Designing a SaaS website that AI answers can quote
How to structure a SaaS site so ChatGPT, Perplexity and Google AI Overviews can understand and quote it: page per question, answer first, prices in text.
- Guide
- October 2026
- 4 min read
AI assistants quote pages that answer one question clearly, near the top, in plain text. Design the site around the questions buyers ask: one page per question, the answer first, prices and facts in text, and pages a crawler can read without running JavaScript. Nothing guarantees a citation, but these choices make one possible.
When a buyer asks ChatGPT, Perplexity or Google’s AI Overviews which tool to use, the answer is built from pages the assistant can read and understand. Most SaaS sites are designed for a person scrolling a homepage, not for a system looking for one clear answer to one question. Most guides on this come from the SEO side. This one is about the design decisions, and we use our own site as the worked example.
One page per buyer question
"What does it cost?", "Does it work with X?", "How is it different from Y?" Each deserves its own page or section with a clear heading.
Collect the questions from sales calls, support tickets and the queries in Search Console. Group them, and give each group a page or a section. Phrase the heading the way the buyer would ask it: "Does it work with Salesforce?" says more than "Integrations", to a person and to a machine.
This is also good information architecture for people. A buyer who wants the price shouldn’t have to read the founding story first.
Answer first, then explain
Open each page with a two or three sentence answer. On formform.co, every guide starts with an "In short" box for exactly this reason.
The answer should stand on its own. No "as mentioned above", no pronouns that depend on the paragraph before, and the product name spelled out. An assistant may lift that one passage without anything around it, so it has to make sense alone.
The writing side of this, with a before and after, is in How to make your work legible to AI.
Write every answer as if it will be quoted without the paragraph around it. It might be.
Put facts in text
Prices, limits, integrations and dates in real text, not in images, tabs that load later, or PDFs.
Tabs and accordions are fine when their content is in the HTML as the page loads. They’re a problem when the content is fetched only after a click. Pricing tables made as images, comparison charts exported from a slide and specs in a PDF are hard or impossible to quote. On our services page, prices are sentences: brand identity from US$6k, websites from US$10k.
Date the facts that change. A price or a feature list with a visible date is easier to trust, for a buyer and for an assistant deciding which source is current.
Structure that machines can read
- Structured data: Organization, Service or Product with Offers, FAQPage, Article.
- An llms.txt that lists your key pages and what each answers.
- A clean sitemap with real last-modified dates.
On this site, every guide carries Article and FAQPage data built from the same text you read on the page, the llms.txt lists each page with a one-line description, and the sitemap dates come from when each page’s source last changed. None of it makes a weak page quotable. It removes guessing from a clear one.
Does the stack matter?
Less than people think. What matters is that the content is in the HTML on first load. Framer, Webflow, WordPress and server-rendered custom code can all do that. Check by viewing the page source.
View the source, not the inspector, which shows the page after scripts have run. Search it for your main answer and your price. If they’re there, crawlers that don’t run JavaScript can read them. If you only find an empty container and a script tag, the page depends on JavaScript, and some crawlers will see nothing. Our site is a single-page app, so we added a step that writes every route to static HTML at build time for exactly this reason.
What can you expect?
Nothing guarantees a citation. Assistants choose their sources in ways that change often and aren’t published. What you control is whether your page has a clear, current, quotable answer when they look. A simple way to track it: ask the assistants your buyers’ questions once a month and note which sources they cite.
A checklist for an AI-quotable site
- A list of the questions buyers ask, each with a page or a headed section.
- Headings phrased the way buyers ask.
- A short answer at the top of every page that makes sense on its own.
- Prices, limits, integrations and dates in text, with dates where they change.
- Organization, Service or Product, FAQPage and Article structured data.
- An llms.txt and a sitemap with real last-modified dates.
- Main answers and prices visible in the page source.
- A monthly check of what the assistants say and whom they cite.
If you want a site designed around the questions your buyers actually ask, send your brief.
Asked & Answered
Often because the answer isn’t on the page in plain words, or it’s buried under marketing copy.
Around the questions buyers ask: one page or clearly headed section per question, a short answer at the top, prices and facts in text, structured data, and content that is in the HTML when the page first loads.
They help machines understand the page. They don’t replace a clear answer.
Not if the content renders in the HTML.
Keep reading
- How to make your work legible to AISearch is becoming answer engines. Being citable by machines is now a design discipline.
- What a SaaS or AI marketing site needs at launchThe pages, homepage message and proof a SaaS or AI marketing site needs at launch, what to skip, and how to make it readable for AI assistants.
- SaaS pricing page design: what makes buyers pick a planHow to lay out a SaaS pricing page: how many plans, what goes in the comparison table, where the FAQ goes, and how to make prices readable for AI answers.