SEO, AEO and GEO: what the difference is, and why ignoring one of the three costs traffic

What you get from this article

• SEO competes for a position in a list, AEO competes for the answer block, and GEO competes for a citation inside generated text.
• The technical base is the same for all three: a fast, crawlable site with structured data and content that genuinely answers.
• Being cited inside the answer block yields 120% more clicks per impression than being left out of it, according to Seer Interactive.
• Content that earns AI citations has sources, dated numbers and a self-contained answer in the first paragraph.
• Measuring GEO needs its own method: Search Console does not show mentions inside ChatGPT.
• Doing only SEO in 2026 is optimizing for the half of search that still produces a click.

Search stopped being a list of links

For twenty years the job was easy to explain: somebody typed a question, Google returned ten blue links, and whoever sat closest to the top took the click. All the SEO that exists was built on that mechanic. Keywords, backlinks, page speed, all of it served to climb a list.

That list still exists, but it shrank. Today the same question can return an answer block written by Google itself above everything, a side panel, a video, a related-questions section, and only then the links. And a good share of questions never reaches Google at all: they are typed straight into ChatGPT, Gemini or Perplexity, which answer in prose and cite a few sources in a footer.

The effect on traffic is measured, not guessed. Ahrefs analysed 300,000 search terms comparing March 2024 with March 2025 and found a 34.5% drop in the click-through rate of the first position when an AI-generated answer block is on the page. Seer Interactive, tracking 53 brands, 5.47 million terms and 2.43 billion impressions, saw the organic click-through rate on searches with that block fall from 3.19% in January 2025 to 1.31% in December of the same year.

Except the story did not stop there, and here comes the figure almost nobody repeats. In the same study, Seer measured a recovery: from 1.31% in December 2025 to 2.36% in February 2026, an 80% rise in two months. The fall did not become a straight line to zero. What search did was change the rules mid-game, knock down everyone optimized for the old rule, and then settle at a new level, lower than the old one and higher than the bottom.

And there is a second number in the same study worth more than the first two combined: being cited inside the answer block delivers 120% more clicks per impression than not being cited. That is not consolation. It is the instruction. If the block is going to appear anyway, the fight stopped being against it and became a fight to be inside it.

That is why new acronyms appeared. Not because somebody needed to sell a service under a different name, but because there are now three distinct destinations for the same content, and each one has its own selection rule.

SEO: the foundation, which did not become the past

SEO is Search Engine Optimization. The goal is to get a specific page to appear as high as possible in the results list for a specific term. The unit of measure is position, and the outcome that matters is the click.

The work has three fronts that anyone who has done it seriously knows: the technical side (the crawler can enter, understand and index the page, and it loads fast on a phone), the content (the page genuinely answers the intent behind the search, rather than repeating the keyword) and the authority (other relevant sites point at yours, which works as a vote of confidence).

Since 2024 the idea that SEO is dead has been going around. It is not, and the reason is technical: the AI systems that answer in prose have to find the material somewhere, and the place they pull from is mostly the search index. A site the crawler cannot reach does not enter the list or the generated answer. The foundation is still under all three floors.

What changed is that SEO stopped being enough. It still delivers the click from the half of searches that still produce clicks, and that half is worth a lot. It is just no longer the whole map. If you are setting this work up now, see how an SEO agency structures that base before anything else.

AEO: optimizing for the answer, even with no click

AEO is Answer Engine Optimization. Here the target is not the position in the list: it is the block of text the search engine assembles at the top of the page, the featured snippet, the answer card, what the voice assistant reads aloud when somebody asks their phone.

The practical difference is large. To win a position, a page has to be the best on the whole subject. To win the answer block, it needs, somewhere in the text, a self-contained passage that answers exactly that question without depending on the previous paragraph. The system cuts. If your best argument only makes sense after three paragraphs of context, it is not cuttable.

