AI search visibility

A friend who runs a small software company asked me a question last month that I couldn’t answer off the top of my head. “When someone asks ChatGPT for the best tool in my category, does it ever say my name?”

So we opened ChatGPT and asked it. It named five tools. His wasn’t one of them. He’d spent three years and a lot of money ranking on page one of Google for those exact terms, and the fastest-growing way people now search had never heard of him.

That’s the whole problem in one sentence. Search is quietly splitting in two. Some people still type into Google and scroll a list of links. More and more just ask ChatGPT, Perplexity, or Google’s own AI summary, read the one answer it gives back, and never scroll at all. If your brand isn’t inside that answer, you’re invisible to those people, no matter how well you rank.

Let me say the plain version of the jargon before we go further, because two ugly acronyms are about to show up a lot. Generative engine optimization (GEO) is the work of getting your brand mentioned and cited inside AI-generated answers. Answer engine optimization (AEO) is the close cousin: shaping your content so a machine can lift a clean, direct answer straight out of it. Same goal from two angles. Show up when the answer is written by a model, not picked from a list.

And the honest take, the one this whole guide rests on: GEO isn’t a new dark art. It’s mostly the same fundamentals you already know (clear answers, clean structure, real authority) pointed at a different reader. The reader is now the model.

This isn’t a niche concern you can park for next year. The size of the shift is already enormous. ChatGPT crossed 800 million weekly users in October 2025, by Sam Altman’s own count at OpenAI’s developer day (reported by TechCrunch). Google’s AI Overviews, the summary box at the top of normal Google results, reached two billion monthly users by mid-2025 (Sundar Pichai on the Q2 earnings call, via TechCrunch). That’s not a fringe behavior. That’s most of the internet’s search habit, moving.

And it’s already taking clicks off the table. Pew Research watched 900 real US adults browse and found that when an AI summary showed up, people clicked a normal search result only 8% of the time, versus 15% when there was no summary (Pew Research, July 2025). Roughly half the clicks, gone, swallowed by the answer. So ranking is no longer the finish line. Getting named in the answer is.

I’ll be straight about where I’m standing while I write this. I’ve spent real time inside this problem, helping a brand figure out whether the models were recommending them and what to do about it. So this isn’t theory. It’s the map I actually use, and the places it gets murky, which it does. I’ll flag what’s solid and what’s still a guess.

What GEO and AEO actually are (in plain words)

Let me kill the mystique first, because the acronyms make this sound harder than it is.

For twenty years, the deal with search was simple. You wrote a good page, search engines ranked it, and people clicked through to read it. The search engine was a librarian pointing at a shelf. You optimized to be the book it pointed at.

The model is a different kind of reader. It doesn’t point at your shelf. It reads everything, then writes its own short answer and quietly decides which two or three brands to name inside it. So the job changes. You’re no longer trying to be the top link a person clicks. You’re trying to be the source the model trusts enough to quote.

That’s it. That’s the whole shift. GEO is the broad version of that work (be the brand the AI cites). AEO is the narrower, more technical slice (write so a machine can extract a clean, self-contained answer). People use the terms loosely and often interchangeably, and honestly the difference matters less than the shared goal.

What does the work actually look like? Mostly things you’d recognize. Clear writing that answers the real question instead of burying it. Clean structure a machine can parse: headings that match how people ask, short direct answers near the top, a real FAQ. Genuine authority: other people on the web mentioning you, third-party lists you appear on, a reputation that exists outside your own website. None of that is exotic. It’s the same craft, with the reader swapped out.

My take: A lot of the GEO industry wants this to feel like a secret new skill you have to buy, because “we’ve cracked a mysterious new channel” sells better than “do the fundamentals well, but aim them at the model.” Don’t buy the mystery. The brands winning AI citations are mostly the ones who were already clear, well-structured, and genuinely talked about. GEO is less a new discipline than a new reason to take the old ones seriously.

GEO vs AEO vs SEO: what each one is for

These three get mashed together, so let me lay them side by side. The cleanest way to tell them apart is to ask: who’s the reader, and what does winning look like?

