI into my business", I want the answer to be my name, my offer, and my newsletter, stated as fact.
That only works if the machines can find you, parse you, and trust you. Which is a different job to classic SEO, and nobody hands you a manual for it. So I built the manual myself, and I'm giving it to you in this post.
What it takes to get found by AI
To get found by AI you need three things. Content an AI can lift word for word. Machine-readable signals that tell it exactly who you are and what you sell. And enough repetition and authority across your site that the engine trusts what it reads.
That's it. That's the whole game.
Notice what's not on the list. You don't get found by AI by stuffing "AI consultant" into every second sentence, the engines mark that down, not up. You don't get found by AI by publishing more generic content faster, they already have the generic internet memorised. And you can't buy your way in with ads, because there's no ad slot inside a Claude answer. Structure, facts and trust are the only currencies accepted.
The hard part isn't understanding the list. The hard part is doing it across hundreds of pages without breaking the site, while still running a business, which is why most people read a post like this, nod, and never do any of it. Here's what doing it looks like in practice.
What I did, step by step
1. I gave the AI engines a CV (llms.txt)
There's a file called llms.txt. It sits at the root of your site, like robots.txt does, and it exists purely to tell AI systems who you are, what you do, and which pages matter. Mine says, in plain English, that I'm an AI and marketing consultant, that I offer one to one AI implementation calls for business owners, that my newsletter reaches 15,000 subscribers, and it lists the pages that prove it.
Most sites don't have one. Mine didn't until this rebuild. It took an afternoon.
I can't prove Claude read that exact file before it recommended me. What I can tell you is that the stranger who emailed me knew my exact offer and my exact newsletter size before we'd ever spoken. Those facts live in my llms.txt and in my structured pages. The machine didn't guess them.
2. I restructured articles so the first 60 words answer the question
AI engines don't cite pages. They cite passages.
Every article I publish now opens with a short block that answers the core question outright, before any storytelling. Each subheading is phrased the way a human asks the question, and the first sentences under it form a complete answer that makes sense quoted on its own. I call these capsules. An AI engine can lift one word for word, attribute it, and move on.
This is the single highest-return change on the list. It costs nothing. It just requires you to stop burying your answer in paragraph nine, which most writers do because it feels more dramatic. Drama doesn't get cited. Answers do.
If you want the deeper technique, I broke down how to get cited on Perplexity in a separate post, and the same structure applies to every engine.
3. I published data nobody else had
In June I published an original study on AI search demand, built from real query data. Original numbers are catnip for AI engines, because they need sources for statistics and most of the internet just recycles the same three reports.
If you have any first-party data at all, sales patterns, delivery times, campaign results, publish it with the date and the method. One real number you own beats ten opinions you borrowed.
4. I told the machines exactly who I am (schema and entities)
Behind the scenes, every important page on my site now carries structured data: who I am, what the organisation is, what each article covers, and how it all connects. I have a Wikidata entry. My profiles all point at each other. The name "Lilach Bullock" resolves to one consistent entity everywhere a machine looks.
This sounds nerdy because it is. It's also how an AI engine gets confident enough to state your offer as fact instead of hedging. Machines recommend what they can verify.
5. I built internal links every single day
AI engines lean on the same crawl infrastructure as search engines. A page nothing links to is a page that doesn't exist. So I built a small fleet of automations that place a handful of internal links every day, connecting new articles to money pages, orphaned pages to busy ones, resources to the posts that mention them. Paced, boring, additive, daily.
I wrote about the day I let Claude loose on my own site to audit exactly this, and the fixes from that audit are still compounding. If you do nothing else this month, find your ten most important pages and make sure at least five other pages link to each one with sensible anchor text.
6. I measured it, and the measurement said I was failing
Here's the uncomfortable bit, because I promised you real numbers, not a highlight reel.
I run a monthly check that asks the big AI engines six generic money questions, things like "best AI marketing consultant" and "AI consultant for small business", and records whether any of them name me. My score in June: zero out of six. Zero in May too.
Meanwhile my organic clicks rose 79% in 28 days, and weekly impressions outside one volatile page grew 22% week on week. The site was clearly getting stronger, yet my own tracker kept telling me no AI engine would say my name.
Then a stranger got my name from Claude on a query my tracker never thought to ask.
That's the lesson I want to tattoo on this whole industry. You get found by AI on specific, high-intent questions long before you win the generic beauty-contest queries. My tracker was testing the finals. My buyer asked a semi-final question, something closer to "who offers one to one AI implementation for founders", and the engine had exactly one confident answer.
The results (the email, the call, the sentence)
On 24 June an email arrived from a growth lead at a US coaching marketplace. Cold, but not clueless. He knew I sell premium one to one AI implementation calls. He knew my newsletter reach. He'd read my positioning back to me better than some people I've worked with for years.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
We got on a call on 2 July. Near the end I asked the question I now ask everyone: "How did you find me?"
