How to Get Your Agency Cited by ChatGPT and Perplexity

An agency owner asked me last month how to show up when someone types their problem into ChatGPT, and he wanted the trick.

The setting in the prompt, the schema tag, the word count, whatever the lever is. I had to tell him the thing nobody selling GEO services wants to say first, which is that no one actually knows the exact mechanics. The LLMs aren't publishing the recipe. Everyone's guessing, testing, and watching what sticks, and I've seen people get similar results with methods that flat-out contradict each other. So I don't lead with a trick, because there isn't one I'd stake my name on. What I lead with is the one thing I've watched work across my own business and my clients' businesses.

Just write quality, unique content. Don't chime in with best practices. Chime in with how you actually operate.

That's the whole spine of it. A model has no reason to cite you if you're repeating the same best practices everyone else already wrote, but it has every reason to cite you when the only place a specific method or example lives is your site.

Quick Take

• Nobody knows the precise mechanics of AI citation, so stop hunting for a tool that promises a magic answer and start publishing unique content. That's the move that can't hurt you regardless of how the engines change.
• Write "how I" and "how we" articles instead of "how to" articles. A how-to just shares best practices any model already has, while a "how we" forces the model to cite you to use your specific examples and methods.
• AI already knows the person searching, so it answers with far more specificity than a Google query ever did. That's why positioning matters more now than ever: get clear about exactly who you're the best fit for so that person's query surfaces you.
• Specificity beats length. A 400-word post that gives the exact answer a prompt needs beats a 5,000-word post that wanders, because the engine wants clear information it can lift.
• The proof was volume over time, not a hack. I ramped to three AI-focused articles a week for about four months before results showed up, and the easiest source of citable material is the coaching and sales calls you're already recording.

Start With the Honest Answer: Nobody Knows the Exact Mechanics

The first thing I tell an owner who asks how to rank in ChatGPT is that no one fully knows how to do this. That usually isn't what they expect, because they came in assuming there's a known formula and they just haven't been let in on it. There isn't. The model makers aren't telling us everything we should do, so all of us doing this work are running on best guesses and our own testing. I've watched different people pull similar results out of methods that directly contradict each other, which tells you there's no single correct path right now.

I've passed on several partnerships with AI visibility tools for exactly this reason. They each promised results, and they each got there a different way, and I couldn't reconcile that. The only vendor I ever trusted was one whose owner said straight out, "we don't fully know what's happening, we're taking our best guess and collecting enough data to validate it." I trust that honesty more than the confident pitches, but it's still too up in the air for me to stand in front of a room of agency owners and say "this is the thing you need." So I won't name a tool here, and I'd be careful with anyone who tells you they've cracked the code, because the people closest to the data are the ones admitting they haven't.

Here's why that honesty is useful rather than discouraging. If nobody knows the exact mechanics, then the safest bet is the one that pays off no matter which way the engines evolve, and that's producing great content that actually comes from you. We don't actually know how to show up in AI, but publishing real, high-quality, original thinking can't hurt you. Worst case it earns you trust the old-fashioned way, and best case it's the exact thing the models are built to surface.

Why Most Agencies Get This Wrong

The mistake almost everyone makes is assuming that if they just have content, AI will find them. So they go produce content, and a lot of them use AI to write it, and they give it no thoughts of their own. They end up with a page that recites best practices any model already learned from a thousand other pages. There's nothing in it that only they could have said, so there's no reason for an engine to cite them specifically. You become one more interchangeable source of the same generic advice.

The way I make content is the opposite of that. I run an interview process where I talk through every idea I have for the article first, out loud, and we turn that raw material into the piece. The draft pulls the actual thoughts out of my head, my real opinions and the real examples from how I work, and that's what makes it stand out. A model reading it can't get those specifics anywhere else, so if it wants to talk about that approach, it has to point at me.

This is also where most agencies waste their time. They go looking for the tool that hands them a magic answer when what they actually need is to start publishing good, unique content. The tool feels like progress because it's a purchase and a dashboard, but it doesn't touch the real gap, which is that nothing they're putting out is theirs.