The thing to notice: these aren’t rivals. They’re layers of the same job. A page that ranks well (SEO) is structured to answer a question cleanly (AEO), and is talked about across the web (GEO) tends to win all three at once. They feed each other.

There’s good evidence for that overlap, not just a hunch. BrightEdge tracked AI Overview citations against organic rankings for sixteen months and found the overlap climbing from 32% to 54.5%, meaning over half the sources the AI cites also rank well the old-fashioned way (BrightEdge, 2025). So a chunk of GEO is just SEO doing its job. Google says this part out loud: in its official guidance, optimizing for its AI features “is still SEO,” because those features run on the same core ranking systems (Google Search Central).

But, and this is the part that makes GEO its own thing, that BrightEdge number cuts both ways. If only half of AI citations overlap with top organic rankings, the other half don’t. Plenty of brands get named in AI answers without ranking number one for the query. Which is the most useful fact in this entire guide.

My take: I’d push back on anyone who tells you GEO is either “totally new and separate from SEO” or “just SEO, nothing to see here.” Both are wrong. It’s mostly SEO fundamentals, with a real extra layer on top: the model weighs what the rest of the web says about you, not just your own pages. So you can’t fake it with on-page tricks alone, and you can’t ignore it as “already covered by SEO.” It’s the same foundation with a different roof.

How to get cited by ChatGPT, Perplexity, and Google AI

This is the part everyone actually wants. So let me give you the real levers, in rough order of how much they matter, and lean on data instead of vibes wherever I can.

1. Answer the question directly, near the top. The model is lazy in a useful way: it wants a clean, self-contained answer it can lift. If your page makes it dig, it’ll find an easier source. So put a plain two-or-three-sentence answer right under the heading, then go deep below it. Write headings the way people actually ask (“how do you check if your brand is cited in AI answers”), not in clever marketing language.

2. Back your claims with sources, quotes, and numbers. This isn’t a guess. It’s the strongest finding from the original academic research on the topic. A team from Princeton and others ran controlled tests on what makes content more likely to show up in AI answers, and the three tactics that worked best were adding cited sources, adding direct quotations, and adding specific statistics. Each lifted visibility by roughly 30 to 40% (Aggarwal et al., “GEO: Generative Engine Optimization,” KDD 2024). Worth sitting with: the same study found that keyword stuffing, the old spammy move, actually lowered visibility by 22%. The models reward content that reads like it knows what it’s talking about, and punish content that reads like it’s gaming them.

3. Build authority off your own website. This is the lever SEO people underrate and the reason GEO isn’t “just SEO.” Models weigh what the broader web says about you. Rand Fishkin’s team at SparkToro tested nearly 3,000 prompts and found AI tools lean heavily toward brands that show up often in documents across the web, and that mentions on Reddit and YouTube carry real weight (SparkToro, 2025). So getting onto the third-party “best X tools” lists, being discussed in your niche’s communities, earning unlinked mentions, all of it feeds the model’s sense that you matter. You can’t write that into your own footer. You have to earn it out there.

4. Get the boring technical bits right. Let the AI crawlers in (don’t block them in robots.txt), use clean headings and lists, add the kind of structured data that tells a machine what your page is. None of this is glamorous and none of it is enough on its own, but it removes the friction that quietly keeps you out.

The pattern across all four: there’s no trick. It’s clarity, evidence, reputation, and not getting in your own way. The same things that made content genuinely good before, now with a machine reading it.

How to check if your brand is even cited

Before you change a single thing, find out where you stand. Most brands never do this, which is wild, because it’s the cheapest, fastest part of the whole job. You can do a rough version yourself this afternoon.

Here’s the manual check, the same one I ran with my friend:

1. Write down the real questions a buyer asks. Not your brand name. The questions people ask before they know you exist. “Best [your category] tool for small teams.” “[Competitor] alternatives.” “How do I [the problem you solve].” Aim for ten to fifteen.

2. Ask each one in a fresh, logged-out session of ChatGPT, Perplexity, and Google’s AI mode. Logged out and fresh matters, because the engines personalize. You want the answer a stranger gets, not the one your own history nudges.

