AI SearchUpdated 9 min read
AI search optimization: how to do it, in 8 steps
AI search optimization is the work of making your pages easy for AI models to find, read and quote, then checking whether they do. In order: let the crawlers in, put the facts in the HTML, write for the questions your buyers ask, lead each section with the answer, add numbers, quotes and sources, publish the pages models tend to cite, get mentioned on the pages they read, and measure it on each AI model. Most of it is SEO you already know. The last step is the new one.
What does AI search optimization mean?
It means getting your brand named, and your pages cited, in the answers that ChatGPT, Claude, Gemini, Perplexity and Google's AI Overviews and AI Mode write. The same job is sold as AEO (answer engine optimization), GEO (generative engine optimization), LLM SEO and AI SEO. The labels differ; the work doesn't. For the definitions and where each term came from, see what is GEO and AEO vs SEO.
This post is the how-to. It assumes your site already works as a website: pages that load, that are indexed, that say something clear.
The 8 steps at a glance
| Step | What you do | How you check it |
|---|---|---|
| 1. Let the crawlers in | Allow the search crawlers of Google, OpenAI, Anthropic and Perplexity | Read your robots.txt |
| 2. Put the facts in the HTML | Prices, tables and answers present before any script runs | View the page with JavaScript off |
| 3. Pick the questions | Write down what a buyer asks a model, in their words | A list of 10 to 20 prompts |
| 4. Lead with the answer | The first sentence under each heading answers the heading | Read each section's first line alone |
| 5. Add numbers, quotes, sources | Replace vague claims with checkable ones | Count the claims that name a source |
| 6. Publish the pages models cite | Pricing, comparison and "X vs Y" pages | Do these pages exist on your site? |
| 7. Get mentioned elsewhere | Appear on the lists and directories AI answers read | Look at which pages your prompts cite |
| 8. Measure on each model | Ask the prompts on a schedule and count | Share of answers that name you |
The 8 steps
1. Let the crawlers in
An AI answer can only quote a page its crawler can fetch. According to OpenAI's crawler docs, OAI-SearchBot decides whether a page can appear in ChatGPT's search answers. Anthropic, Perplexity and Google have their own: Claude-SearchBot, PerplexityBot and Googlebot. Open your robots.txt and make sure none of them is disallowed on the pages you want cited. Blocking a training crawler and allowing a search crawler is a separate choice from blocking both.
2. Put the facts in the HTML
OpenAI's and Anthropic's crawlers download JavaScript files but don't run them, according to Vercel's crawler data (December 2024), which may have changed since. A price, a feature table or an answer that only appears after scripts run doesn't exist for them. Turn JavaScript off in your browser and reload the page. If the fact is gone, move it into the server-rendered HTML.
3. Pick the questions
Models answer questions, so start from the ones your buyers ask. Write down 10 to 20 in their words: "best [category] for [type of company]", "[your product] vs [competitor]", "how much does [category] cost". Open ChatGPT, Gemini and Perplexity, ask each one, and note who is named and which pages are cited. That list is your work plan for the rest of the steps.
4. Lead with the answer
The first sentence under a heading should answer the heading, in a form that still makes sense quoted alone. "AEO tools cost from $29 to $599 a month" beats three paragraphs of context. Google's own guide to optimizing for its generative AI features says the advice for AI Overviews and AI Mode is the same as for search: it's still SEO, with no special file and no rewriting content for AI.
5. Add numbers, quotes and sources
This is the step with research behind it. The 2023 paper that named GEO, GEO: Generative Engine Optimization, rewrote source pages in different ways and measured how much of each AI answer came from them.
| Change to the page | Visibility score (higher is better) |
|---|---|
| Keyword stuffing | 17.8 |
| No change | 19.5 |
| Easy-to-understand wording | 22.2 |
| Citing sources | 24.9 |
| Adding statistics | 25.9 |
| Adding quotations | 27.8 |
Stuffing keywords scored slightly worse than leaving the page alone. Numbers, quotes and sources scored best. These are scores from the paper's Table 1 (position-adjusted word count), from a lab benchmark on 2023-era engines, so read them as a direction. What to do: replace "many customers" with the figure and its source, and quote the person or document the claim comes from.
6. Publish the pages models cite
Models don't cite pages evenly. In our run of 30 buyer questions in ChatGPT, 18 of the 87 pages it cited were pricing pages (the data). Most of the citations that weren't a vendor's own site were comparison articles and "best X" lists. If you don't have a pricing page that states the price, a page that compares you with the alternatives, and a page for each question buyers ask, you've left the pages models cite most to someone else.