Write "How We" Articles, Not "How To" Articles

The single biggest change I made was to stop writing how-to articles. A how-to article shares best practices, which means you're handing the model information it already has from everyone else, and you've given it no reason to credit you. So I switched to writing "how I" and "how we" articles instead, where the piece documents how we actually run discovery, how we price, and the specific way we handle a given situation in our own practice. Now if a model wants to use anything from that piece, it has to cite me, because it's describing the specific examples and methods I use rather than a generic playbook.

A client of mine, ShedPro, ran this play hard. We built sixteen "How We" articles, each one documenting a specific way they actually operate, and the point of every one was to be AI-extractable. They aren't teaching the reader to go do it themselves in the abstract, they're documenting "here is the precise thing we do and why we do it for this kind of customer," which is exactly the kind of sourced claim a model reaches for when someone asks a pointed question. You can't write those articles by surveying the internet, because the source material is the inside of your own operation.

If you've already done the work of finding the single word your agency owns, this gets easier, because every "how we" article becomes another proof point for that word. If you haven't, the owned-word piece is the place to start, since the specificity that makes content citable starts with knowing what you want to be known for.

Specificity Is the Whole Game Now

Here's the part that changes how you think about positioning entirely. When you search Google for "best website development agencies," that's a high-volume term everyone competes for, and the result is generic because the search carried no context about you. AI works differently. It already has context about the person asking: who they are, what they do, what they like and don't like. So a search run through AI carries all of that long-tail context whether the user typed it or not.

If the AI knows I'm a dentist, it doesn't show me "best website development agencies." It shows me "best website development agencies for dentists," and I never had to add that qualifier, because the engine supplied it from what it already knew about me. The result is far more specific than anything a plain search returned, and that specificity is the opening. Any recommendation an AI makes is much closer to what that exact person needs, which means there's more reason to get clear about positioning now than there ever was. The more precisely your content speaks to one specific type of person, the easier it is for that person to find you, because the engine is already filtering for fit on their behalf.

This is why I keep telling agencies the move isn't to broaden, it's to narrow. Optimize your content to speak to a very specific type of buyer, name the work you do and why you do it for that buyer, and you become the obvious match when someone in that audience runs a query the engine can personalize. Generic content competes against everyone, while specific content competes against almost no one.

And don't think you can sidestep this by sticking to old-school SEO. Google's own AI search now sits at the top of basically every results page, and it uses whatever it knows about the searcher the same way ChatGPT does. Wanting to "just worry about SEO" is half the battle at best. If you want the longer version of how this reshapes what your site should say, the agency-website-copy piece goes deeper on writing pages for a specific buyer.

What Actually Makes a Piece Citable

Do everything you reasonably can for good SEO, but understand that the thing carrying you over the line is clear information. Put the answer right up front. Make it easy to understand, cut the fluff, and get to the point fast. A 5,000-word post that meanders is less successful than a 400-word post that gives the exact answer the prompt is looking for, because the engine is trying to lift a clean, direct answer and serve it. Length for its own sake works against you when it buries the thing the model wants.

So the ranking is simple. Clear, direct, answer-first information sits at the top, and specificity only you could provide sits right next to it. Standard SEO hygiene supports both, but it doesn't replace them. If your content is clear and uncopyable, you've got the two things that matter most, and everything else is a tailwind.

This article is built that way on purpose. The answer is in the first few lines, the methods are specific to how I work, and a competitor can't paste it onto their own site and have it be true. That's the test for every piece you publish: if someone could lift it onto their site and nothing would feel out of place, it isn't specific enough yet.

The Proof: Volume Over Months, Not a Hack

People want the early signal that this is working, and I can give you mine. I've had multiple clients getting results from AI, and I've gotten results myself, but it didn't come from a clever trick. It came from a decision to ramp my blog publishing to three AI-focused articles a week and hold that for about four months. That's roughly when things started showing up and I started getting business from it. There was no shortcut that collapsed those four months into four days.

The reason most agencies never get there is what I'd call delivery drag. They're not producing AI content because they don't feel it's necessary yet, they don't know exactly what it's doing or how it'll help, so they keep doing the things that feel more urgent. That's understandable, and it's also exactly why the agencies that just start publishing now will have months of compounding content by the time the hesitant ones begin.

