28 October 2026 EN ES
The First Users

How products get their first real users

Channels

AI Answers: A First-Hundred User Check

A founder checklist for testing whether AI answers can send your first users: category fit, shallow comparable pages, and signup tagging.

Illustration: AI Answers: A First-Hundred User Check

You have no distribution yet. You are still replying to every signup by hand. So when someone says AI answers will be your growth channel, you should be suspicious. AI answers can be a real first-user signal, but only when your category gets AI referrals, your key page is shallow and comparable, and you can tag the signups separately.

AI platforms sent 770.7 million referral visits per month to websites worldwide from June 2025 through May 2026, up 117.4% from the prior year. Generative AI sites received 9.5 billion monthly visits worldwide over the same period, up 70% from the prior year. That is a large audience, but it is not a promise that your product will appear in an answer. For a founder with no distribution, the useful question is narrower: can this send a handful of real signups while you are still doing the unscalable work?

The category decides whether AI can matter

Start with the category, not the model. Marketplaces drew the largest AI referral volume, averaging 46.8 million monthly visits from June 2025 through May 2026, with news and travel close behind. Marketplaces grew 237.3% year over year in AI referral traffic, even though they already had the largest volume base. Beauty posted the steepest year-over-year gain in AI referral traffic, up 312.5% from June 2025 to May 2026, while fashion climbed 278.2%. Products near a marketplace, a comparison-heavy consumer category, or a category where people ask for recommendations may matter.

Look for your category in the AI referral traffic ranking. Absent or only a tiny slice of the ranking, do not treat AI as a channel. Present and growing, move to the page check.

The page must answer the question directly

58.8% of AI referrals reached homepages, whereas ChatGPT-cited pages were usually located two or three folders down. That split is a warning. A pricing page, a comparison page, or a product page that answers a direct question has a better chance of being cited. A buried blog post with a signup form at the bottom asks a model to do research for you. Shallow means the page is easy to find from the homepage, not hidden behind a chain of subfolders. A comparable page is one a model can check against other answers.

Web-source citations in ChatGPT answers increased by more than five times over a year, reaching 6.8%. The trend is moving in your favor, but the base is still small enough that you should not plan a launch around it. For the first hundred, treat citations as a bonus signal, not a substitute for asking people you know why they would use the product.

Tagged signups are the only proof

Tracking AI signups is the part founders skip because it feels like growth work before there is growth. Do it anyway. Without a separate view of AI-referred signups, you will mistake a lucky mention for a channel. Add a simple segment for AI domains, or add a source tag to links you share in places where AI answers may point back. Then watch for a pattern: repeated signups from the same answer, the same category, or the same page.

Write down the source answer you can see. A screenshot may not prove a channel, but it can prove a question worth answering. Keep the note next to the signup so you can compare the question, the page, and the result.

Run the three checks before you build

  • Category fit: your category appears in AI referral traffic rankings with meaningful volume or growth.
  • Page shape: your key page is a comparable, checkable page, not a deep, buried page.
  • Tracking: you can tag AI-referred signups separately in analytics.

Do this check before you spend another hour on prompts. Pass all three: treat AI answers as a small referral source and put your best answerable page in front of the questions your users already ask. Fail any: do not build a content engine around a rumor.

Advertisement