SEO and GEO are not the same thing

A plumber we know got three new jobs last month from people who told him, point blank, that ChatGPT recommended his shop. He didn't run an ad. He didn't rank first for "emergency plumber" in his city. He'd just spent two years writing plain answers to the questions customers actually call about, and somewhere in that pile of training data and live retrieval, his name came up.

That's the shift worth paying attention to. A growing slice of buying research now ends before anyone sees a list of blue links. People ask ChatGPT, Perplexity, or Google's AI overview "who should I hire for X near me," read the answer, and act on it. If your business isn't part of that answer, you're invisible for that search, and you never even show up in the analytics to know it happened.

So people have started calling the work of showing up in those answers GEO, generative engine optimization. Some call it AEO, answer engine optimization. The acronym doesn't matter. What matters is that it's related to SEO but it isn't the same job, and treating them as one thing leads to wasted money on both ends.

What the two jobs actually are

Classic SEO is about earning a position on a results page. Someone types a query, Google returns ten links, and your goal is to be near the top of that list. The whole game assumes a list exists and that a human will scan it and click.

GEO is about being the source a model quotes when there is no list. You ask Perplexity a question and it writes you a paragraph, maybe with a few citations underneath. Nobody scrolls ten options. The model decided, on your behalf, which sources to trust and synthesize. The win isn't a rank. The win is getting named, cited, or summarized accurately inside that generated answer.

Different mechanics, different definition of winning. One puts you on a shelf next to nine competitors. The other tries to make you the answer.

Where they overlap

Here's the part that keeps this from being a fad you have to chase separately. Most of what makes a model want to cite you is the same stuff that's made good SEO work for fifteen years.

Clear structure helps both. A page with a real heading hierarchy, where the answer to a question sits right under a heading that asks it, is easier for Google to rank and easier for a model to lift cleanly. Credible, specific content helps both. Models, like search engines, lean toward pages that say something concrete and verifiable instead of fluff. Consistent facts across the web help both. If your business name, address, and what you do read the same on your site, your Google profile, and the directories you're listed in, both systems trust the signal more.

And page performance matters more than people expect on the GEO side. Retrieval systems crawl and fetch pages live to answer some queries, and a slow or flaky page is a page that gets skipped. We treat site speed and Core Web Vitals as table stakes for both, not a separate SEO chore.

So if you do the GEO groundwork and the AI thing somehow cools off, you haven't wasted the effort. You've improved your ordinary search presence at the same time. That's why we call it low-regret. There aren't many marketing bets where the downside is "well, your regular SEO just got better."

Where they genuinely differ

The overlap is real, but a few things are specific to GEO and worth doing on purpose.

Write for the question, not the keyword. Search optimization trained everyone to target phrases like "CRM software small business." People don't talk to chatbots that way. They ask "what's a good CRM for a 5-person plumbing company that hates spreadsheets." Content that answers the messy, full-sentence version of a question is what gets pulled into an answer. We spend time mapping the actual questions a buyer asks out loud, then making sure a page answers each one directly and early, in the first sentence or two, not buried in paragraph nine.

Make your facts machine-readable. This is where structured data earns its keep. Schema markup (the JSON-LD blocks that label a page as a product, a service, an FAQ, a local business, a review) gives a model unambiguous facts instead of asking it to infer them from prose. Your hours, your service area, your prices, your founding date. When those are marked up explicitly, there's less room for a model to guess wrong about you. We've watched AI answers get a business's service area flat wrong because the only source was a sentence buried in an About page. Schema fixes that class of problem.

Give crawlers a map. A newer convention worth adopting is an llms.txt file at the root of your domain, the way robots.txt has lived there for decades. It's a plain-text file that points AI crawlers at the pages you most want them to read and understand, with short descriptions. It's not magic and not every model honors it yet, but it costs almost nothing to add and it makes your intent explicit. We put one on this site. Pairing it with clean schema is most of the technical work.

None of that requires reinventing your website. It's a layer of clarity on top of content that should already be good.

The honest part about what you can't control

Now the caveat, because this is where the hype gets dangerous.

You do not control what a model says. You optimize the inputs. Anyone promising you a guaranteed spot in ChatGPT's answer, or a fixed citation in Google's AI overview, is selling you something they can't deliver. These systems are probabilistic. The same question can produce different answers on different days, the models retrain on schedules nobody outside the labs controls, and there's no ad slot to buy your way into the organic answer.

What you can do is stack the odds. Be the clearest, most consistent, most credible source on the questions you want to own, and you become a likely candidate for citation. Likely, not guaranteed. We're upfront with clients about that distinction because the alternative is selling a lottery ticket and calling it a strategy.

There's also no real-time dashboard for this yet, the way there is for search rank. You can spot-check by asking the models your buyers' questions and seeing whether you come up, and you can watch for referral traffic and the "I found you through ChatGPT" comments. But if someone shows you a tool claiming a precise "AI visibility score," treat the number with suspicion. The measurement is genuinely immature. We'd rather tell you that than pretend otherwise.

So what should a small business actually do

Start with the boring foundation, because it pays off in both search and AI answers: fast pages, clean structure, accurate and consistent business info everywhere you're listed. Then write honest, specific answers to the real questions your customers ask, the full-sentence ones. Add schema so the facts aren't left to guesswork. Drop in an llms.txt. Then keep publishing content that's actually worth quoting.

That's most of it. It's unglamorous, and it's the work that holds up whether the next big channel is an AI assistant or something nobody's named yet. If you want help with the citation side specifically, that's exactly what our get cited by AI search work is built around.

Common questions

Is GEO going to replace SEO? No, and we'd be skeptical of anyone who says it will. Plenty of people still type a query and scan links, especially for transactional searches where they want options. GEO is a second front you fight on, not a replacement. The good news is the foundational work serves both, so you're rarely choosing between them.

Can you guarantee my business shows up in ChatGPT? No, and neither can anyone else. We can make your site the kind of clear, credible, consistent source these models tend to cite, which meaningfully improves your chances. But the output belongs to the model, not to us or you. Guaranteed placement isn't a real product.

How long before this shows results? Slower than a paid ad, comparable to SEO. Models retrain and re-crawl on their own timelines, so changes you make today might surface in answers weeks or a couple of months out. It's a compounding investment, not a switch you flip.

If you're not sure whether your site is even legible to these systems right now, that's a fair place to start, and it's a quick thing for us to check. Tell us the questions your best customers ask before they buy, and we'll look at whether your site answers them in a way both Google and the AI assistants can actually use. No pressure and no jargon. Get in touch and we'll take a look.