Catch AI-assistant first users without ads

You don't need a paid funnel to catch the first users an AI assistant is already sending you. If someone asks an assistant for a tool and your product gets named, that click is a high-intent visit. Treat it like a small, hand-built page. ChatGPT referral traffic produced a 4.7% lead conversion rate, while Google organic traffic produced a 1.9% lead conversion rate, a roughly 3x difference.

Most founders miss it because the visit often arrives as Direct traffic. The assistant's answer points to your site, but the browser does not always tell you where the person came from. That makes the traffic look like a random walk-in, even when it is the warmest lead you have.

The referral is hiding in your analytics

First Page Sage measured ChatGPT referral conversion rates using anonymized data from more than 160 client companies collected from May 2025 through September 2026. Across 3.4 billion sessions on 218 client sites, B2B SaaS organic sessions declined 29% to 29.4% of a normalized baseline, while AI platforms supplied 11.8% of SaaS sessions.

In a First Page Sage study covering more than 150 companies, visitors arriving from AI platforms converted at 14.2%, while Google organic visitors converted at 2.8%. Ahrefs' internal data showed that AI search visitors made up 0.5% of total traffic but generated 12.1% of signups, a 23x conversion advantage. The first hundred users do not need a growth engine. They need a page that matches the question, a tag that shows the source, and a reply that sounds human.

Attrifast's 200-site cohort found that 71% of visits referred by ChatGPT reached sites without referrer headers and were classified as Direct traffic in GA4. Attrifast's platform pass-through data, which tracks whether a referrer survives the click, showed Perplexity preserved referrers 50–70% of the time, Gemini 40–60%, Claude 30–50%, ChatGPT 15–30%, and Google AI Overviews less than 3%.

Build a page that matches the assistant's answer

Start with the questions your product actually answers. Write down the assistant-style prompts that would lead to your tool: a comparison, a use case, a workflow, or a narrow problem. If you cannot name the prompt, you cannot build the page. If you can, you have your first page. Do it before the next AI referral.

An assistant sends someone who has just read a recommendation, not a generic homepage visitor. The page should answer the same question the assistant answered, in plain language, and then offer the next step. If the assistant said your tool helps teams write release notes, the page should open with that use case, not a logo and a tagline.

Keep the page narrow. A narrow page can be tested, updated, and followed up by the same person. The unscalable part is the job now.

The conversion action should be obvious. A trial, a demo request, a form, or a direct email address works. Do not ask the visitor to learn your whole product. Answer the question, show the relevant proof, and make the next step easy enough to take on a phone.

Run the checklist before the next spike

Use the smallest setup that keeps this traffic from disappearing into Direct.

• Find the assistant-style question in your traffic by checking Direct visits and new search terms.
• Make a dedicated page that answers that question and ends with a clear action.
• Add a source tag or referrer-capture script so assistant visits can be separated from Direct.
• Set a conversion event for signup or trial and view it by source.
• Reply personally to each early lead and note what the assistant said.

When a lead arrives, pick up the same thread. Ask what they are trying to do, send the relevant page, and offer a human reply. The manual follow-up is also the signal. It tells you whether the assistant is sending the right people.