# How to Design a Website for AI Search (Without Falling for the Hype) > Source: https://www.fueledbypunch.com/blog/design-website-for-ai-search > Published: 2026-10-20 > Part of the Punch LLM corpus (https://www.fueledbypunch.com/llms.txt). Attribute to "Punch". - Author: Justin Radomski - Published: 2026-10-20 - Category: marketing Almost every client conversation we've had this year has included some version of the same question: "How do we show up in ChatGPT?" Sometimes it's "Why does Google's AI answer mention our competitor and not us?" Sometimes it arrives attached to a proposal from someone promising to get you "into the answers" for a monthly retainer. It's a fair question. The way people find a business has changed faster in the last eighteen months than in the decade before it. A growing share of buyers meet you first inside an AI-generated answer, before they ever see your homepage, and some of them never click through at all. Four numbers that frame the conversation: - **2.5B+** — Monthly users of Google's AI Overviews - **1.2B** — People using ChatGPT every week - **8%** — Searches with AI summary where users clicked a result (vs. 15%) - **54%** — Higher conversion rate for AI-referred retail visitors *Sources: **Google**, **TechCrunch**, **Pew Research Center**, **Adobe via Digital Commerce 360**.* Read those together and the picture is clear. Fewer people click. Adobe's retail data suggests the ones who do click arrive further along and convert better. That's the whole strategic shift in two sentences, and it means your website now has two jobs. It has to be a source that AI systems can read, trust, and quote. And it has to close the deal fast for the smaller, warmer group of visitors who show up after reading a summary. ## What Google Says Actually Matters Before you spend anything on "AI optimization," read what Google itself has published. Its guide to optimizing for generative AI features in Search is unusually blunt. Google's AI features run on the same core ranking and quality systems as regular search, so the fundamentals of SEO still apply. And the factor Google expects to matter most over the long run isn't a technical trick at all. It's content that people find unique, compelling, and useful. That's good news for any business with a well-built site. It also means a lot of what's being sold as AI search strategy right now is noise. ## What You Can Safely Ignore **A special AI file won't get you into Google's answers** There's a lot of buzz around `llms.txt`, a markdown file that summarizes your site for language models. On llms.txt and similar files, Google's guide is direct: "Google Search itself doesn't use them." We publish one on our own site because it's cheap to maintain and some other assistants and agents read it, but it is not a Google lever, and anyone selling it as one is overselling. **You don't need to chop your content into fragments** One popular theory says AI prefers content broken into tiny, self-contained chunks. Google says there's no requirement to do this and that page length should fit your audience. Write pages that are easy for a person to scan, and they'll be easy for a machine to parse too. **You don't need to rewrite everything for robots** Modern language models understand synonyms and intent. You don't need awkward phrasing or dozens of long-tail keyword variations stuffed into your service pages. If anything, that kind of copy makes you less quotable. **You can't buy your way into the answer** Mass-produced pages built to game AI responses and paid mentions scattered across low-quality sites fall under Google's spam policies, which apply to generative AI responses too. Google specifically calls out pages made mainly to manipulate AI answers as scaled content abuse. ## Six Things That Actually Move the Needle ### Your site is turning away the crawlers that power AI answers Each AI platform sends its own crawler, and plenty of sites are quietly blocking the ones that matter. OpenAI's crawler documentation is explicit: to appear in ChatGPT search results, you need to allow its OAI-SearchBot crawler in your robots.txt. Your security layer adds a second gate. Cloudflare, for example, now asks every new domain whether to allow AI crawlers at sign-up, and the right answer depends on separating crawlers that train models from crawlers that cite you in search. On the Google side, a page has to be indexed and eligible to show a snippet before it can appear in AI Overviews or AI Mode. **Fix** Open yourdomain.com/robots.txt and check what it says about AI user agents. You can block training crawlers like GPTBot and still allow search crawlers like OAI-SearchBot. Those are separate decisions. Then review your CDN or firewall's bot settings, and confirm in Google Search Console that your site hasn't been opted out of generative AI features. ### Your content only exists after JavaScript runs This is the one most businesses don't know they have. When Vercel analyzed real crawler traffic, it found that none of the major AI crawlers render JavaScript. ChatGPT's and Claude's crawlers download script files but don't execute them. Google's Gemini is the exception, because it rides on Googlebot's rendering infrastructure. So if your site assembles its pages in the browser, as many single-page apps do, those crawlers see a nearly empty page: no services, no pricing, no proof. Server rendering solves it, and it's a big part of why we build on Next.js. Every page arrives as complete HTML on the