thrivelia.com — AI Site Grade

Analysis

I have enough to write a thorough audit. Let me compile the findings.

The Cold-Knowledge Hallucination Gap

Thrivelia's site is fully open to every AI crawler tested — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, ChatGPT-User, OAI-SearchBot, Bytespider, Applebot-Extended, and anthropic-ai all return 200 with full content (1.3–1.5 MB each, no blocking, no JS shell). The robots.txt has a blanket Allow: / for * and names zero AI bots — no GPTBot, no ClaudeBot, no PerplexityBot, no Google-Extended rules at all. The site also has an llms.txt that is unusually sophisticated: it documents a Wix MCP endpoint (/_api/mcp) with 6 available tools for AI agents to query business details, search the site, and even call Wix APIs on a visitor's behalf. An openai-domain-verification TXT record confirms OpenAI has verified the domain for potential GPT Actions or custom GPT integration.

Cold-Knowledge Hallucination Gap

When queried cold, a frontier LLM described Thrivelia as a retention agency founded by Val Geisler — a well-known retention marketer who actually founded a completely different agency (Val Geisler / Retention Marketing). The LLM also hallucinated a "Retention OS" framework and claimed the agency has "limited press coverage" and "no widely known client list." In reality, Thrivelia is run by Trayan (named repeatedly in case studies and the challenge page) and has five detailed, public case studies with named clients (Pulsetto, Sintra AI, Crucial FOUR, BionicGym, VoomVaya) showing specific revenue lifts — $9K to $800K/month for Sintra, $1.5M YoY growth for BionicGym, $332K added in 30 days for VoomVaya. The cold model knows nothing accurate about the actual founder, clients, or results.

Schema and Content Fragmentation

The homepage has only two schema types — LocalBusiness (with address country "BG" — Bulgaria) and WebSite — but no Service, Product, FAQPage, Review, or AggregateRating schema anywhere on the site. Every case study page, the services page, the pricing page, and the challenge page have zero JSON-LD. The /pricing page shows placeholder text ("I'm a benefit" repeated) with generic Silver/Gold/Platinum plans at €350–€500 — contradicting the homepage's positioning as a high-end agency for 6-to-8-figure DTC brands. The /contact page contains Wix default placeholder text ("I'm a paragraph. Click here to add your own text...") still live. The /blog page returns "Check back soon" — no posts exist. The /services page yields only 57 words of visible text, suggesting heavy JS rendering that leaves AI crawlers with thin content despite the large HTML payload.

External Signal Void

Web searches for "Thrivelia" return zero results on DuckDuckGo — no press mentions, no Reddit threads, no reviews, no third-party articles, no directory listings. The site has no external backlink footprint visible via search. Despite claiming "150+ brands thrive with us" and operating "on all 7 continents," the brand has no discoverable external presence outside its own domain. This creates a paradox: the case studies are substantive and specific, but no third party has written about or reviewed the agency anywhere public.

Wix Infrastructure and JS-Rendering Risk

The site runs on Wix (server header "Pepyaka", Wix request IDs, Wix static assets). Several key pages (/services, /casestudies, /pricing, /contact) return very thin visible text (19–148 words) from a plain GET despite large HTML payloads (650KB+), indicating client-side rendering that may leave AI crawlers with incomplete content. The llms.txt MCP endpoint is a forward-leaning solution to this — it gives AI agents a structured API path to bypass the JS rendering issue — but standard crawlers hitting those pages via sitemap will find near-empty shells.

Findings

1. Cold LLM knowledge hallucinates founder, framework, and client results High

   A frontier LLM queried cold described Thrivelia as founded by Val Geisler with a 'Retention OS' framework, while the actual founder is Trayan and the site has five detailed case studies with specific revenue lifts. The model knows nothing accurate about the founder, clients, or results.

   What to change: Publish structured data (Person, Organization, CaseStudy, Review) and ensure key facts are prominently visible in plain HTML to ground AI models.

2. Zero external backlinks or third-party mentions found High

   Web searches for 'Thrivelia' return zero results on DuckDuckGo — no press mentions, Reddit threads, reviews, or directory listings. Despite claiming '150+ brands thrive with us,' the brand has no discoverable external presence outside its own domain.

