On launch day, an AI startup's website needs to do five things: say in one plain sentence what the product does and for whom, show the product working, give some signal of price, explain what happens to the user's data, and offer one clear next step. Behind that, it needs to be crawlable by Google and by AI assistants from the first hour, with public docs if developers will use it. Everything else, from the blog to the careers page, can wait a few weeks.

What goes on the first screen?

People decide fast. Nielsen Norman Group found that users often leave web pages within 10 to 20 seconds, and concluded that a page must communicate its value proposition within 10 seconds to earn minutes of attention. On launch day, many visitors arrive from one post or one link and give you exactly that long.

The first screen needs four things:

1. A headline that names the job. "Turn sales calls into CRM notes before the next meeting" passes. "AI-powered revenue intelligence" fails, because every competitor could write it.
2. One line on who it is for. "For sales teams on HubSpot" tells the right visitor they are in the right place, and tells the wrong one to leave.
3. A look at the real product. A screenshot, a short loop or a play button on a demo, showing the result, not a dashboard full of empty charts.
4. One primary button. "Try it free" or "Join the waitlist". A second, quieter link such as "Watch the demo" is fine. Five buttons is not.

Our Possena home page replaced a typical AI headline with one sentence in the visitor's own words and put the starting box right on the first screen. Our guide on making software's value clear in the first minute goes through the full page.

How do you show proof when you have no customers yet?

You cannot show logos you do not have, and you should never invent them. A launch-day site can still give visitors plenty of reasons to believe it:

• The product's own output. A real before and after: the messy input, the finished result. For an AI product this is often the strongest proof there is.
• The founders. Real names, real photos, what they did before and why they built this. Google's guidance on helpful content says trust is the most important part of E-E-A-T, encourages accurate authorship information, and calls fabricated creator profiles, such as AI-generated headshots or made-up credentials, a form of deception (Google Search Central).
• Your own test results, with the method. A line such as "We ran X real support tickets through it and Y% needed no edits" is believable if you say how you tested. Only publish numbers you measured.
• Beta users and design partners, quoted by name only with their written permission.
• Honesty about the stage. "Launched this week. Free while in beta." Visitors forgive being early. They do not forgive being misled.

How should you show the product working?

For an AI product, a demo is not optional. Visitors have seen too many promises and want to watch the thing work. Give them, in order of priority:

1. A 60 to 90 second recorded demo near the top of the home page, captioned so it works with the sound off.
2. A way to try the product with no setup, if at all possible: paste a document, upload a file, type a question.
3. Optionally, a clickable product tour for people who want to explore without signing up.

Show the product on realistic inputs, including one where it handles something tricky well. If it sometimes gets things wrong, say what the user can do about it. That honesty reads as confidence.

Do you need pricing on launch day?

You need a price signal, even if final pricing is not set. Nielsen Norman Group found that business customers rank pricing as the information they need most online, that they go to competitors' sites when a site hides prices, and that companies hiding costs come across as evasive. Showing a price also tells people whether they are in the right product category at all.

Any of these works on day one:

• Published plans with a free tier.
• "Free during beta. Paid plans will start at about $X per user per month."
• "From $X a month" with a note on what changes the price.
• For enterprise-only products, a typical range for a typical team size.

"Contact us for pricing" as the only option is the weakest choice for a product nobody has heard of yet.

What should you say about trust and data handling?

For AI products, data handling is a buying question, not a legal footnote. In Cisco's 2024 Data Privacy Benchmark Study of 2,600 privacy and security professionals, 27% said their organization had banned generative AI applications for the time being, and 63% had set limits on what data can be entered. If your buyer works at a company like that, the security team will read your site before anyone signs up.

Write a plain-language page, linked from the footer and from the sign-up screen, that answers:

• What data you collect and why.
• Whether customer data is used to train any model, yours or a provider's.
• Which AI model providers and other subprocessors handle the data.
• Where data is stored, how long you keep it, and how a user deletes it.
• Who to contact about security, with a real email address.

Only claim certifications you hold. If a SOC 2 audit is in progress, say "in progress", not "compliant". A privacy policy and terms of service should be live before the first sign-up.

Waitlist or open sign-up?

If you choose a waitlist, still show the demo and the price signal. A waitlist with nothing behind it asks for trust without giving any reason.

Do you need docs on launch day?

If developers will connect to your product, yes. Docs are part of the product, and for many technical buyers the docs are the real home page. At minimum, publish:

• A quickstart that gets to a working result in a few minutes.
• An API or integration reference, with real request and response examples.
• Limits, errors and pricing for usage.
• A changelog, even if it has one entry.

Keep docs public and in normal HTML text, not only inside a logged-in app or a PDF. That is what lets Google and AI assistants answer "Can it do X?" correctly. Non-technical products still need a short help page that answers the first ten questions people will ask.

How do you get found by Google and AI assistants from day one?