That changes how the text is written. A question as a section heading, the answer in two or three sentences right below it, and only then the depth. Short, direct definitions. Tables when there is a comparison. Numbered steps when there is a process. None of that is a trick: it is formatting that makes the cut easier, and it happens to be easier for people to read too.

One honest warning about the outcome. Winning the answer block often means the person got what they wanted and did not click. And even when cited inside it, the click does not return to the old level: in the Seer study, results cited in the block still perform 38% below searches where the block simply does not appear. In other words, the best position inside the new format is still worth fewer visits than a search without it.

That turns the decision into business strategy, not a technical tweak. Which content do you give away in the answer, knowing you will lose part of the click but gain brand recall? And which content do you keep behind the visit, because it only makes sense whole? That is the question no tool answers for you.

The answer block did not take your click. It changed the price of the click, and now part of the payment is content given away for free.

GEO: how an AI decides who to cite

GEO is Generative Engine Optimization. The target is being mentioned by name inside the text that ChatGPT, Gemini, Claude or Perplexity write when somebody asks something in your field. It is not appearing in a list or a block: it is being the reference the model cites.

The mechanics differ from the other two and are worth understanding. When an assistant answers a current question, it usually runs searches of its own, reads a few pages and writes new text based on what it read. It does not copy your page: it extracts claims from it. So what has to be impeccable is not keyword density, it is the claim. Clear, verifiable sentences, with a number and a date, attributed to a source.

There is a second, slower path, which is mentions spread across the internet. Models learn on public text, and a brand cited in many different places tends to come up more often even without a live search. That cannot be bought with links: it is built with editorial presence, with content other people repeat, with real participation in the subject.

The detail most people ignore: the AI has to be able to attribute the claim to somebody. A text with no author, no date, no source, published by a company with no institutional page confirming who it is, is a text the model reads and cannot credit. Explicit authorship is not vanity, it is citation infrastructure. It is the same reasoning behind how the AI Benchmark publishes where every number comes from instead of showing only the ranking.

What the three have in common, and where they fight

The shared part is bigger than it looks, and that is good news: a fast, crawlable site with HTTPS, correct structured data, headings that genuinely describe the content and text that answers the intent of whoever arrived serves all three destinations at once. Nobody needs three sites.

The fight shows up in two places. The first is the click: SEO exists to bring the person in, AEO frequently solves their problem before bringing them in. The second is length: SEO rewards depth, and a 2,000-word page usually ranks better than a 300-word one; AEO rewards brevity, because an answer block is two sentences.

The way out is not to choose: it is to structure the page in layers. The short, cuttable answer at the top, the depth below, the source visible at the end. Whoever only wants the answer takes it and leaves, and the search engine has something to cut. Whoever wants to understand stays, and the system has enough depth to recognize the page as an authority on the subject.

The article you are reading is built exactly that way, on purpose. The direct answer opens the page, the conclusions come right after, the body develops them and the sources close it. It is not a layout coincidence.

• Shared: speed, crawlability, structured data, clear headings, search intent actually met.
• SEO only: authority through links, URL architecture, competing for long-tail terms.
• AEO only: cuttable passages, question-and-answer format, FAQ markup, lists and tables.
• GEO only: verifiable claims with source and date, identified authorship, presence in third-party texts.

Where to start, in order

The order matters more than the list. Doing GEO on a site the crawler cannot reach properly is spending money on the top floor of a building with no foundation.

First, the technical side. Check that pages are indexed, that the sitemap exists and is up to date, that the site opens fast on a phone over a poor connection, that every page has its own title and description. That is week one, and without it nothing else pays off.

Second, content architecture. List the questions your customers actually ask, in the words they use, not in the words your industry uses. One question per page. If two questions are answered together, that is one page with two sections.

Third, format. Every page opens with the answer in two or three sentences. Then comes the depth, in sections whose headings are also questions. At the end, a frequently asked questions section with structured data markup. Every number gets a source and a date.

