Search is no longer just Google
For twenty years we optimized sites for a list of ten blue links. But when someone asks ChatGPT what is the best tool for X, they do not see ten links: they see a direct answer with two or three names. If your brand is not in that answer, you do not exist for that person.
AEO (Answer Engine Optimization) is the work of getting AI models to understand you, trust you, and cite you when they answer a question relevant to your product. You will also see it called GEO (Generative Engine Optimization): same idea, different name. This guide explains how an answer engine works inside, what exactly it looks at, a four-week plan for a small app and how to measure whether it works.
How an answer engine answers
Worth understanding, because it changes where effort pays off. When you ask an assistant with search enabled (ChatGPT with search, Perplexity, Google's AI summaries, Claude or Copilot), three things happen within seconds:
- It retrieves. It runs one or more searches against its index and gathers a handful of candidate pages. If your page is not indexed or cannot be read, it is not in the handful.
- It reads and selects. It pulls from those pages the fragments that answer the question. The paragraph that answers directly and stands on its own wins, not the longest page.
- It writes and cites. It composes the answer and adds links to the sources it used. That citation is the "visibility" AEO is after.
And when you ask without search, the model answers from memory: with what it learned in training. There, what has been said about you across the web for months matters, not what you changed yesterday.
Why it matters right now
AI assistants have become the first place many people research: what to buy, what stack to use, who to hire. These answers are built from what the model was able to read and understand on the web. If your site is a black box to them, your competitor gets the mention. And unlike Google, there is no second page here: either you get cited or you do not.
What an answer engine looks at
- Accessibility. That your content can be read without running JavaScript, because many bots do not render, and that
robots.txtdoes not block GPTBot, ClaudeBot, PerplexityBot or each provider's search bots. - Entity clarity. That it is clear who you are: the same brand name on every page, an
Organizationblock with links to your profiles, one sentence that defines you repeated verbatim on the home page, in the meta description and inllms.txt. - Citable content. Paragraphs that answer a question without depending on the rest of the page, with the answer in the first two sentences and concrete facts (figures, names, dates) instead of adjectives.
- Structure. One heading per question, one page per topic, real frequently asked questions. Models chunk by headings; each chunk should stand on its own.
- Freshness. Visible dates and updated content. For "best X in 2026" questions, recency weighs.
- External authority. Mentions on other sites, reviews, tool directories, comparisons. The model trusts what several sources repeat.
- Structured data. JSON-LD consistent with what is visible on the page. More in why AI cites some sites and not others.
A four-week plan for a small app
Week 1: the mechanics
Check that your key pages return real text in the HTML and not an empty div filled in by JavaScript; if your app is an SPA generated with Lovable, Bolt or v0, read why AI-built apps are invisible to AI. Review robots.txt, publish an llms.txt and add Organization and SoftwareApplication on the home and product pages. All of this is an afternoon's work.
Week 2: the content
Write the sentence that defines your product (what it is, for whom, what it achieves) and repeat it verbatim everywhere. Create one page per question your customer would ask an AI: "what is X", "X vs Y", "how to do Z with X". Put the direct answer at the top and the nuances after. Turn your real frequently asked questions into a section with FAQPage markup.
Week 3: entity and presence
Complete your public profiles with the same description (GitHub, LinkedIn, X, Product Hunt, tool directories in your category). Get the first external mentions: a comparison on a blog, a listing in a directory, a helpful answer in a community. The model learns who you are from what others say about you.
Week 4: measurement
Open ChatGPT, Claude and Perplexity and ask the same questions in all three: what your brand is, what it does, alternatives, and "recommend me a tool for [your problem]". Write down with the date whether you appear, whether what they say is correct and whom they confuse you with. Repeat monthly.
The mistakes that spoil it most often
- Blocking GPTBot or ClaudeBot with a copied
robots.txttemplate "to protect the content". - The product name and value proposition inside an image or a video.
- One long page that tells everything, without headings that say what each part is about.
- Changing the brand name or description between pages and profiles.
- Marking up
FAQPagewith questions that do not appear on the visible page.
How to measure AEO
Three sources. The monthly round of questions, which gives you presence and accuracy. Server logs filtered by the AI bots' user agents, which tell you whether they read you and which pages. And web analytics, where visits arriving from chatgpt.com or perplexity.ai show up as a traffic source: when someone clicks the citation, it shows.
AEO does not replace SEO
AEO leans on many classic SEO foundations: useful content, good structure, authority. The difference is the output format and the moment you are found. Think of it as a new layer on top of what you already do, not a replacement. We develop it in SEO vs AEO.
Scanaris checks your AEO for you
The AI visibility pillar of Scanaris reviews your llms.txt, your structured data, whether your content is readable without JavaScript and whether you block AI bots in robots.txt, and gives you a ready-to-paste fix prompt for each thing that is missing. The visibility tracker runs the round of questions for you: every day it checks whether you show up when someone asks for a recommendation in your category and how they describe you. These are read-only checks on public information; it never touches your site.
Frequently asked questions
Are AEO and GEO the same thing?
In practice, yes: both names describe optimizing to be cited in AI-generated answers. GEO comes from academia; AEO, from marketing. The techniques are the same.
How long until it shows?
Live visits and search, days to weeks: as soon as your pages are readable and indexed, they can appear as a source. The model's memory, until its next training run, which you do not control. That is why the external presence work is worth starting early.
Is it worth it for a small app?
Especially for a small app. On Google you compete with sites that have years of authority; in an AI answer, with the clarity of your page. A new product explained well can be cited before a big one explained badly.