"Through Claude AI."
Three words. Nine weeks of unglamorous structural work, and the payoff arrived as three casual words on a Wednesday evening call.
I want to be precise about what this is and isn't. It's one lead, not an avalanche. I won't pretend one call means the pipeline problem is solved, because I promised myself I'd never dress up a data point as a trend. But it is proof of mechanism. The path exists: a buyer asked an AI engine a question, the engine answered with my name and my facts, and money-adjacent conversation followed. Everything I do to get found by AI from here is just widening that path.
And one more thing worth saying out loud: he didn't find me on Google. The search clicks are lovely, but the enquiry with a budget came through an engine my analytics can barely see. If you're only measuring clicks, you're already measuring the wrong war.
What I'd do differently
Four things, honestly.
I'd create llms.txt in week one, not week six. It's the cheapest signal on the list and I sat on it because it felt too simple to matter.
I'd test long-tail buying questions in my citation tracker from day one, not just the generic queries. Tracking "best AI marketing consultant" flattered my competitors and hid my progress.
I'd ask "how did you find me?" on every single call, in writing, every time. I nearly didn't ask. The most useful marketing data I collected this quarter came from one question I almost skipped.
And I'd write this post sooner. Documenting how to get found by AI while I was building the system would have made every step sharper, because nothing exposes a fuzzy process like explaining it to someone else.
Should you bother getting found by AI?
If you sell expertise, yes, and you should start this quarter. The work to get found by AI is front-loaded: most of it is one-off structural fixes that keep paying while you sleep. The buyers moving to AI-first research are exactly the time-poor decision makers you want. The engines are answering with or without you. The only question is whose name comes back.
Here's the honest bill. You need a site you control, a clear offer stated in plain English somewhere a machine can read it, a few months of patience, and either the technical confidence to add files and schema or an AI workflow that does it for you. You don't need a big budget. I did this with automation, stubbornness, and a rebuild I was doing anyway.
And you need to resist the urge to fake it. AI engines cross-check. The facts about my business that Claude repeated to my lead were true, current, and consistent across my site, my LinkedIn, and my data. That consistency is the strategy. There's no shortcut where you get recommended for things you can't prove, and there shouldn't be.
Related reading: ai implementation help non technical founders.
Related reading: Best AI Coaches for Entrepreneurs in 2026.
Frequently asked questions
How long does it take to get found by AI?
For me, nine weeks from starting the structural work to the first attributable sales call, on a twenty-year-old domain. On a newer site expect three to six months before AI engines cite you on specific questions. The generic queries take longer. Start with the questions your real buyers ask when they already know what they need.
Do I need to be technical to get found by AI?
No, but you need to be systematic. llms.txt is a text file. Answer-first writing is a habit. Schema can be handled by plugins or AI-powered lead generation workflows you set up once. If you can publish a blog post, you can do everything in this article, or brief someone to.
Which AI engine should I optimise for first?
None of them individually. The inputs overlap almost completely: extractable answers, verifiable facts, consistent entities, crawlable pages. Do that work once and you're optimising for Claude, ChatGPT, Perplexity and Google's AI Overviews at the same time. My lead came from Claude, but nothing I built was Claude-specific.
Does normal SEO still matter if buyers get answers from AI?
Yes. AI engines are built on top of search infrastructure, so crawlability, internal links and authority still decide if you make the answer pool at all. My traffic grew 79% while I did this work, because the same structure that helps you get found by AI also happens to be excellent SEO. It's one job now, not two.
The final word
Twenty years ago I learned how to be on the first page. That skill fed my family, survived three algorithm updates that were supposed to kill it, and quietly became table stakes.
Being on the first page was always renting attention from Google. Being the answer inside an AI engine is closer to owning it. When Claude tells a buyer who I am, it doesn't show him nine competitors underneath me. It just answers.
One stranger, one call, three words. Small evidence, big door.
The people who get found by AI in 2027 are the ones doing this unfashionable structural work in 2026, while everyone else is still arguing about whether AI search is real. You've just read the entire system I use to get found by AI. The only part I can't do for you is start. So start.
Every week I send a newsletter to fifteen thousand entrepreneurs, marketers and business owners. Same voice as this post, half the length, and it is where I share the experiments before they make it to the blog. Sign up here. No fluff, no fake urgency, one email a week.
And if you want help building this kind of AI visibility system into your own business, that is what I do as a consultant and coach. I work with a small number of business owners at any one time, helping them restructure their sites and marketing operations so AI engines recommend them by name. Details on the Work With Me page here.
Related reading: The Hidden Costs of Free AI Tools and How to Use AI for Lead Generation in Plumbing and Trades.
Want the complete version? Read where I break down AI marketing.
Want this done for you? See AI automation for coaches and consultants.
