Generative Engine Optimization

GEO, explained.

Generative Engine Optimization is the practice of making your business findable, quotable and citable by the machines people now ask instead of searching. ChatGPT. Perplexity. Google's AI Overviews. Claude. The engines that don't hand you ten links, they hand you an answer.

You will also see it called AIO, AI optimization, or answer engine optimization. Same thing. The acronym is not settled because the discipline is about three years old. The problem it addresses is very real, and it is not waiting for the naming to resolve.

The short version.

SEO was about ranking on a page. GEO is about existing in the layer those pages now get assembled from.

The engines don't share an index. They do share a taste in sources. GEO is about being the kind of source they reach for.

For twenty-five years the game was to appear in a list and win the click. Now the machine reads the list on the user's behalf, assembles an answer, and names a handful of sources. If you're not one of them, you weren't in the conversation, whatever position you held.

This isn't algorithm gaming. It's making sure the machines have your facts straight.

Why this is happening now.

Two findings are worth building a strategy around, and neither is mine.

Steal this diagram, credit is appreciated.Download SVG

Semrush found in 2025 that a visitor arriving from AI search was worth roughly 4.4 times a traditional organic visitor in conversion value. In the same year Ahrefs reported that 12.1% of its signups came from AI search on just 0.5% of its traffic. That's a 23x premium on a rounding error of a traffic source.

Ahrefs also measured a 34.5% drop in click-through rate for the top organic result when an AI Overview sits above it. Put those together and the picture isn't subtle. Fewer clicks are coming. The ones that come are worth considerably more.

This is the part most businesses haven't internalized. Traffic is going down and value per visit is going up at the same time. If you're still measuring your marketing purely on sessions, the number will fall and you'll conclude something is broken, when in fact the metric quietly stopped describing reality.

How answer engines actually pick their sources.

Nobody outside these companies has the exact recipe, and anyone selling you one is guessing with confidence. But the behavior is observable, and it is consistent enough to work with.

They retrieve, then they synthesize. The model doesn't remember your website. It searches, pulls passages, and writes an answer from what it pulled. Which means the unit of optimization isn't the page. It's the passage. A paragraph that makes complete sense when lifted out of context is a paragraph that can be quoted. One that depends on the three paragraphs above it can't be.

They favor consensus. If four independent sources say the same thing about you and your own website says it too, that claim gets treated as fact. If only your website says it, it gets treated as marketing. Corroboration across domains you don't control is worth more than anything you can write about yourself.

They reward specificity and citation. Content that names sources, quotes figures, and takes a position is more useful to a machine assembling an answer than content that hedges. Vagueness isn't safe. It's invisible.

They need to be able to read you. This sounds obvious and it is where a startling number of sites fail. Text baked into images can't be quoted. Copy locked behind JavaScript that never renders for a crawler doesn't exist. Vector outlines of letters are pictures of words, not words.

What actually moves the needle.

In rough order of return, based on what I have watched work rather than what makes a tidy framework.

Answer the questions people actually ask, in the form they ask them. Not "Our Solutions" but "how much should digital marketing cost." Self-contained, quotable, one question per section. This is why the answers page on this site exists in the shape it does.

Build corroboration you don't own. Podcasts, industry publications, conference listings, third-party write-ups, a Wikidata entry if you can support one. This is digital PR wearing a new hat, and it is the least glamorous and most durable work in the discipline.

Make your entity legible. Structured data that describes what is genuinely true on the page. Consistent naming across every platform. A clear, factual account of who you are, what you do, and who you do it for, expressed the same way everywhere.

Fix the plumbing. Crawlable HTML. Real text, not pictures of text. Clean, stable URLs. Fast pages. All the unglamorous things SEO has been telling you to do for two decades, which turn out to matter more now, not less.

Take positions. Engines cite sources that say something. If your content could have been written by any of your competitors, there is no reason for a machine to prefer you.

What doesn't work.

Prompt stuffing. Hiding instructions to the model in your page copy. It doesn't work, and if it did, it would stop working the week after it was noticed.