Fourth, the signature. Who wrote it, what that person actually does, when the text was published and when it was reviewed. An institutional page confirming the company exists, with an address and a way to get in touch. That feeds all three destinations at once.

How to measure each one without fooling yourself

SEO has mature instrumentation. Search Console shows impressions, clicks, average position and the term that brought each visit. It is the most reliable data in this whole story, and it is free.

AEO is measurable indirectly. The most useful signal is a term that gains impressions and loses clicks: if a term appears far more often and brings proportionally fewer people, the answer is probably being delivered before the click. That is not automatic failure, but it has to go into the calculation.

GEO is where almost everyone fools themselves, because there is no official dashboard. What works today is the manual, tedious method: build a list of twenty to thirty questions a customer would ask, run that list through the main assistants once a month, and note in how many you were cited and in how many the competitor was. That is field research, not an automatic report. Anyone promising an exact number of AI mentions is estimating and calling it measurement.

And there is the signal worth more than all the others: people arriving and saying they heard about you in an AI answer. It is well worth asking that on first contact. It is the only genuinely reliable attribution data in this channel today.

The most expensive mistake is treating this as three projects

The common confusion is thinking these are three separate services, with three budgets, three teams and three reports. No company needs that, and anyone selling it that way is selling the same work three times.

It is one job with three destinations. The content is the same, the technical structure is the same, what changes is the delivery format and the way you measure. A well-built page competes for the list, the answer block and the AI citation without you doing anything different for each.

The symmetrical mistake is equally expensive: carrying on doing 2019 SEO, focused exclusively on position and keyword volume, and wondering why the position improved and the visits did not. That is happening to a lot of companies right now, and the report looks good while the phone goes quiet.

If you want to understand where your site stands across those three destinations today, that is exactly the diagnosis that opens an AI consulting engagement at ROO3. Before writing a single new line, the first step is looking at what already exists and finding out which of the three floors is missing.

Frequently asked questions

Did SEO die when AI arrived?

No. AI systems have to find the content somewhere, and the main place is still the search index. A site the crawler does not reach appears neither in the list nor in the generated answer. What changed is that SEO stopped being enough on its own, because a meaningful share of searches now ends without a click.

What is the practical difference between AEO and GEO?

AEO targets the answer block the search engine itself assembles on the results page, usually by cutting passages from existing pages. GEO targets being cited by name inside the text an assistant such as ChatGPT, Gemini or Perplexity writes. In the first case the search engine cuts you; in the second, the model rewrites and credits.

Do I need to hire three different services?

No. It is one job with three destinations. The technical base and the content are shared; what changes is the delivery format and the way you measure. Anyone selling SEO, AEO and GEO as three separate contracts is charging three times for the same effort.

How do I know whether AI systems are citing me?

There is no official dashboard. The method that works today is building a list of twenty to thirty questions a real customer would ask, running that list through the main assistants once a month, and noting in how many you appeared. It is manual research, and anyone promising an exact automatic number is estimating.

How long does it take to see results?

The technical part usually shows within weeks, because the search engine reindexes quickly after a fix. New content competing for position takes two to six months, depending on how contested the term is. AI citation is the slowest and least predictable, because it depends on the model finding and trusting the material.

Does structured data help you appear in AI answers?

It helps indirectly. Structured data markup makes explicit what is a question, what is an answer, who the author is and when it was published, and that makes both the cut by the search engine and the attribution by the model easier. It is no guarantee of citation, but it is one of the few signals you control entirely.

Sources

• Ahrefs: study on the drop in clicks with AI Overviews
• Seer Interactive: AIO Impact on Google CTR, 2026 update
• Google Search Central: structured data guide

Rodrigo Fávaro

Founder of ROO3, a marketing and technology agency in São José do Rio Preto, Brazil. Builds AI products running in production (Tobia, gerar.app, Pense Mercado) and maintains the AI Benchmark, a public ranking of AI models. See ROO3 AI consulting.

X @rodmf LinkedIn rodrigofavaro