How to Optimize Your Website for AI Search

Ellie at Willa, Editorial lead, Willa · Published July 28, 2026 · Updated August 25, 2026

More and more of your audience is skipping Google's blue links and asking ChatGPT, Claude, or Perplexity instead, and getting a single synthesized answer. If your website isn't built for how those engines read, you're not in it. At Willa's live session "Master AI Search Visibility: How to Be Found in the New Era of Search," Emily Richardson, founder of Known (KNWN), with over five years of marketing experience and a practice helping brands shift from traditional SEO to AI visibility, walked through how AI search works under the hood and her three-step process for optimizing your website for it.

How is AI search different from Google search?

Emily's shorthand: "Google uses words and AI uses math."

On Google, you type a 4 to 7 word query, Google matches it against its keyword index, ranks links, and you click through and synthesize an answer yourself. AI search queries average 27 to 35 words, about five times longer. The engine runs a "query fan-out," breaking your question into two to four sub-questions it searches in parallel. It then uses retrieval-augmented generation (RAG) to retrieve chunks of content, filter and rank them, and synthesize one answer. RAG works on vectors and embeddings, mathematical similarity rather than keyword matching. As Emily put it: "In AI search, you are the answer. You're not just a link."

And the two systems genuinely don't mirror each other. Emily shared an Ahrefs study that ran 15,000 prompts and found that, on average, 81% of the web pages surfaced by large language models aren't ranked on Google for the same question. Only 8% overlap with Google's top 10. You can get cited by AI while ranking poorly on Google, and vice versa.

The behavior shift is already here. Emily cited figures showing 71% of people in the US have tried AI search and 14% use it daily; when Google shows an AI Overview, 60% of those searches end without a single click; and 41% of consumers trust AI search more than paid ads, rising to 68% among regular AI users.

Her most encouraging point: unlike Google, where ads increasingly priced smaller players out of the results page, AI search has no ads yet. Everybody starts in the same place, so small brands can get ahead right now.

How do I write content that AI search will cite?

Step one of Emily's process: content. To get cited, your content needs to sound like a helpful expert: clear, conversational, closer to a forum answer or tutorial than a polished ad. Build trust the EEAT way (expertise, experience, authority, trust): show first-hand experience, cite credible sources, and acknowledge nuance instead of pushing your product.

Her sharpest filter: "Create content that only you can make." AI can't go into the world and have experiences. That's only us. So lead with proprietary data, real-world experience, and unique perspectives. Her test: if you can single-shot prompt the whole article with zero expertise and zero unique data, it shouldn't be on your website. The AI can generate it itself, so it has no reason to cite you.

Emily shared a telling example from her client work with Owala water bottles. When Known ran 100 prompts across the four major AI search platforms, the second most-cited URL wasn't Owala's website (that came fourth). It was a mom's personal blog of detailed, week-long water bottle reviews: how they fit in the car, what her kids liked, whether the teacher allowed them in class. AI valued that first-hand experience over the brand's own polished pages.

How should I structure my content so AI bots can read it?

AI bots don't read your whole page the way Google does. They grab only the chunks that align with the query. Emily's structural rules follow from that:

• Front-load everything. She cited data that 44% of AI citations come from the first 30% of page content. "AI bots are lazy. They do not want to read the full page."
• Turn headings into questions. Headers phrased as questions are twice as likely to be cited. Make your H1s, H2s, and H3s match what someone would actually ask, and put a summary answer directly underneath each one.
• Use bullets, numbered lists, and tables. Listicles and tables make up roughly 50% of citations for B2B queries in ChatGPT. The AI would rather reuse a page that's already structured than synthesize a judgment from ten sources.
• Add an FAQ section. Emily cited a 40% increase in citation weight for pages with FAQs. Easy version: paste your article into whatever AI you use and have it generate the FAQs. Redundant? Doesn't matter. Add them.
• Keep pages focused. Around 900 to 1,400 words per page is the sweet spot, and when one member asked whether multiple linked pages beat an infinite-scroll site, Emily confirmed it: shorter, clearly-headed, interlinked pages win.

She drove the point home with a live exercise: attendees played AI bot and hunted for "what are the benefits of an e-bike?" on two pages with similar content. On the dense wall of text, hands went up slowly; on the structured page, almost instantly. "Think if you were a robot: how would you like to read this?"

What technical setup does my website need for AI search?

