Case Study: How 1 Niche Site Earned 144,321 AI Citations in a Quarter

I built BakingSubs to test one question: can a niche site become the source AI engines quote, without ads, backlinks, or social? Over 12 months it earned 144,321 Microsoft Copilot citations in its peak quarter and pulled 5,000+ daily Google clicks. Here is what worked, what I never did, and how a coach or consultant copies the parts that matter.

Key takeaways

• BakingSubs hit 144,321 Microsoft Copilot citations in a single quarter and 5,000+ daily Google clicks, in 12 months, with no ads, no backlinks, and no social.
• The mechanism is the Citation Cluster Method: groups of posts that each answer one specific question a buyer would type into an AI engine, all linked around a clear topic.
• I never built a backlink campaign, never ran ads, never posted on social. Those are not part of the model.
• For coaches and consultants, the same approach works when each post answers one real buyer question (not a topic, a question) and your About page makes it obvious who you are.
• Copilot rewarded the site fastest. ChatGPT and Perplexity followed once the cluster had enough depth that no single page was the only answer.
• Do not copy the recipe volume. Copy the structure: one clear expert, tightly clustered questions, plain answers, real opinions.

What BakingSubs actually is, and why a baking site matters to coaches

BakingSubs is a niche site about ingredient substitutions in baking. Nothing about coaching. The reason it matters to you is that the mechanism that made it cited has nothing to do with the niche. It works because of how the content is shaped, not what the content is about.

Here is the shape. Every page answers one specific question someone would type into ChatGPT or Copilot. "Can I use Greek yogurt instead of sour cream in cheesecake." "What happens if I swap butter for oil in banana bread." Each answer is direct in the first sentence, then expands with the why, the ratio, and the failure case. No long intros. No personal stories at the top. The answer first.

A coach's site has the same job, with different questions. "Is a life coach worth it at 35." "How does executive coaching work if my company is paying." "Can a business coach help if I am pre-revenue." Each of those is a real buyer query right now. Most coaching sites answer none of them on a dedicated page. They have a Services page, an About page, a blog full of mood-board posts, and that is it. AI engines have nothing to quote.

The 12-month timeline, in plain numbers

I want to give you the real arc so you can calibrate expectations. Month one through three: nothing. Indexed but invisible. I kept publishing. Month four through six: Copilot started citing single pages, a few times a week, then a few times a day. Month seven through nine: citations started compounding because the cluster was deep enough that one question linked naturally to four others. By month ten through twelve, Copilot citations hit 144,321 in the quarter and Google traffic was running over 5,000 clicks a day.

What changed at month four was not a tactic. It was that the cluster crossed a depth threshold. When you have 8 posts on one tight topic, AI engines see one decent source. When you have 40 posts that all reference each other and all answer different but adjacent questions, AI engines start treating you as the source on that topic. Most coaches I talk to give up at month three because the dashboard is empty. The results start to snowball right after that point, not before.

For coaches the timeline is broadly similar. A life coach who picks one specific audience (say, women returning to work after maternity leave) and publishes 25 to 40 tight question-answer posts over 6 to 9 months will start to see citations in month four to seven. Not faster. Not slower. The threshold is real.

What I published, and what a coach version looks like

Every post on BakingSubs has the same skeleton. The question as the H1. A direct answer in the first one or two sentences. A short table of the ratio or the swap. A "when this works" section. A "when this fails" section. A short FAQ. That is it. No 300-word intro about my grandmother's kitchen.

For a coach, the skeleton transfers almost cleanly. Take an executive coach working with new VPs. A post titled "Does executive coaching work for first-time VPs" should open with: "Yes, for two specific reasons, but only if the company is paying and the coach has worked with VPs in your industry before." Then a short section on when it works. Then a short section on when it does not. Then an FAQ.

The hardest part is the "when it does not work" section. Coaches almost never write these. They are trained in sales copy that says yes to everyone. AI engines reward the opposite. When Claude sees a page that says "this is not for you if X, Y, Z," it treats that page as more trustworthy than the page that promises everyone results. That is one of the strongest signals you can send, and most coaches refuse to send it. Health coaches in particular face a higher bar here because AI engines weight health-related content more carefully.

What I never did, and why you can skip it too

I never built a backlink. Not one outreach email. Not one guest post. The site has organic backlinks because pages got cited, but I did not chase them. Most SEO advice from 2019 to 2023 told you to spend 80% of your time on link building. That advice is now actively wrong for expert-led sites, because AI engines weight on-page signals, author identity, and topical depth far more than link counts.

I never posted on social. No LinkedIn, no Instagram, no Twitter. Zero. The site grew because people asked Copilot a question and Copilot quoted BakingSubs. That is a different traffic source than social entirely.

I never ran ads. Not search, not display, not retargeting. The economics of paid traffic for content sites have been getting worse for years and AI engines do not care whether you run ads.

