GEO Strategy

If you are reading this, I will assume you have already decided AI search is worth investing in. If you have not, that is an earlier question, and I worked through it in Is your brand ready for AI search. This is for the stage after the decision, when the real question becomes: how much of this channel can you actually influence?

It is the question committed leaders ask me most. Will their effort move anything, or are they funding a black box? The answer decides the return on what they have signed off, so get clear on what you control, what you influence indirectly, and what you do not control at all.

The reassuring part, backed by evidence rather than hype, is that most of this channel is more influenceable than it looks. LLMs cite what they can find, what they can lift, and what they have reason to trust. Two of those are firmly in your hands. The third you earn over time. The mechanism shows you where the leverage sits, so start there.

How an LLM builds an answer, and why that helps you

Picture ChatGPT or Gemini not as an expert answering from memory, but as a very fast researcher. Ask a question that needs current or specific information and the assistant does not reach into its training for the answer. It runs a search, pulls back pages, reads the relevant parts, and writes a summary with sources attached. The industry calls this retrieval-augmented generation. The term does not matter. The behaviour does: it looks things up, then writes.

That is the root of your leverage. Because it quotes what it looks up rather than what it once learned, it cites your live content, the thing you can change. Improve the page and you improve what it can quote tomorrow.

One detail shapes how you measure. Each assistant reads a different library: ChatGPT’s search leans on Bing, Gemini and Google’s AI Overviews on Google’s index, Perplexity on its own, Claude on the open web. There is no single AI to optimise for, which is why the same question yields different brands in different tools.

Here are the three things that have to be true, ordered by how much you control each one.

The first lever: being findable, and you already own it

This one is overlooked, and it is pure SEO. You have been pulling it for years.

If the model searches and reads the results, being absent makes you uncitable, whatever your brand is worth. Findability is the cost of entry, and it is in your control. It is also where the “search is finished” crowd goes wrong, because the two are tightly linked. One analysis of 432,000 keywords found 97% of AI Overviews cite at least one source from the top twenty organic results. A separate study tracked that overlap over sixteen months and watched it climb from a third to over half, higher still in health, insurance and education.

The link is not perfect. Ahrefs found the share of AI Overview citations from top-ten pages fell to 38% from 76% in under a year, and the chat assistants track rankings even less closely. So ranking well raises your odds, the foundations are shared, and you control them, but ranking alone is no longer enough. That gap is the space GEO occupies. None of this makes your SEO investment redundant. It is the platform the rest of your effort stands on.

The second lever: being easy to quote, the one most teams miss

Findability gets you considered. This lever decides whether you are used, and it is almost entirely in your hands, because it comes down to how you write.

The model skims for a clear, self-contained passage it can lift and attribute. These systems quote passages, not whole pages, so a page that ranks well but buries its answer loses to a clearer one. Princeton and Georgia Tech ran the first controlled experiment on this: adding clear citations, direct quotations and specific statistics raised how often content was used in AI answers by up to 40%, and mid-ranking pages gained the most.

There is no trick to it. It rewards clarity, specificity and evidence, the same things that make content good for a reader. This is your highest-leverage, lowest-cost move, because it is how you write what you were publishing anyway.

The third lever: being trusted, which you earn rather than pull

The first two levers you control directly. This one you only influence, and the distinction is where budget gets wasted.

Models lean toward credible sources and brands discussed across the wider web, not just on their own site. Reddit is the most cited source across the major assistants, with YouTube, LinkedIn, Wikipedia and the big editorial names close behind, and G2 showing up for business software. Your own site is rarely the whole answer. You are assembled from what the rest of the web says about you.

We’ve seen some interesting insights in the tracking data we’ve collected with reconnAI that add nuance to this. Citations spread thin: even the top domain on any platform rarely exceeds 5% of citations, so breadth beats owning one ranking. And brands that get talked about get cited more, with one study linking a heavy Reddit presence to several times more ChatGPT citations.

Brands that are widely and credibly discussed have stronger signals of what they are known for, and those signals lift ranking and citation alike. You cannot make a model cite a thread or buy your way in. But you can influence the cause, by being worth talking about, through earned coverage, real reviews, community presence and video. This is digital PR and brand-building, which serious SEO always included. It is slower and indirect, but real, and it compounds.

What you do not control, and shouldn’t chase

Knowing what you cannot move is worth as much as knowing what you can. You do not choose which library each assistant reads or how it weights sources, and those change without notice. The engines diverge: source overlap between two major assistants can be as low as one in nine, so ChatGPT is no guide to Perplexity, and neither tells you what Gemini shows in another market.

Do not chase these. They are conditions to manage, not knobs to turn. Manage them with breadth, and by watching results per engine rather than assuming one reading speaks for all.

Where to put the effort you have committed

Everything above tells you which levers to pull. It does not tell you where to start — and that depends entirely on your industry. Who is already winning the answers? Which sources do the models lean on in your category? Where are you already showing up, and where are you invisible? Without that picture, you are guessing at priorities and hoping the budget lands somewhere useful.

That is the gap we built reconnAI to close. It reads the landscape of your industry across ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews, and shows you:

  • Who’s winning — your share of voice against competitors, and how it shifts by engine and market, so you can see who owns the answers you want.
  • Which sources are being cited — the domains, threads and publications the models actually draw on in your category, so you know where credibility is being built.
  • Where to engage — the specific places worth your effort, rather than a generic “get on Reddit” guess.
  • Where you’re visible, and where you’re not — how your brand is described when it does appear, and the questions where you are absent entirely.

In short, it turns a black box into a map. You stop guessing where to start and begin with what the evidence in your own market is telling you. From there, the build follows the guide to building a GEO strategy from the ground up.

The bottom line

This channel is more influenceable than its mystique suggests. You control whether you are findable and whether you are easy to quote, where the fast returns sit. You earn trust over time, and you measure the rest rather than chasing it. The brands that get a return are not hunting for a secret switch. They are spending on the levers they hold, building on solid SEO foundations, and watching closely enough to know what works. If you have committed the budget, that is where it should go.

Matt Johnson
Co-founder, reconnAI