7. Get mentioned elsewhere
Many of the pages an AI answer cites aren't yours: the "best X" lists, directories, reviews and comparison articles. Look at which pages your prompts cite and which of them you're missing from. Getting onto them is closer to PR and link building than to writing: pitch the list's author with a fact they can use, or get listed in the directory. Reddit is the exception worth knowing: it ranked in Google's top 10 for 28 of the 30 questions but ChatGPT cited it 0 times in our run, though Google's AI Overviews still do.
8. Measure on each model
Rankings don't tell you whether a model names you. Of the 87 pages ChatGPT cited in our run, 2 were in Google's top 10 for the same question. Run your prompts on each AI model on a schedule, and count the share of answers that name your brand and which pages they cite. Answers change from one run to the next, so one run proves little and a trend over weeks tells you something. ChatGPT tags its source links with utm_source=chatgpt.com (all 87 in our run), so you can filter its visits in your analytics. Some AI apps send no referrer, and their visits land in "Direct". Aleyda Solis's AI search checklist builds a step around the split between being cited and being clicked.
By hand this works for about 5 prompts. Past that, a tool does the counting; the AEO tools compared covers what each one reads and costs.
What doesn't work
- Stuffing keywords. It scored below doing nothing in the GEO paper.
- Writing a separate version of the page for AI. Google says it's not needed, and a page that reads badly to people reads badly to models.
- Treating one AI model as all of them. In our own answers for two products, 91% of the sites cited were cited by one AI model only. A result on ChatGPT says little about Gemini or Perplexity.
- Promising a result. Nobody controls what a model says tomorrow. Anyone selling "rank #1 in ChatGPT in 30 days" is guessing with your money.
What does ChatGPT say when you ask it?
On October 11, 2026 we asked ChatGPT, with web search on, "how to optimize for ai search". Its answer opened with a list of eight strategies, the first being to create content that directly answers a specific question, which is step 4 here. Of its first 15 source links, 7 went to Google's own pages and 3 to OpenAI's documentation. None was one of the guides that rank on Google's first page for the same search, apart from Google's. The pages that explain how the systems work are the pages the systems cite.
When isn't AI search optimization worth your time?
When the SEO basics aren't done. If your pages aren't indexed or answer nothing clearly, fix that first. It helps both.
When your buyers don't ask AI. A local trade where everyone phones for a quote gets more from Google Business Profile than from ChatGPT.
When you want a guarantee. There isn't one.
How we checked
- Google: the US results page for "ai search optimization" and "how to optimize for ai search" on October 11, 2026, read through cloro. Top results include Google's guide, NP Digital, Semrush, Reforge, Walker Sands, HubSpot and Digital Marketing Institute.
- Search volume: 1,300 a month for "ai search optimization", difficulty 33; 210 for "how to optimize for ai search" (DataForSEO Labs, US, October 2026).
- ChatGPT: the ChatGPT app, US, web search on, the same day, through cloro. One run; ask again and the sources can change.
- The paper: read on arXiv on October 11, 2026; scores from Table 1, position-adjusted word count.
- Our 30-question run: September 30, 2026, method and limits in GEO vs SEO.
- Not tested: whether following these steps raises your citations. They follow the sources linked; we haven't run a controlled test on a page.
Step 8 is the part that's tedious by hand and the part an AI visibility tool does for you. Picked asks your buyers' questions every day on the AI models in your plan (Claude less often), shows which pages each answer cited, and writes the article for a question you lose. To see where you stand before anything else, run the free AI visibility checker.
Frequently asked questions
What is AI search optimization called?
It goes by several names: AEO (answer engine optimization), GEO (generative engine optimization), LLM SEO and AI SEO. They describe overlapping work: getting your brand named and your pages cited in answers written by AI models.
How do I optimize for AI search?
Let the AI companies' search crawlers fetch your pages, put your facts in the HTML, write for the questions your buyers ask, lead each section with the answer, add numbers, quotes and sources, publish the pages models cite most (pricing, comparisons), get mentioned on the lists AI answers read, and measure the result on each AI model.
Is SEO still worth it in 2026?
Yes. Google's guidance for AI Overviews and AI Mode is that it's still SEO, and ChatGPT reads pages its crawler can find. What changed is that ranking alone doesn't tell you whether AI answers name you: in our 30 questions, 2 of 87 pages ChatGPT cited were in Google's top 10.
Do I need a tool for AI search optimization?
Not to start. Ask ChatGPT, Gemini and Perplexity your buyers' questions and write down who is named and which pages are cited. That works for about five prompts. Past that, counting takes longer than fixing, which is where a tool helps.
How is AI search optimization different from SEO?
SEO is judged by where a page ranks. AI search optimization is judged by whether an answer names you and cites your page. The work overlaps heavily; what's added is writing for quotability and measuring per AI model. AEO vs SEO compares the two.