Step two is making sure bots can actually get in. Most sites are fine, Emily stressed, but it's miserable to optimize for months and then discover AI was blocked the whole time. Her checklist:

1. Check your robots.txt. Add /robots.txt to your domain and search the page for "disallow." "If bots can't crawl you, you do not exist." If GPTBot or other AI crawlers are disallowed, fix it in your platform's settings.
2. Keep key content in plain HTML. AI bots can't process JavaScript-rendered text, so important content needs to be server-side rendered. Quick check: paste your page's HTML into Claude or ChatGPT and ask whether any content is JavaScript-only.
3. Add alt text and transcripts. AI doesn't watch videos. Embed your YouTube videos or podcasts on your site with the full transcript. That transcript is citable content.
4. Load fast. Emily cited data that pages loading in under 0.4 seconds average 6.87 AI citations versus 2.1 for pages over one second. She recommends targeting about half a second.
5. Tidy your architecture. Keep important pages one to two clicks from your homepage, use descriptive internal links (like /guides/protein-types: "AI loves guides"), and don't leave orphan pages with nothing linking to them.

How do I build AI visibility beyond my own website?

Step three: broadcast far and wide. Large language models don't rely on links the way Google does. They rely on patterns of words across the whole web. "They don't just read your website, they read the internet's version of you."

The practical consequences:

• Keep your brand identity identical everywhere. Use the same name, phrasing, and description across your website, LinkedIn, Crunchbase, and every profile, so AI sees one unified brand. Emily shared a cautionary client story: Klarity Health (spelled with a K) is still being confused by ChatGPT with an unrelated Klarity Diagnostics, because AI matches on the name itself, not the link. If you're naming a business right now, check that nobody in your industry shares it.
• Cultivate user-generated content. Emily cited data that domains with a Reddit presence are four times more likely to be cited by ChatGPT, and those with G2, Trustpilot, or Capterra presence three times more likely. Independent blog reviews and YouTube round-ups are gold. One warning from the Q&A: don't try to seed Reddit yourself. Emily spent a week attempting it and got flagged on every comment. Leave it to specialist agencies.
• Repurpose one answer everywhere. Pick a high-intent question you want to own, and if you're a small brand, niche it down the funnel ("is plant-based protein as effective as whey for people who are pregnant"). Publish it on your domain with schema markup, summarize it on LinkedIn, record a YouTube short, post it on Medium, and pitch PR earlier than you think you should. Publishers are getting cited heavily in AI search.
• Consider YouTube for niche B2B. If nobody has made a video about your specialty (Emily's example was reverse osmosis chemistry), you'll get cited almost by default.

How can I start optimizing for AI search this week?

Emily closed with a do-it-now exercise. Find a real question people ask about your industry (Google's "People Also Ask" boxes are a good source) and niche it down to your audience. Make that question a heading on your site, answer it in the first sentence underneath, then add tips and examples below, formatted with the structure rules above.

Then ask that same question in ChatGPT and open the Sources tab to see which sites it's pulling from. "It often will just show you the roadmap to how to get cited": the listicles you could pitch, the blogs and formats winning for your query.

Key takeaways

• AI search works mechanically differently from Google (longer queries, query fan-out, retrieval by vectors), and 81% of AI-cited pages don't rank on Google for the same question, so it needs its own strategy.
• Publish content only you could make: first-hand experience, proprietary data, a real point of view. If one AI prompt could generate the whole article, it won't get cited.
• Structure for lazy bots: question-phrased headings with the answer front-loaded, bullets and tables, an FAQ section, focused 900 to 1,400-word pages.
• Do a one-time technical check: robots.txt allows AI crawlers, key content in server-rendered HTML, transcripts for video, sub-half-second load times, no orphan pages.
• AI reads the internet's version of you, not just your site. Keep your brand identity consistent everywhere and amplify one answer across LinkedIn, YouTube, Medium, review sites, and PR.

This was part one of Emily's AI search series at Willa. Her follow-up session covers the other half of the equation: how to pick queries you can actually win, score them, and track whether ChatGPT mentions your brand. And if the level-playing-field theme resonates, our session on getting app downloads with no marketing budget runs on the same scrappy-distribution thinking.

Join us for the next one

This guide came from a live Willa event, where members played AI bot in Emily's crawling exercise, asked questions about everything from Reddit to pricing, and are already voting for a follow-up workshop on their own websites. Willa hosts expert-led sessions like this almost every week for women building careers with AI. See what's coming up on our events page, or become a member and learn it with us in real time.

• AI search
• AEO
• SEO
• ChatGPT
• content strategy
• marketing

Join Willa, the community where women learn AI together.

All posts