   What to change: Build backlinks through guest posting, PR, directory listings, and client testimonials on third-party sites.

3. Key pages return thin visible text due to client-side rendering High

   Pages like /services, /casestudies, /pricing, and /contact return only 19–148 words of visible text despite large HTML payloads (650KB+), indicating heavy JS rendering that leaves AI crawlers with incomplete content.

   What to change: Implement server-side rendering (SSR) or static generation for key pages to ensure AI crawlers receive full content.

4. No JSON-LD schema on case studies, services, pricing, or challenge pages High

   The homepage has only LocalBusiness and WebSite schema. Every case study page, the services page, the pricing page, and the challenge page have zero JSON-LD. This limits AI understanding of the agency's offerings, results, and credibility.

   What to change: Add JSON-LD schema (Service, Product, FAQPage, Review, AggregateRating, CaseStudy) to all relevant pages.

5. robots.txt does not name any AI crawler Medium

   The robots.txt has a blanket Allow: / for * and names zero AI bots — no GPTBot, ClaudeBot, PerplexityBot, or Google-Extended rules. While currently all bots are allowed, the lack of explicit rules means the site cannot selectively manage AI crawler access.

   What to change: Add explicit rules for AI crawlers (e.g., GPTBot, ClaudeBot) to allow or disallow as desired.

6. Pricing page contains placeholder text and contradicts premium positioning Medium

   The /pricing page shows placeholder text ('I'm a benefit' repeated) with generic Silver/Gold/Platinum plans at €350–€500, contradicting the homepage's positioning as a high-end agency for 6-to-8-figure DTC brands.

   What to change: Replace placeholder text with real pricing tiers, benefits, and value propositions that match the premium brand positioning.

7. Contact page contains Wix default placeholder text Medium

   The /contact page includes Wix default placeholder text ('I'm a paragraph. Click here to add your own text...') still live, which appears unprofessional and may confuse AI crawlers.

   What to change: Remove all default placeholder text and replace with actual contact information and copy.

8. Blog page returns 'Check back soon' with no posts Medium

   The /blog page returns only 'Check back soon' — no blog posts exist. This misses an opportunity for content marketing and AI visibility.

   What to change: Publish blog posts regularly to build topical authority and attract AI crawlers.

9. Services page has only 57 words of visible text Medium

   The /services page yields only 57 words of visible text, suggesting heavy JS rendering that leaves AI crawlers with thin content despite the large HTML payload.

   What to change: Ensure services page content is server-rendered or statically generated to provide full text to crawlers.

10. Case studies listing page has only 19 words of visible text Medium

    The /casestudies page returns only 19 words of visible text, likely a JS shell, making it difficult for AI crawlers to discover individual case studies.

    What to change: Implement SSR or static generation for the case studies listing page to expose links to individual case studies.

What's working

• All AI crawlers allowed with full content access — Every tested AI crawler (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.) returns 200 with full content (1.3–1.5 MB each), no blocking, and no JS shell. The site is fully open to AI indexing.
• llms.txt documents Wix MCP endpoint with 6 tools for AI agents — The llms.txt is unusually sophisticated: it documents a Wix MCP endpoint with 6 available tools for AI agents to query business details, search the site, and call Wix APIs. This forward-leaning solution helps AI agents bypass JS rendering issues.
• OpenAI domain verification TXT record present — An openai-domain-verification TXT record confirms OpenAI has verified the domain for potential GPT Actions or custom GPT integration, indicating proactive AI readiness.
• Five detailed case studies with specific revenue lifts — The site has five detailed, public case studies with named clients (Pulsetto, Sintra AI, Crucial FOUR, BionicGym, VoomVaya) showing specific revenue lifts — $9K to $800K/month for Sintra, $1.5M YoY growth for BionicGym, $332K added in 30 days for VoomVaya. These provide strong social proof.
• Sitemap submitted with 24 URLs — A sitemap.xml is present and contains 24 URLs, helping search engines and AI crawlers discover site content.
• LocalBusiness schema on homepage with address — The homepage includes LocalBusiness schema with address country 'BG' (Bulgaria), providing basic location data to AI crawlers.