Building SiteWhisper — AI chatbot that answers from your real website pages
Hey IH — solo founder here.
I built SiteWhisper because most “AI chatbots” invent answers. Mine crawls your site and only answers from what’s already on your pages (plus PDFs / Q&A you add). One script tag. Free plan with no card.
What I’m doing to sell it so far:
- Mint a live 30-day demo on the prospect’s own site
- Quote a sentence the bot pulled from their pages
- Email the founder as Michael from michael@trysitewhisper.com
It’s working better than FAQ-style cold email. Still early — free users and a handful of demos out.
Product: https://sitewhisper.online/
Curious what would make you try this on your own site tomorrow — pricing, a Shopify one-click, or something else?
For me, a useful demo would include one question the site can't answer, not just one it can. A prospect could ask about a refund exception that isn't documented and see whether the bot admits the gap and offers a handoff. That would help me judge the 'only answers from your pages' promise before installing it.
Thanks, yes — that’s the behavior I want to show. SiteWhisper won’t invent a refund exception: the dashboard shows unanswered prompts, and I can add the missing Q&A/PDF (or hand it off) before it answers. That’s the main proof of the “only from your pages” promise.
That helps. I'd show the visitor's side of that flow too: what they see when the bot can't answer, and how the handoff continues. An unanswered prompt in the dashboard helps the owner, but I'd want the visitor to leave with a clear next step rather than an unexplained wait.
Cool concept — are you already live with a landing page, or still in build mode? Curious how you're planning to explain it to non-technical visitors.
Thanks! We’re live now with a landing page and an interactive demo. I’m keeping the explanation simple: add one script tag, we crawl the site (plus any PDFs/Q&A), and it answers from that content only. The live demo quoting their own page makes it easier for non-technical folks to get.
Your demo is a measurement instrument disguised as a product demo. When you quote a sentence the bot pulled from their pages, you're not proving capability - you're proving measurement precision. That's why it beats generic cold email: most "AI solutions" sell you a promise; you're selling them a read-back of their own data. Measurement source (crawled vs claimed) determines trust credibility.
Thanks, that’s exactly the trust loop I’m aiming for. We crawl the site and the live demo quotes a sentence from the customer’s own pages, so it’s a read-back of crawled data—not a generic promise. When content is missing, the dashboard shows unanswered prompts so I can add Q&A/PDFs rather than invent.
The personalized demo seems like a meaningful signal, especially if it’s outperforming generic FAQ outreach. When prospects respond, is the “only answers from your site” trust angle what gets them interested, or is the live demo itself doing most of the selling?
Thanks, I think the live demo does both. It runs on the prospect’s site and quotes their own pages, so they can see the source instead of taking a generic claim on faith. If there’s a gap, we don’t invent an answer.
That’s a stronger signal — the demo is effectively proving the trust claim rather than just explaining it. I’d be interested in digging into whether that changes what prospects actually do next. Happy to continue privately — what’s the best email to reach you on?
I would test an agency workflow before prioritizing a Shopify one-click install. Let someone crawl one client site, then show which answers are grounded in current pages and which pages are stale or conflicting. Agencies need to trust the source before putting a bot in front of a client, and a small report they can share in a review call may be a stronger wedge than another install option.
Thanks, that’s a great agency workflow. We support multiple sites on paid plans (one script per site), and the dashboard surfaces unanswered prompts so you can add Q&A/PDFs instead of guessing. Shopify’s script tag works now; the App Store app isn’t shipped yet.