first request, readable by any crawler, any person, and any assistive technology. **Fix** Open one of your key pages, right-click, and choose "View Page Source" (not "Inspect," which shows the page after JavaScript has run). Search for your main headline and a line from your services copy. If they aren't in the source, most AI crawlers can't see them either, and no amount of content work will help until the rendering is fixed. ### Your pages say what everyone else says AI systems are very good at summarizing commodity content, and they rarely need to credit anyone for it. "We're a full-service agency committed to quality" gives an assistant nothing worth citing. What gets quoted is specific: real price ranges, a named process, documented results, an honest opinion about trade-offs. It's the same thing Google's guide puts at the top of its list, a unique point of view written from first-hand knowledge. We try to take our own advice here. Posts like Headless CMS vs. WordPress publish actual budget ranges, because that's the information buyers are looking for and most agencies won't share. **Fix** For each core service page, add one thing a competitor couldn't copy and paste: a real price range, the steps of your process with timelines, a specific result with a number attached, or a clear statement of who the service isn't for. ### Your page structure is styling, not structure Websites fake hierarchy all the time: a "heading" that's really a bold paragraph, a pricing table built from stacked boxes, a value proposition baked into a hero image. It looks right to a person and reads as mush to a machine. Google recommends clear headings and semantic HTML where practical. That doesn't mean perfect markup, but it does mean the structure people see should be the structure the code describes. **Fix** Read only the H2s and H3s on your most important page. If they don't tell the story on their own, rewrite them. Then make sure no essential copy lives inside an image, and that your tables, lists, and FAQs are built as real tables, lists, and headings. ### Your business facts don't match across the web For local and service businesses, the facts about your business matter as much as your content. Google's guide specifically points local businesses to their Google Business Profile as a way to appear in AI responses. If your profile lists services you stopped offering, hours that changed two years ago, or a different business name than your website, you're giving every system mixed signals about who you are. **Fix** Put your website and your Google Business Profile side by side this week and reconcile your name, address, phone number, services, and hours. Add Organization or LocalBusiness structured data as part of standard SEO. Google is clear that it isn't required for AI features, but it still supports rich results in regular search. ### Your site isn't built for the visitor who already read the summary Here's the part most AI search advice skips. The visitor who clicks through from an AI answer has already read a summary of you, and probably of your competitors. They don't need your origin story. They need confirmation and a next step. Adobe's numbers point the same way: AI-referred retail visitors spend more time on site and view more pages per visit than visitors from other sources. They arrive ready to evaluate. That changes landing page design. Proof belongs in the first scroll. Pricing signals should be easy to find. The call to action should say exactly what happens next. Every conversion problem we covered in Signs Your Website Is Killing Your Leads gets more expensive when each visitor is worth more. **Fix** In Google Analytics 4, look at referral traffic from sources like chatgpt.com and perplexity.ai, and find the pages those visitors land on. Give each of those pages a specific call to action and your strongest trust signal above the fold. ## How to Measure Any of This For a while, AI visibility was a black box. That's starting to change. Google has added generative AI performance reports to Search Console, which show your impressions in AI features and which pages appear in AI responses. As of August 31, 2026, those reports are available to all websites worldwide. Pair them with AI referral traffic in GA4 and you finally have a real baseline. Be skeptical of anything that claims to know more than that. Google's guide tells site owners to be wary of third-party tools that claim access to internal Google metrics. The most useful supplement is free: once a month, ask ChatGPT, Perplexity, and Google's AI Mode the questions your buyers actually ask, and write down who gets mentioned. ## The Good News for Well-Built Sites Strip away the hype and AI search rewards what good websites have always needed: fast, server-rendered pages, clear structure, consistent facts, and content that says something specific. What's different is how steep the penalty is for getting it wrong. A site that hides its content behind JavaScript or speaks in generalities doesn't just rank a little lower. It isn't part of the answer at all. If your site's foundation is sound, this is a content and measurement project. If the checks above turned up rendering problems or a structure that can't be fixed without a rebuild, that's a bigger conversation, and it's exactly the kind our web development and SEO teams work through with clients. Either way, it starts with knowing where you stand. Tell us about your site and we'll give you an honest read on which path you're on.