There is no special trick, and that is good news. Google says that to appear in AI Overviews or AI Mode a page must be indexed and eligible to show in Search with a snippet, and that there are no additional requirements, no special schema, and no need for new machine-readable files or AI text files (Google Search Central). So the basics carry most of the weight:

1. Do not block crawlers by accident. Pre-launch sites are often left behind a password or a "noindex" tag. Remove both on launch day.
2. Allow the AI search crawlers in robots.txt. OpenAI says OAI-SearchBot is used to show websites in ChatGPT search and recommends allowing it; GPTBot is separate and relates to model training. Anthropic says blocking Claude-SearchBot may reduce your visibility in Claude's search results, while ClaudeBot relates to training. You can allow search crawlers and block training crawlers if you prefer.
3. Put the important words in text. What it does, who it is for, pricing, integrations and data handling should be readable HTML, not only inside images or video.
4. Set up Google Search Console and submit a sitemap. Google's SEO starter guide notes that a sitemap is optional and that some changes take hours while others take months, so do this on day one, not day thirty (Google Search Central). Check that you are indexed by searching site:yourdomain.com.
5. Give every page a unique title and description that says what the page is about in plain words.
6. Get mentioned elsewhere. Launch posts, directories, founder profiles and partner pages are how both Google and AI assistants first learn you exist. Make sure each one describes the product the same way your home page does.

An llms.txt file is a community proposal, not a formal standard. It does no harm, but it will not replace any of the steps above. Our guide on getting recommended by ChatGPT covers what happens after launch.

Launch-day checklist

What we can't promise

A good launch site makes sure the people who arrive understand the product and can try it. It does not create demand on its own, and no one can promise when Google, ChatGPT or Claude will start recommending a brand-new product. That depends on crawling, on how often others mention you, and on whether the product is genuinely the best answer to someone's question. Project 66 Studio builds the launch website and parts of the product such as onboarding, demos and forms; building a whole company or product is the work of Project 66 Venture Studio.

How Project 66 Studio can help

We build launch websites for software and AI companies, with the demo, price signal, data page and Google and AI search setup done before day one. A new website starts from $799, and setting up SEO, AEO and GEO together is $990 once. See our work for software and AI, check pricing, or start a project.

Build your plan Start a project

Questions people ask

What is the most important part of an AI startup's launch website?

The first screen. Nielsen Norman Group found that people often leave pages within 10 to 20 seconds, so the headline must say what the product does and for whom, with a look at the real product and one clear button. If that screen fails, nothing further down the page gets read.

Should an AI startup launch with a waitlist or open sign-up?

Open sign-up is better when someone can get value alone in a few minutes, because you learn whether people actually use the product. A waitlist makes sense when the product is not ready for strangers or costs too much to run at scale. Either way, show the demo and a price signal so visitors have a reason to trust you.

How can a startup show social proof with no customers?

Show the product's real output, named founders with real backgrounds, and test results you measured yourself with the method explained. Quote beta users only with written permission. Never invent logos, testimonials or numbers, which Google's guidance treats as deception.

Do I need an llms.txt file to appear in ChatGPT or Google AI Overviews?

No. Google says there are no special files, schema or AI text files needed to appear in AI Overviews or AI Mode; the page must simply be indexed and eligible for a snippet. For ChatGPT search, OpenAI recommends allowing its OAI-SearchBot crawler in robots.txt. llms.txt is a community proposal, not a formal standard.

What should an AI product say about data privacy on its website?

Say in plain language what data you collect, whether it trains any model, which AI providers and subprocessors handle it, where it is stored, how long you keep it, and how users delete it. Cisco's 2024 study found 27% of organizations had banned generative AI apps for the time being, so many buyers' security teams will check this page first. Only claim certifications you actually hold.

How much does a launch website cost at Project 66 Studio?

A new website or rebuild starts from $799. Setting up SEO, AEO and GEO together costs $990 once, or one of them $490 once. We confirm the final price before any work begins.

Read next

• How to make a SaaS demo video that sells
• Make your software's value clear in the first minute
• Possena: an AI website that shows its value in the first minute

Sources

1. How Long Do Users Stay on Web Pages?, Nielsen Norman Group (read 26 Oct 2026).
2. State the Price to Give B2B Sites a Competitive Advantage, Nielsen Norman Group (read 26 Oct 2026).
3. Creating helpful, reliable, people-first content, Google Search Central (read 26 Oct 2026).
4. More than 1 in 4 Organizations Banned Use of GenAI Over Privacy and Data Security Risks, Cisco (read 26 Oct 2026).
5. AI features and your website, Google Search Central (read 26 Oct 2026).
6. SEO Starter Guide, Google Search Central (read 26 Oct 2026).
7. Overview of OpenAI Crawlers, OpenAI (read 26 Oct 2026).
8. Does Anthropic crawl data from the web, and how can site owners block the crawler?, Anthropic (read 26 Oct 2026).
9. The /llms.txt file, llmstxt.org (read 26 Oct 2026).