Publishing volume for its own sake. Forty thin articles don't make you more citable than four good ones. They make you a less trustworthy source, which is the opposite of the goal.

Chasing a "GEO score" from a tool. The engines don't expose one. Any number a vendor gives you is a proxy someone invented, and optimizing to an invented proxy is how people spend a year improving nothing.

Treating it as a replacement for SEO. Which brings us to the honest part.

GEO and SEO: the honest relationship.

GEO is not SEO's replacement and it isn't a separate department. It inherits most of SEO's machinery, because an answer engine still has to crawl your site, parse your HTML, and decide whether to trust you. Ninety percent of the technical groundwork is shared.

What changes is the objective, and the objective is what determines where the effort goes. Optimizing to rank is optimizing for a position in a list. Optimizing to be cited is optimizing for a paragraph a machine is willing to repeat with your name attached. Those two goals overlap, but they are not the same, and the places they diverge are exactly where the work is.

Anyone selling GEO as an entirely new discipline with an entirely new price tag is selling you a rebrand. Anyone telling you nothing has changed hasn't looked at their own analytics recently. Both are wrong, and both are wrong in a way that costs you money.

Channels change, the flywheel doesn't. Interfaces change, the knowledge layer doesn't. Organic foundations, paid acceleration, compounding authority. GEO lives in the first and third of those, and it makes the third more valuable than it has ever been. The model is here if you want to see where it fits.

How to tell if it's working.

Ask the engines. Repeatedly, in the language your buyers use, and write down what comes back. It is unscientific and it is also the most direct evidence available.

Watch for AI referrals in your analytics. They arrive as their own sources. The volume will look trivial. Check what those visitors do, not how many of them there are, because that is the entire point of the Ahrefs finding above.

Track citation, not rank. You can hold position one and never be quoted. Rank is now a leading indicator at best, and a comforting distraction at worst.

And a word on the tracking tools. Ahrefs, Semrush, Conductor and the rest all sell AI citation trackers now. This is opinion rather than settled fact, but I think they're structurally broken.

Answer engines personalize hard. What comes back depends on who's asking, what they asked before, and where they're sitting. A tracker queries from one place, as one profile, and writes the result to a database. That's rank-tracker logic pointed at a system that doesn't have ranks. You get a number. It's precise, it's repeatable, and it's measuring something that isn't happening to your actual customers.

I cite those companies' research elsewhere on this site and I rate it. The trackers are a different thing. The best intel available to you is already sitting in your own analytics, and it costs nothing. If you want to know where to look, ask me.

Common questions.

Is GEO just SEO with a new name?

No, but it is not a separate discipline either. It inherits most of SEO's machinery, since answer engines still need to crawl, parse and trust a page. What changes is the objective. SEO competes for a position in a list. GEO competes to be the sentence the machine repeats.

Do I need to do GEO if my SEO is already good?

Good SEO gives you a real head start, because crawlable, well structured, trustworthy pages are the raw material answer engines feed on. It does not finish the job. Ranking first for a query nobody clicks anymore is not the same as being the source an answer engine quotes.

How do I know if GEO is working?

Ask the engines directly and repeatedly, watch for referral traffic from AI assistants in your analytics, and track whether your brand appears in answers to the questions your buyers actually ask. Rankings are a poor proxy, because you can rank well and still never be cited.

Can you pay to appear in AI answers?

Not in the way you can buy an ad slot. Answer engines assemble responses from sources they judge relevant and credible. Advertising is arriving in some products, but it sits alongside the answer rather than replacing how the answer is built.

Want to know if any of this applies to you?

The honest answer for a lot of businesses is: partly, and not yet urgently. For some it's already costing them. Working out which one you are takes one conversation, and I will tell you plainly if the answer is that you have bigger problems.

Start a Conversation

I keep the ongoing roster deliberately small. Strategy doesn't scale like software, and I'd rather be great for a few than adequate for many.

Back to the glossary