Every few months, someone declares SEO dead. This time the killer is supposed to be AI search, and the industry has reacted the way it always does: by inventing new acronyms. AEO. GEO. LLMO. AISEO. Half of the definitions contradict each other, and most of them come from people who sell software for the acronym they just invented.
I want to offer a different starting point, because I am in a good position to compare the claims against reality. I run SEO and GEO for one of Finland’s largest consumer websites. And from where I sit, two things are true at the same time.
First: the news of SEO’s death is greatly exaggerated. In my daily work, organic search still brings massive amounts of traffic and revenue compared to other channels. Not “still hanging on” massive, but “clearly one of the most valuable channels” massive. Anyone telling you to stop investing in SEO in 2026 is asking you to walk away from an asset that still performs.
Second: my own search behaviour has already changed. I notice it all the time. I ask an AI assistant something, I read the answer, and I do not click through to any website, unless I have a real reason to visit the site itself: something to buy, something to log in to, something to do. The click is no longer the default. It has become intentional. And I am not alone. Pew Research found in 2025 that when Google shows an AI summary, users click a traditional result in about 8% of visits, compared to 15% when there is no summary.
These two facts sound like a contradiction. They are not. They describe different layers of the same system. Understanding those layers is exactly what SEO, AEO and GEO separate from each other. So let’s define the terms properly, look at what the evidence supports, and then look at how much of the “new” work is simply good SEO with a new name.
The 30-second answer
One honest note first: the industry uses these terms in conflicting ways, which is part of the confusion this article is about. So these are my working definitions. In this article, AEO means extractive, direct-answer systems, and GEO means generative, synthesized answers.
SEO (Search Engine Optimization) means optimizing to rank in a list of search results. The unit of competition is the page. The reward is the click.
AEO (Answer Engine Optimization) means getting your text chosen as the direct, single answer to a query: featured snippets, voice search, zero-click answer boxes. The unit of competition is the passage. The reward is being picked as the answer.
GEO (Generative Engine Optimization) means making sure your brand or site gets synthesized, mentioned and cited as a recommended resource inside broader conversational AI answers: ChatGPT, Perplexity, Gemini, Claude, AI Overviews. The unit of competition expands beyond the page: to your brand, its content, and the sources that describe it across the web. The reward is presence in the answer, often without any click at all.
The most important thing to understand: these are layers, not replacements. Generative engines pull heavily from the web’s search indexes. In early 2025, Seer Interactive found that 87% of ChatGPT Search citations matched Bing’s top organic results. Since then the mix has shifted: independent tests in 2025 and 2026 indicate that ChatGPT also taps Google’s index for many users, and OpenAI runs its own search crawler, so no single engine’s rankings predict citations as neatly anymore. The principle, however, has only grown stronger: your content has to live in some search index to be retrieved at all. If a page cannot be crawled and indexed, no GEO tactic will save it. SEO is the foundation. AEO made content easy to extract. GEO extends the competition from your own website to the entire conversation about your brand.
Side by side
| Dimension | SEO | AEO | GEO |
|---|---|---|---|
| Optimizes for | Ranking in results | Being chosen as the answer | Being cited in AI-generated answers |
| Unit of competition | Page | Passage | Brand as a whole, plus its sources |
| Where it happens | Search result pages | Snippets, voice, answer boxes | ChatGPT, Perplexity, Gemini, Claude, AI Overviews |
| How much you control | Your site | Your content structure | Only partly: in discovery queries, most cited sources are third-party |
| Key metrics | Rankings, clicks, organic revenue | Answer ownership | Share of voice, citation rate, accuracy |
| How stable it is | Fairly stable | Fairly stable | Varies: the same question can give a different answer tomorrow |
| Age of the evidence | 25 years | About 8 years | 2 to 3 years, and still thin |
Good GEO is mostly good SEO, and the evidence supports this
Here is the claim I most want to push back on: the idea that GEO and SEO are somehow in conflict, that you have to choose one, or that optimizing for AI means doing something completely different from what good SEO has always been.
I have been looking for real conflicts between the two. I keep not finding them.
Start with the mechanics. AI assistants with web access do not have their own private map of the internet. They query search indexes, retrieve pages, and build their answers from what they find. This means the classic SEO basics are not made useless by AI search. They are its requirement. A page that cannot be crawled cannot be retrieved. A page that cannot be retrieved cannot be cited. (This is also why one of the most common GEO failures I see is surprisingly basic: sites whose bot protection silently blocks AI crawlers, or whose rendering setup leaves important content out of the initial HTML. I have written about why a site can look complete in a browser while giving AI readers almost nothing to work with. You cannot win a game you have locked yourself out of.)
The data points in the same direction. In early 2025, a study by Seer Interactive found a strong connection (a correlation of about 0.65) between ranking on page one of Google and being mentioned in AI answers, and later analyses have kept finding the same direction: pages that rank well are far more likely to be cited. The connection is not perfect, and every engine has its own taste: Profound’s analysis of AI citations shows that ChatGPT, Google AI Overviews and Perplexity each favour quite different sources. But “the same URLs do not always get cited” is very different from “SEO does not matter”. The foundations still transfer. The finish line moved.
And the tactics? Look at what the research on AI visibility actually recommends: clear answer-first structure, quotable statements, cited statistics, expert quotes, structured data, passages that work on their own. Now compare that to what AEO best practice has recommended since featured snippets appeared, and what quality-focused SEO has recommended for over a decade. It is the same list. (I tested one part of this myself: whether FAQ content actually matters for AI search. The answer was, fittingly, yes and no at the same time.)
