← The AI Hype Audit — all 268 verdicts

PARTLY Partly real: 3D scroll sites are real tech, but 'comment RABBIT for the prompt' is a ManyChat auto-DM sales funnel, not a one-prompt website Facebook · Sep 2026

The claimReel: "Client: I don't want basic website / Me: How's this?" over a bioluminescent 'Enter the Rabbit Hole' 3D scroll site. Caption: "Build this Custom 3D scroll Website — Follow and Comment 'RABBIT' for the prompt AND workflow — DM me to get a modern website built for your business!" Sep 5, 2026.

The reel is a textbook comment-to-DM funnel. ManyChat's own guide (published Sep 16 2024, updated Jul 6 2026) describes it exactly: a viewer comments a keyword, a bot DMs 'your link, your offer, or your freebie in seconds,' optionally gated behind a follow. The caption's real ask is the last line: 'DM me to get a modern website built for your business.' The 'prompt and workflow' is the lead magnet; the developer's paid build is the product. Meta's Transparency Center classes posts that 'explicitly request engagement (such as votes, shares, comments...)' as engagement bait, and a 1kreach analysis (Apr 26 2026) calls the 'comment X to get the link' DM funnel the most penalised pattern on Reels (though it cites no data). On the tech: scroll-driven 3D sites are real and well documented. Codrops (Nov 19 2025) shows the stack (Three.js/R3F, GSAP ScrollTrigger, GLB models, custom shaders) and rates it intermediate-to-advanced. A 2026 developer guide puts a basic version at 2-4 hours with prerequisites, a commercial build at 16-40 hours, and self-learning to commercial quality at 6-12 months. Claude Code can scaffold Three.js+GSAP scroll pages (MindStudio, Mar 17 2026: 'an afternoon, under $10', but with 3-4 revision rounds and a warning that video+Three.js+scroll animations 'equals slow page on mid-range devices'). No single prompt hands you a finished, hosted, branded 3D site with your own model. Agency pricing for a focused 3D hero site: $20K-$50K, 8-12 weeks (Utsubo cost guide). And Utsubo's May 20 2026 SEO guide notes an empty canvas 'for the first three seconds' makes LCP three seconds; their fix was a static image on mobile.

What holds up

  • Funnel mechanic confirmed: ManyChat 'comment keyword -> auto-DM link/freebie' guide (Sep 16 2024, updated Jul 6 2026); ManyChat Help documents Reel comment triggers and a follow-gate before the link is sent.
  • Meta Transparency Center defines engagement bait as posts that 'explicitly request engagement (such as votes, shares, comments, tags, likes...)'; 1kreach (Apr 26 2026) says the 'comment X to get the link' funnel is the most penalised Reels pattern (no data cited).
  • Tech stack verified: Codrops tutorial (Nov 19 2025) builds a cinematic 3D scroll site with Three.js/R3F, GSAP ScrollTrigger, GLB models and GLSL shaders, rated intermediate-to-advanced, source on GitHub.
  • Effort verified: a 2026 developer guide lists a basic scroll-3D build at 2-4 hours (skills assumed), commercial first build 16-40 hours, standard project 4-6 weeks; MindStudio (Mar 17 2026) says Claude Code can scaffold one in an afternoon for under $10 in tokens, with 3-4 revision rounds.
  • Cost and performance verified: Utsubo cost guide prices a focused 3D hero site at $20K-$50K / 8-12 weeks (custom 3D adds $15K-$80K); Utsubo SEO guide (May 20 2026) says an empty canvas for three seconds means a three-second LCP and recommends a static image on mobile.

What doesn’t

  • The keyword comment triggers a bot, not a human; the DM will contain a link (freebie, template, community, or booking page), and the caption's closing line shows the goal is selling you a build.
  • 'The prompt' cannot ship a finished site: you still need a 3D model (Spline export or .glb), GSAP ScrollTrigger tuning, mobile fallbacks, a domain and hosting, and several debugging rounds.
  • Heavy WebGL on a local-business site is a Core Web Vitals liability: the canvas can become the LCP element, and main-thread shader compile and .glb parsing hurt INP, especially on phones.
  • Meta treats 'comment X to get it' posts as engagement bait; whatever reach the reel earned came from that mechanic, not from proof the workflow works for real clients.
  • No client site, no live URL, no before/after is shown, just a demo animation that matches freely available Codrops/Wawa Sensei/Spline tutorials.

The catch

Comment RABBIT and a bot DMs you a link; the 'prompt' gets you a generic 3D demo at best, and the real offer is 'DM me to build your site.' A 3D scroll hero is achievable with tools like Claude Code plus Spline, but it takes hours to weeks of skilled iteration and can slow the one thing a local business site must do: load fast on a phone.

How to actually do it

  • Do not comment the keyword. If you want the 'prompt,' know it is bait for a sales DM; ask for a live client URL and run it through PageSpeed Insights on mobile before taking it seriously.
  • Decide if you even need 3D: for a local business, a fast, mobile-first site with strong Google Business Profile and clear booking/contact beats a WebGL hero. Ask 'what does this animation do for my customer?'
  • If a 3D moment genuinely fits your brand, scope it as one hero element with a static-image mobile fallback, and budget real hours (16-40 for a commercial build) or agency money ($20K+), not '$0 and a prompt.'
  • Test any delivered site on a mid-range Android on cellular; if LCP is over 2.5s or scrolling stutters, the 3D is costing you leads and rankings.

Expect a bot DM with a link and a pitch; the technique is real, the 'one prompt' is not, and a 3D scroll site is usually the wrong spend for a local business.

Confidence
Medium
Posted by
a web-dev creator page

See the original claim →

We test hype for free. We build the real thing for a living.

Thirty minutes, no pitch — and you'll leave with something useful either way.

Book a call with Todd or start with the free Business Checkup →

The Verdict Weekly

Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.

© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.