How to Optimize for AEO (7 Steps, No Consultant)

To optimize for AEO, you first stop guessing and start watching which AI systems already read your site. You need to know how to optimize for AEO in a way that doesn't require a consultant or a big budget, and it comes down to measurement, structure, and patience.

Search is changing. SEO still matters, but AI answers increasingly decide who gets found. When someone asks ChatGPT or Perplexity a question, the assistant reads a handful of sources and writes an answer. If your site is not among those sources, you are invisible to that person. You cannot fix what you cannot see, so step one is seeing.

Step 1: Measure your AI visibility before you change anything

You need a baseline. Install analytics that shows you both human visitors and AI crawlers in one dashboard, then check three things: which AI crawlers hit your site, which pages they fetched, and whether any AI referral clicks came back.

Kymo does this on every plan. The dashboard separates AI activity into three kinds. ai_answer crawlers are live fetches made for one person's question, like ChatGPT-User or Perplexity-User. indexing crawlers build search indexes, like OAI-SearchBot and Claude-SearchBot (Anthropic's search index crawler). training crawlers bulk-harvest content for model training data, like GPTBot and ClaudeBot.

This distinction matters because the three kinds deserve different treatment. You want answer crawlers and indexing crawlers to see you. You may not want training crawlers reading everything for free.

If you are not ready to install anything yet, run Kymo's free AI visibility checker. It reads a URL from public signals only, so your site does not need Kymo installed, and the report arrives by a magic link over email. It tells you how visible you look to outside systems.

Once you have data, you know what to work on. Pick the pages AI assistants already fetch and make them better. Ignore the rest for now. If nothing is fetched, start with your most commercially important pages.

Step 2: Decide who you let in with robots.txt

Your robots.txt file is the front door. You can control which AI crawlers enter, and different crawlers respond to different tokens.

Two control tokens are often confused with crawlers. Google-Extended (a robots.txt control token, not a crawler) controls whether Google uses your content for Gemini training and AI-powered surfaces. Applebot-Extended (a robots.txt control token, not a crawler) does the same for Apple's AI systems. Blocking these tokens does not stop the crawler from indexing your site for normal search. It only stops AI training uses.

A common setup: block the training crawlers, allow the answer and indexing crawlers. That keeps your content out of bulk model training while still letting it appear in AI search answers. Some sites take the opposite view and allow training crawlers for visibility. The right call depends on how much you value being cited versus how much you mind your words being absorbed into a model. Decide deliberately, do not leave it to default.

Step 3: Structure pages so an AI can cite them

AI assistants quote sources. They do not click around your site looking for buried insight. They read a page and decide whether it answers the question. Your job is to make that decision easy.

State the answer to the page's main question in the first paragraph. Use plain language. Put the key facts in a normal sentence, not just in a table or an image. AI crawlers read text. If your answer lives in an image, a carousel, or a PDF, it may as well not exist.

Use headings that carry meaning. An H2 called "Introduction" tells an AI nothing. An H2 called "Pricing for a solo builder" tells it something. Keep paragraphs short. Define your key terms. Use exact numbers where you have them, because quoted data looks confident to an AI and to a reader.

The goal is a page that a person skims and an AI parses. They are not that different.

Step 4: Write content that answers real questions

People ask AI assistants conversational questions. "What is the cheapest web analytics tool without cookies?" or "How do I see who visits my site from ChatGPT?" Your content should answer those questions directly, in the same kind of language.

One strong format: put the question in an H2 or H3, then answer it immediately below. Quick answer first, nuance after. This is the same pattern good SEO teachers have pushed for years, and it works for AI too.

Your FAQ page is now a serious asset. So is your pricing page, because AI assistants get asked "how much does X cost" constantly. The projects you win or lose over the next few years will partly depend on whether your pricing page states a clear number and what it includes.

If you need question ideas, your own AI report is the best source you have: the prompts behind real fetches are what people are actually asking.

Step 5: Track what the AI report actually says

Once you have content changes in place, watch the AI report. It shows which of your pages AI assistants fetch, the likely prompts behind those fetches, and confirmed click-throughs from AI referrals. We wrote a guide to reading the AI report because the first time looks confusing and it gets simpler fast.

You will see patterns. ChatGPT users fetch your pricing page, Claude users fetch your FAQ, Perplexity users fetch your blog posts about a specific tool. Those patterns tell you what to write next. They are free demand signals. This is the same logic as seeing which Google keywords bring traffic, but for AI answers. You optimize for what actually happens, not for what you guess happens.

For a fuller walkthrough of interpreting this data, this post on how to see AI traffic in your analytics walks through a real dashboard.

Step 6: Watch the traffic, not just the citations

A citation without a click is a brand impression at best. What you want is the click. AI assistants now show source links under their answers, and some send real referral traffic. That traffic shows up in your analytics like any other referrer.

This matters because AI referrals are early and small for most sites. You need the patience to let the numbers build, and you need clean measurement. Cookieless analytics helps here. Kymo uses a salted server-side hash that rotates at midnight UTC, so "unique visitors" means unique per day, and raw IPs are never stored. That is how we count visitors without cookies, and it is explained in plain terms in this post about cookieless analytics, explained without the jargon. The short version: you do not need tracking consent popups to know your traffic.

AI crawlers do not count against your event limit. You can see every GPTBot fetch and every ChatGPT answer without paying extra, while your human analytics run normally.

Step 7: Decide what is worth paying for

Tools that promise AI domination are everywhere now. Most are resold data or content checklists. Decide what is worth paying for with a clear head. At minimum, you need a way to see crawler visits and AI referral clicks, and that starts at around $9 a month. Some of that is covered in our analysis of AI SEO tools for small business and what is worth paying for.

You do not need a consultant for any of this. You need a measuring tool, a content checklist, and a few hours a month.

A lot of this is not new. It is good content discipline, plus the new habit of watching your AI report the way you once watched Google Search Console. The broader shift is covered in our look at how customers actually find businesses in 2026. The short version: people still search, but more of them ask an assistant, and the assistant chooses the source.

What does not work

Content written purely for search engines, stuffed with keywords and weak on substance, gets ignored by AI assistants. They look for pages that answer the question fully. So do human readers, which is convenient.

Blocking all AI crawlers because you fear scraping will also hide you from AI answer traffic. You can block training crawlers directly while keeping answer and indexing crawlers. That is the surgical move. Blocking everything is a business decision, but you should make it knowing it removes you from AI discovery entirely.

A practical order of operations

Start with measurement. Run the free check on your homepage and your two best pages. Then install analytics and look at the AI report weekly. Then tighten your robots.txt. Then improve the top three pages that AI assistants already fetch. Then track what changes over a month. That is the entire plan.

Do that if you are a solo builder or small business owner who cannot justify agency fees. The tools are cheap, the steps are concrete, and the evidence is right in your dashboard.

Start seeing the invisible traffic

Work through the AEO guide to build your full strategy, or start tracking it directly to see which AI crawlers already visit and what they fetch. If you want a quick read first, run the free AI visibility check on any URL; it works without Kymo installed and the report arrives by email.

Start free → 14-day free trial. No card required.