3. For each answer, write down three things. Are you named at all? If you are, is what it says accurate and flattering, or stale and wrong? And which competitors get named instead of, or ahead of, you?

4. Read the rest of the web in the answer. Notice which sources the AI is pulling from: a particular Reddit thread, a “top 10” listicle, a review site. Those sources are the doors into the answer. If you’re missing from them, that’s your fix list.

That’s a real audit. It tells you the only thing that matters at this stage: does the AI recommend you, and if not, what is it reading instead? There are paid tools that track this at scale and over time (which you’ll want eventually, because answers shift week to week), but you don’t need one to learn whether you have a problem. You need fifteen good questions and twenty quiet minutes.

My take: The reason I push people to do the manual check first, before buying any tool, is that it makes the problem real in a way a dashboard never does. Watching ChatGPT confidently recommend three competitors and skip you is a different feeling than reading a “visibility score.” It tends to turn “I should look into GEO someday” into “okay, what do we fix this week.” Do the twenty-minute version yourself at least once.

Can you win AI search without ranking number one?

This is the contrarian heart of the whole thing, so let me make the case slowly.

The old game was brutal for small brands. To get traffic from a competitive term, you basically had to outrank everyone, which meant out-spending and out-waiting bigger competitors for years. The number-one spot was the prize, and it was usually taken.

The new game has a different shape. Remember that BrightEdge finding: only about half of AI citations overlap with the top organic rankings (BrightEdge, 2025). The model isn’t just reading the top of Google and copying it. It’s synthesizing from a wider pool: forum threads, “best of” lists, comparison articles, places you can earn your way into without winning a years-long ranking war.

So the move for a smaller brand isn’t “outrank the giant.” It’s “be the clearly-explained, well-cited, frequently-mentioned answer the model reaches for.” Be the one with a genuinely clear page on the exact question. Be the one mentioned in the Reddit thread the AI keeps pulling from. Be on the third-party list. You can do all of that without ever touching the number-one organic spot.

Count it out with me. If the AI answer names two or three brands, and citation only half-depends on top rankings, then there’s a real seat at the table that isn’t won by raw authority. That seat goes to clarity and reputation. Which a focused small brand can build a lot faster than it can climb to number one.

I want to be honest about the limit, though. “You don’t need to rank number one” does not mean “you can ignore SEO.” The two are tangled. Strong rankings still feed citations, and a brand nobody links to or talks about anywhere will struggle no matter how clean its pages are. The point isn’t that authority stopped mattering. It’s that the path to getting cited is wider than the old one-spot race, and that width is the opening.

The wedge path: audit, fix, then keep watch

Let me put the whole thing into a path you can actually walk, because “do good fundamentals” is true but not very actionable on a Monday morning.

Step one: audit. Run the manual check above, or have someone run a deeper version. The output is a clear picture: which buyer questions you show up for, which you’re invisible on, what the AI says about you when it does name you, and which competitors are eating your lunch. This is the cheap, high-value step, and most brands skip straight past it to “let’s make content,” which is backwards. You can’t fix a gap you haven’t found.

Step two: fix the obvious gaps. With the audit in hand, the fixes usually sort themselves into a short list. Pages that bury the answer get a clean direct answer up top. Claims without evidence get real sources, quotes, and numbers (the best-tested lever, remember). And the off-site holes get worked: the listicles you’re missing from, the communities where your category gets discussed, the comparison pages that name everyone but you. This is the part that takes real effort, and it’s also where most of the win lives.

Step three: keep watching. This is the step people forget, and it’s why GEO is ongoing rather than one-and-done. AI answers aren’t fixed. They shift as the web changes, as the models update, as competitors do their own work. The same Ahrefs data that tracks this shows the click impact getting bigger over time, not smaller (Ahrefs, 2026), which tells you the ground is still moving. A brand that audited once in 2025 and walked away is already out of date. So you check it on a rhythm, monthly is plenty for most, and you keep the answer accurate and you keep earning the mentions.

That’s the wedge. Audit to see clearly, fix what’s broken, then watch it so it stays fixed. It’s not magic and it’s not a one-time project. It’s a small, repeatable habit that compounds, exactly like good SEO always was.