I have not yet found a credible, evidence-backed GEO tactic that inherently conflicts with good SEO. You can of course execute badly: over-structure a page, write it so generic that it converts nobody. But then the problem is the execution, not the discipline. So when someone offers to “do GEO” for you as something separate from SEO, ask them what exactly they would do differently, and what evidence supports it. The honest answer is that the work on your own site overlaps almost completely. The real difference is somewhere else: outside your site. That brings me to the part that actually is new.
The genuinely new part: top of mind is built on sites you don’t own
This is the shift I find most important. When people ask AI engines discovery and comparison questions (“what is the best option for X”), study after study finds that most of the cited sources are third-party: review sites, forums, media coverage, YouTube, Wikipedia, LinkedIn. Analyses of large citation datasets show that Wikipedia is ChatGPT’s single most cited source, and on Perplexity, Reddit alone makes up almost half of the top cited sources. A June 2026 synthesis of more than 680 million tracked citations confirms the same pattern. Brand-owned pages barely register on these lists.
To be fair, the numbers depend on the type of question. A Yext study of 6.8 million citations found that for local and branded queries, most citations come from sources brands control, such as their own sites and listings. But in the discovery moment, the moment when a buyer does not yet know your name, the answer is built largely from what other people have written about you.
I see this play out at national scale in very concrete ways. A story that Finland’s largest news outlet wrote about one of our products has become a more cited source in AI tools than our own product pages. Read that again: the earned article beats the owned content on the very surface where buying decisions increasingly begin. If your SEO team ever needed a business case for working closely with the PR team, this is it.
And it does not stop at traditional media. My own monitoring is small and directional, but in it I see social content appear as a cited source in AI answers more and more often: individual LinkedIn posts and articles, even Instagram profiles. The engines do not draw a line between “official” and “unofficial” sources. They cite what they can find and what they have learned to trust.
For marketing teams, this breaks an old wall in the org chart. SEO and PR have been separate functions with separate budgets and separate KPIs for as long as I have worked in this field. AI engines do not care about our org charts. Visibility is no longer only about optimizing your own content. It is about how your brand is talked about everywhere else. If I had to name the single biggest strategic change GEO brings, it is this one.
What the evidence does and does not show
Because this field is full of vendor marketing, let me be precise about the state of the research, including the inconvenient parts.
The paper everyone cites, the Princeton study that launched the term GEO and its famous “up to 40% visibility improvement”, is real but conditional. The gains were measured in a setup where the source was already available to the model. The study shows that presentation affects citation. It does not show that these tactics get you discovered in the first place, and it says nothing about lasting traffic effects. A 2026 academic review of 45 GEO studies came to a sobering conclusion: the terminology is inconsistent, the metrics are not standardized, and AI visibility is not a ranking you can climb step by step. It is a pipeline with many stages, and you can only see parts of it.
One of the few analyses based on real log data found that optimized pages got clearly more AI referral traffic than untouched pages: 6.1x growth versus 3.5x on the same site. The untouched pages grew too, simply because ChatGPT itself is growing so fast. Once you account for that, the effect of the optimization work itself was closer to 1.8x, and the authors themselves call the result suggestive rather than proven. Promising, not proven.
So, in summary: the foundations (crawling, indexing, authority) are well established. That is the SEO layer. The extraction tactics (structure, quotability, schema) have reasonable support. That is the AEO layer. The off-site brand footprint clearly matters. The citation data makes that hard to argue against. What has no evidence behind it: magic llms.txt files, secret “AI schema”, and anyone who guarantees you a spot in ChatGPT’s answers. (Full disclosure on llms.txt: yes, my own site has one. No, Google says no AI system currently uses it. And yes, Anthropic, the company behind Claude, has one on their own site anyway. Make of that what you will. It takes five minutes and does no harm, which is more than I can say for most tactics in this category.) The answers vary from day to day. If someone promises you a guaranteed position, that promise tells you everything you need to know.
What I would actually do differently in 2026
If good GEO is mostly good SEO, what changes in practice? Honestly assessed, four things:
- Measure a new thing. Add AI share-of-voice tracking to your reporting: a fixed set of buyer questions, run regularly across the main AI engines, with a log of whether and how your brand appears. You cannot manage something you have never once looked at.
- Decide, on purpose, which AI crawlers you allow. Some companies have good reasons to block certain bots: content rights, or the difference between training crawlers and search crawlers. The failure mode is not blocking itself. It is blocking or allowing by accident, through a robots.txt line or a CDN default you did not know existed. Check GPTBot, PerplexityBot, ClaudeBot and friends, and make it a decision instead of an accident.
- Bring earned media into your search strategy. Treat PR, reviews, community presence and expert commentary as visibility infrastructure, not as a separate discipline. Budget accordingly.
- Keep doing excellent SEO and answer-first content. Not as a legacy activity, but as the foundation everything above sits on. The channel that still drives the revenue deserves better than obituaries.
Common misconceptions
“GEO replaces SEO.” It sits on top of it. No index presence, no retrieval, no citation.
“GEO means optimizing your own content for AI.” Mostly it means earning presence on sites you do not own. In discovery questions, your own content is a small share of what gets cited.
“AEO and GEO are the same thing.” Different surfaces and different units of competition. AEO competes for a passage. GEO competes for a brand.
“You can guarantee AI visibility.” The answers change from day to day, and the research is two years old and full of open questions. Anyone selling guarantees is selling snake oil.