SEO, AEO and GEO: the difference and where to start

When someone asks Google, ChatGPT or Perplexity a question, three things can happen to your brand: it shows up as a link, it is chosen as the answer, or it is cited as a source. Each of these has its own name and its own techniques. But they do not compete. They are layers of the same job.

The three in one table

	SEO	AEO	GEO
Stands for	Search Engine Optimization	Answer Engine Optimization	Generative Engine Optimization
Goal	Rank well in search results	Be the direct answer	Be cited in AI answers
Where it shows	Google and Bing results pages	Featured snippets, answer boxes, voice assistants	ChatGPT, Gemini, Perplexity, Google's AI Overviews and AI Mode
What to optimize	Crawling, speed, search intent, authority	Short, structured answers, questions as headings	Verifiable facts, a clear brand and author, mentions on trusted sources

SEO: being found

SEO is still the foundation. If a search engine cannot crawl, index and understand your page, nothing that comes after it works. In practice there are four fronts:

Technical: pages that load fast (Core Web Vitals), work well on mobile and do not block crawling by accident.
Intent: each page answers one question or one need, not ten at once.
Content: clear writing, with depth and the experience of someone who does the work.
Authority: other relevant sites that mention and link to yours.

AEO: being the answer

AEO starts from a simple observation: many people want the answer, not a list of links. Search engines and assistants build that answer from pieces of pages. To be the piece they pick:

Write the question as a heading, the way people search for it.
Answer in the first two lines, with no detour, and go deeper afterwards.
Use lists, numbered steps and tables when the content is a list, a how-to or a comparison.
Mark the page up with structured data (schema) so machines understand each part. It helps with reading the page, but it does not guarantee a visual highlight in search.

GEO: being cited

Generative AI answers in running text and, in many cases, shows the sources it relied on. GEO works so that your brand is among them. Nobody controls what an AI answers, so the goal is to improve the odds:

Be factual and verifiable. Data, dates, definitions and concrete examples are easier to cite than vague opinion.
Make it clear who is speaking. Author, company and field of work on stable pages, with the same details on public profiles such as LinkedIn.
Get mentioned outside your site. Interviews, third-party articles, communities and relevant directories shape how AI understands who you are.
Allow access. Check your robots.txt: if it blocks AI crawlers such as GPTBot, ClaudeBot and PerplexityBot, your content stays out of those answers.
Optional: an llms.txt file summarizes the site for language models. It is a proposal that costs little to publish, but there is no guarantee that platforms use it.

Where to start

Order matters, because each step prepares the next:

Check the technical base. Add the site to Google Search Console, see what is indexed and measure Core Web Vitals. Fix whatever blocks crawling before producing new content.
List the real questions. Collect 10 to 20 questions that customers and prospects ask in meetings, in support and on WhatsApp. They are worth more than a keyword list.
Write one page per question. A direct answer in the first two lines, depth afterwards, real examples and, at the end, the next step for people who want to talk.
Structure and sign. Apply structured data, keep the author and company visible, and keep the details consistent between your site and your profiles.
Open the crawlers and measure. Keep robots.txt open to the search engines and to the AI crawlers you want citing your site, and track the results.

How to measure each one

SEO: impressions, clicks and average position in Search Console; organic traffic and conversions in GA4.
AEO: whether your pages appear in highlights and answer boxes, tracked in Search Console and with manual checks of your main questions.
GEO: a periodic test, asking AI tools about your topics and noting whether the brand is cited, plus referral traffic from AI domains in GA4.

None of these measurements is perfect, and GEO is still the hardest. That is why it is worth recording the questions you tested and the date, so you can compare month over month.

Common mistakes

Treating SEO, AEO and GEO as rival disciplines, with separate teams and budgets.
Writing for the robot and forgetting the person who will read it.
Publishing generic AI text with no experience or data of your own. It is what AI cites least.
Skipping the technical base and jumping straight to "optimizing for AI".
Not measuring, and deciding by gut feeling.

Quick questions

Does SEO still matter now that AI exists?

It does, a lot. AI tools that search the web depend on pages that can be crawled and understood, and many of the sources they cite come from well-ranked sites. SEO is the base of the other two layers.

Do I need to create an llms.txt?

It is not required and there is no guarantee that platforms use it. Since it is simple and cheap to publish, it may be worth it, as long as the rest (content, crawling, brand) is in order.

How long until results show?

It depends on where the site starts. Technical fixes can show effects in weeks. Authority and citations in AI take months and depend on publishing consistently.

If you want to apply this to your site or your company, talk to me.