This page is the starting point for all of it. As I publish deeper, hands-on essays on each piece (how to run the audit properly, how to write for AI extraction, how to earn the off-site mentions), they’ll live under GEO and AEO. For now, the guide above is the map.

Frequently asked questions

What is generative engine optimization (GEO)?

Generative engine optimization (GEO) is the practice of getting your brand mentioned and cited inside the answers that AI tools like ChatGPT, Perplexity, and Google’s AI Overviews generate. Instead of optimizing to rank in a list of links a person clicks, you’re optimizing to be one of the sources a model trusts enough to name in its own written answer. In practice GEO is mostly familiar fundamentals (clear answers, clean structure, real authority and mentions across the web) aimed at a new reader: the model. The original academic research found that backing your content with cited sources, direct quotes, and specific statistics raised visibility in AI answers by roughly 30 to 40%.

What’s the difference between GEO, AEO, and SEO?

SEO (search engine optimization) gets your page ranked in the list of links so a person can click it. AEO (answer engine optimization) structures your content so a machine can lift a clean, self-contained answer straight out of it. GEO (generative engine optimization) gets your brand cited inside an AI’s own written answer, weighing not just your page but what the whole web says about you. They overlap heavily and the same well-built page can win all three. The difference is the reader: a person who clicks, a snippet engine that extracts, or a model that writes prose and chooses who to name.

How do you get cited by ChatGPT or Perplexity?

You get cited by doing four things well. First, answer the real question directly and near the top of the page, so the model can lift a clean answer. Second, back your claims with cited sources, direct quotes, and specific numbers, which is the strongest tested on-page lever. Third, build authority off your own site: get onto third-party “best tools” lists, get discussed in communities like Reddit, earn mentions across the web, because models weigh what others say about you. Fourth, get the technical basics right by letting AI crawlers in and using clean, machine-readable structure. There’s no trick to it. It’s clarity, evidence, and reputation.

Can you rank in AI search without ranking number one on Google?

Yes. Research tracking AI citations against organic rankings found only about half of cited sources also rank in the top organic spots, which means plenty of brands get named in AI answers without holding the number-one position. Models synthesize from a wider pool than the top of Google, including forum threads, comparison lists, and review sites you can earn your way into. So a smaller brand can win a citation through clarity and reputation rather than raw ranking authority. It does not mean SEO stops mattering, since strong rankings still feed citations, but the path to getting cited is wider than the old single-spot ranking race.

How do you check if your brand shows up in AI answers?

Run a manual check in about twenty minutes. Write down ten to fifteen real questions a buyer asks before they know you (best-tool questions, competitor-alternative questions, problem questions), then ask each one in a fresh, logged-out session of ChatGPT, Perplexity, and Google’s AI mode. For each answer, note three things: whether you’re named at all, whether what it says is accurate and current, and which competitors get named instead. Also note which web sources the AI is pulling from, because those are the doors into the answer. Logged-out and fresh sessions matter, because the engines personalize and you want the answer a stranger gets.

Is GEO worth doing for a small business?

For most small businesses, yes, and the cheapest part (the audit) is worth doing no matter what. AI search is already huge and growing, with ChatGPT past 800 million weekly users and Google’s AI Overviews past two billion monthly users, and AI summaries are measurably cutting the clicks that used to reach websites. The opening for a small business is that you don’t need to outrank giants to get cited, because models weigh clarity and reputation across the web, not just ranking position. Start with the free twenty-minute check to see whether you have a problem. If the AI is recommending competitors and skipping you, that’s a real, fixable gap and worth the effort.

Want to know if AI recommends you?

If you read this far, the obvious next question is the one my friend asked me: when someone asks ChatGPT or Perplexity for the best option in your category, does it ever say your name? Most brands have genuinely no idea, and not knowing is its own risk.

I’m happy to just find out for you. I’ll run a quick look at how you show up across the main AI engines (named or not, what it says, who’s beating you) and send back a short, plain list of what I’d fix first. No deck, no pitch, no commitment. Just a straight read on whether the AI is recommending you or quietly sending people to someone else.

If that’s useful, come grab a slot and tell me your category and a couple of competitors. That’s all I need to start looking.