getmagic.com — AI Site Grade

Analysis

Magic (getmagic.com) — AI-Visibility Audit

The site has an llms.txt file and grants every major AI crawler full 200 access with real content, yet the cold LLM knowledge about the brand is stuck in early 2023 — it knows nothing about Magic 24/7, the AI-human hybrid model, or the pricing structure that the site now prominently features.

Crawler Access

All major AI bots — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Applebot-Extended, anthropic-ai — receive a full 200 response with the same 669KB payload as a browser. Bytespider (ByteDance) is the sole bot blocked at 403 by Cloudflare. The robots.txt contains no AI-bot-specific directives, only a generic Disallow: /wp-admin/ rule. The site runs on Cloudflare with strong security headers (HSTS, CSP, X-Frame-Options) and a 15.5M-second cache TTL. No JS-rendering risk exists: the homepage delivers 1,373 words of visible text on a plain GET.

llms.txt and Sitemap

An llms.txt exists at /llms.txt — generated by Yoast SEO v26.7 — and surfaces 50+ curated URLs across pages, posts, customer stories, eBooks, features, and services. This is an unusually mature AI-friendly signal for a WordPress site. The sitemap index contains 8 sub-sitemaps covering ~300+ URLs. The blog is actively publishing (most recent post dated May 6, 2026).

Cold-Knowledge Gap

The LLM's prior knowledge (trained through early 2023) describes Magic as a "subscription-based virtual assistant service" founded by Mike McDerment (FreshBooks co-founder), focused on US-based vetted talent. The actual site in 2025-2026 has evolved significantly: it now sells a hybrid human+AI model ("Magic AI" + "Magic 24/7"), offers three explicit pricing tiers ($199-$540/week), and positions itself as a virtual assistant agency with industry-specific assistants (transportation, warehousing, media, non-profit, construction). The cold model has no awareness of Magic 24/7, the AI assistant layer, the "top 0.5% of applicants" claim, or the "Magic Match" pairing system.

Schema Posture

Every page uses Yoast-generated JSON-LD with WebPage, Organization, BreadcrumbList, and WebSite types. The Organization schema includes sameAs links to Facebook, X, LinkedIn, and Instagram. However, no Product, Service, FAQPage, HowTo, or Review schema exists anywhere — despite the site having a pricing page with three plans, a comparison table ("Magic vs hiring in-person"), and a reviews page with dozens of testimonials. The reviews page contains 4,300+ words of user testimonials but has zero Review schema markup.

External Signals

The site has no discoverable press coverage from major outlets (TechCrunch, Forbes, Inc.) in current search results. The DNS TXT records reveal an anthropic-domain-verification token, confirming active ClaudeBot/Anthropic relationship management. The brand's X handle is @magic (a strong social asset). The reviews page aggregates tweets from notable figures (Austen Allred, Dana Loesch, Adora Cheung) but these are embedded as text quotes, not structured data. No Reddit threads or third-party review aggregator pages surfaced in search.

Findings

1. LLM cold knowledge stuck in early 2023, missing Magic 24/7 and AI-human hybrid model High

   LLMs trained through early 2023 describe Magic as a subscription-based virtual assistant service founded by Mike McDerment, with no awareness of the hybrid human+AI model, Magic 24/7, three explicit pricing tiers ($199-$540/week), or industry-specific assistants.

   What to change: Publish a detailed llms.txt with a comprehensive summary of the current product offerings, pricing, and AI-human hybrid model to update LLM knowledge.

2. No Product or Service schema on pricing page Medium

   The pricing page lists three plans with weekly rates but lacks Product or Service schema markup, reducing the chance of rich results in AI-generated answers.

   What to change: Add Product or Service schema to each pricing tier with name, description, price, and currency.

3. Reviews page lacks Review schema markup Medium

   The reviews page contains over 4,300 words of user testimonials but has zero Review schema markup, preventing search engines from displaying star ratings in rich snippets.

   What to change: Add Review schema to each testimonial with author, rating, and review body.

4. No FAQPage schema on how-it-works or feature pages Medium

   Pages like /how-it-works and /feature/magic-24-7-assistant contain FAQ-style content but lack FAQPage schema, missing an opportunity for rich results.

   What to change: Add FAQPage schema to pages with question-and-answer content.

5. No HowTo schema on how-it-works page Low

   The /how-it-works page describes a step-by-step process but lacks HowTo schema, reducing its visibility in AI-generated step-by-step answers.

   What to change: Add HowTo schema with steps, images, and estimated time.

6. Bytespider (ByteDance) blocked by Cloudflare Low

   Bytespider receives a 403 response from Cloudflare, preventing its content from being indexed by ByteDance's AI models.

   What to change: Allow Bytespider access if coverage in ByteDance AI products is desired.

7. No discoverable press coverage from major outlets Medium

   Web searches for press coverage from TechCrunch, Forbes, or Inc. returned no results, limiting external signals that could boost AI visibility.

   What to change: Pursue press coverage or publish case studies that are likely to be picked up by media.

8. No third-party review aggregator pages surfaced Medium

   Searches for reviews on Trustpilot, G2, or Reddit returned no results, reducing external validation signals.

   What to change: Encourage customers to leave reviews on third-party platforms and link to them from the site.

9. Customer testimonials embedded as plain text quotes, not structured data Low

   The reviews page includes quotes from notable figures but they are plain text, not marked up with schema, limiting their impact on AI-generated answers.

   What to change: Add Review schema to each testimonial quote.

What's working

• llms.txt file published with 50+ curated URLs — An llms.txt generated by Yoast SEO surfaces 50+ curated URLs across pages, posts, customer stories, eBooks, features, and services, providing a strong AI-friendly signal.
• All major AI bots receive full 200 access with real content — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, Applebot-Extended, and anthropic-ai all get a 200 response with the same 669KB payload as a browser, with no JS-rendering risk.
• Robots.txt has no AI-bot-specific blocks — The robots.txt only disallows /wp-admin/ and contains no directives blocking AI crawlers, ensuring maximum discoverability.
• JSON-LD Organization schema with sameAs links — Every page includes Yoast-generated JSON-LD with Organization schema containing sameAs links to Facebook, X, LinkedIn, and Instagram, helping AI models verify brand identity.
• Anthropic domain verification token present — DNS TXT records include an anthropic-domain-verification token, indicating active relationship management with ClaudeBot/Anthropic.
• Brand owns @magic handle on X — The brand's X handle is @magic, a strong social asset that can be used for brand signals and verification.
• Blog actively publishing with recent content — The blog is actively publishing, with the most recent post dated May 6, 2026, providing fresh content for AI crawlers.
• Sitemap index with 8 sub-sitemaps covering 300+ URLs — The sitemap index contains 8 sub-sitemaps covering approximately 300+ URLs, ensuring all pages are discoverable.