For coaches this is the part that lands hardest. You have been told for a decade that you need to be on LinkedIn daily, that cold DMs work if you do enough of them, that referrals dry up unless you stay top of mind. When referrals dry up the real fix is not more outreach, it is making sure that when a buyer asks ChatGPT "who should I hire," your name shows up. Outreach is a tax you pay because your site is invisible. Fix the site and the tax goes away.

The three things to copy, ranked

If you copy nothing else, copy these three, in this order.

One: pick one specific audience and write only for them. BakingSubs is for people doing substitutions in home baking. Not pastry chefs. Not bread bakers. Substitutions. That is it. The narrower the audience, the easier it is for AI engines to decide who to recommend you to. A "coach for high-performing professionals" is invisible. A coach for "first-generation immigrant women in tech leadership roles" is highly citable, because the description is specific enough that an AI engine can match it to a query.

Two: write one post per question, not one post per topic. A "topic" post tries to cover everything about an area in 3,000 words. A question post answers one specific thing in 1,500. AI engines pull from question posts because the question matches the buyer's prompt almost word for word. Topic posts get ignored. The pattern of building tight clusters around real questions is the heart of the method.

Three: make your About page a Person, not a brand. BakingSubs has a real person with a real face on the About page, with the right schema tags so Claude and Copilot can tell it is a human, not a faceless brand. Most coaching sites read like agencies. AI engines do not recommend agencies when buyers ask for "a coach." They recommend people.

What NOT to copy from BakingSubs

A few things worked for the recipe niche that would actively hurt a coach.

Do not publish at high volume just to publish. BakingSubs has hundreds of pages because the niche has hundreds of distinct ingredient substitutions. Coaching does not. A coach who tries to publish 200 posts will start writing thin content that ChatGPT explicitly downweights. Thin content gets ignored by ChatGPT and dragging down a strong site is easier than people realize. For most coaches, 25 to 50 deep posts in a tight cluster beats 200 shallow ones.

Do not skip the personal voice. BakingSubs is mostly transactional content where the reader wants a clear answer, not a personality. Coaching is the opposite. Buyers are hiring you, not your method. Your posts need to read like you. Your opinions, your judgment calls, your "I disagree with the common advice here because..." moments. Those are what get cited when a buyer asks ChatGPT "who has a different take on X."

Do not assume one channel is enough. BakingSubs got disproportionate love from Copilot first. ChatGPT followed later. Perplexity later still. If you optimize for only one engine you will be fragile. The Citation Cluster Method works across engines because the underlying signals (clear author, tight topic, specific answers, real opinions) are what all four reward.

How a coach starts this week

If you want to run a smaller version of this experiment, here is the first week.

Day one: write down the single most specific audience you can serve. Not "professionals." Not "leaders." A specific kind of person, in a specific stage, with a specific problem. Day two: write down the 20 questions that person would type into ChatGPT in the week they decide to hire someone. Day three: rewrite your About page so a 12-year-old reading it can tell who you are, who you help, and what makes you different. Day four through seven: write the first three posts, each answering one of those 20 questions, each with the skeleton above.

Then run the AI Visibility Check and see which of those questions you already show up for in ChatGPT, Claude, Perplexity, and Copilot. The gap between where you are and where you want to be is what the rest of the plan attacks. For the deeper system, The AI Citation Playbook walks through the same build I did on BakingSubs, translated for an expert-led service business.

Frequently asked questions

How long does it take to get cited by ChatGPT or Copilot as a coach?

For most coaches building a tight cluster from scratch, first citations land in month four to seven. That assumes 25 to 40 question-answer posts on one specific audience, a clear Person-based About page, and direct answers in the first sentence of each post. Less work means a longer wait, not a worse outcome.

Do I need backlinks to get cited by AI search engines?

No. BakingSubs hit 144,321 Copilot citations in a quarter without a single outreach email. AI engines weight author identity, topical depth, and clear on-page answers far more than link counts. Backlinks are not harmful, they are just no longer the lever. The bigger SEO shift is worth understanding before you spend time on link building.

Can I copy the BakingSubs model if my niche only has 20 possible questions, not 200?

Yes, and that is actually the normal case for coaches and consultants. You do not need 200 posts. You need 25 to 50 deep ones that all interlock. Volume past that point hurts more than it helps, because thin content drags down strong content under the helpful-content classifier.

What if my site is brand new with zero traffic?

Same plan, same timeline. BakingSubs started at zero. Traffic is a downstream effect of the structure, not a prerequisite. What you need at month zero is one clear audience, one clear author, and the willingness to write one question-answer post per week for six months.

Do I still need to post on LinkedIn or Instagram?

Not for this to work. I did zero social and the citations still happened. If you enjoy posting on LinkedIn or it brings you real conversations, keep doing it for those reasons. But do not pretend it is the engine. The engine is the site, the cluster, and the answers.

The shortest version of all of this: pick one audience, answer their real questions one post at a time, and make it easy for an AI engine to tell who you are. The numbers I hit on BakingSubs were not a recipe-niche fluke. They were a structure that the engines reward, and the structure works for coaches and consultants too. Run the visibility check, see where you stand on the four engines, and decide which question to answer first.