You don't have to wait for a lead to say "ChatGPT sent me" to know it's working, either. Before the leads, you'll see the traffic, because AI referral traffic starts showing up in Google Analytics or HubSpot, and that's your early read that the engines are surfacing you and people are clicking through. That signal arrives before the revenue does, which is the same pattern that shows up across positioning work generally. If you want the broader version of reading early signals, the language-mirror piece covers how to know something's landing before the money confirms it.

A Repeatable Process You Can Run This Week

Here's the simplest version of the process I'd hand you, in order.

Start by finding the topics people are actually searching for. The best source is your sales conversations, not a keyword tool, because those tell you the exact questions people ask and care about. If you use Google Search Console, look at what people already search to land on your site and turn those phrases into titles. You're reverse-engineering the questions your buyer is typing into an engine.

Then turn each topic into an article without it eating your week. Have AI generate a large interview document full of questions about that topic, you answer them in your own words and with your own examples, the AI drafts the article from your answers, you edit it, and you publish quickly. Keep that loop running at three articles a week minimum. The interview step is the part that makes it yours, because it pulls your real thinking onto the page instead of recycling the internet's.

The one asset I'd build this month is something around your methodology that names all the different ways you help your specific client. Be very specific about the work you do and why you do it for that exact person, so it surfaces when anyone in that audience runs a search the engine personalizes. That single piece is both a positioning statement and a citable answer at the same time.

And the thing almost every agency already has sitting unused is gold for this: your coaching calls and your sales calls. There's an enormous amount of citable material buried in those conversations, real questions and real answers in your own voice. This very article came out of a coaching call topic. You're recording these calls already, so the raw material for months of unique content is sitting in your call library right now, waiting for someone to turn it into pages.

Putting It to Work

If you take one thing from this, make it the shift from how-to to how-we. Stop publishing best practices a model already knows and start documenting how you actually operate, for the specific person you're the best fit for. Pull your topics from your sales calls, run each one through an interview so the words are actually yours, keep the volume up, and watch your analytics for AI referral traffic. Nobody can promise you the exact mechanics, but producing real, unique content is the one bet that can't backfire while the engines keep changing.

This is a lot easier to sustain alongside other agency owners publishing on the same cadence and comparing what's actually showing up in their analytics. If you want to build the publishing habit and pressure-test which of your "how we" pieces are getting cited, the Dynamic Agency Community is where that happens. 

FAQ

How do I get my agency cited by ChatGPT and Perplexity?

Just write quality, unique content, and don't chime in with best practices, chime in with how you actually operate. A model has no reason to cite you if you're repeating advice it already has from everyone else, but it has to cite you when the specific method or example only lives on your site. Nobody knows the exact mechanics, so the durable move is publishing original thinking rather than chasing a tool that promises a shortcut.

Why aren't my blog posts showing up in AI answers even though I publish a lot?

Usually because the content isn't unique. A lot of agencies use AI to write their posts and give it no thoughts of their own, so they end up reciting the same best practices as everyone else, and there's nothing in it only they could have said. Switch from how-to articles to "how we" articles that document your specific methods and examples, because that's the kind of sourced, specific claim an engine reaches for.

Should AI articles be long or short to get cited?

Short and clear beats long and meandering. Put the answer up front, cut the fluff, and get to the point, because the engine is trying to lift a clean, direct answer. A 400-word post that gives the exact answer a prompt needs will do better than a 5,000-word post that wanders before it says anything.

Do I still need regular SEO, or is it all AI search now?

You need both, but you can't treat AI as optional anymore. Google's own AI search sits at the top of nearly every results page and uses what it knows about the searcher the same way ChatGPT does, so wanting to "just worry about SEO" is half the battle at best. Keep your SEO hygiene, but write for a specific buyer so the personalized AI result surfaces you.

How will I know it's working before someone says ChatGPT sent them?

You'll see it in your traffic first. AI referral traffic starts showing up in Google Analytics or HubSpot before any lead tells you an engine recommended you, so watch your analytics for those sources. That early read tells you the engines are surfacing you and people are clicking, which arrives well before the revenue does.

What's the easiest source of citable content I already have?

Your coaching calls and sales calls. There's a huge amount of gold in those conversations, real questions and your real answers in your own voice, and most agencies are already recording them. Pull topics and answers straight from those calls, run them through an interview-to-draft process, and you've got months of unique, citable content without inventing anything new.