zepz.io — AI Site Grade

Analysis

Zepz.io is a holding-company corporate site that AI crawlers can read freely, but the cold-knowledge gap is severe — LLMs describe Zepz as a consumer remittance app (the old WorldRemit identity), while the site positions itself as a B2B payments group powering WorldRemit and Sendwave, with no consumer-facing product on the domain itself.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, ChatGPT-User, anthropic-ai, Applebot-Extended — receive a full 200 response with identical HTML content (59,778 bytes) to a browser baseline. Google-Extended and Bytespider are blocked with HTTP 429 (rate-limit wall), a selective exclusion that prevents Google's AI training crawler from reading the site while allowing its search crawler through. The robots.txt is a bare Yoast-generated file with a single User-agent: * Disallow: rule and no AI-bot directives at all. No llms.txt exists (404). The site runs on WordPress behind nginx on Cloudflare DNS (ns.cloudflare.com) with WP Engine hosting (A records at 192.0.78.x, WP Cloud SPF include). No JS-rendering risk — all pages serve full HTML content on first GET.

Cold-Knowledge Gap

An LLM queried cold about "Zepz" recalls it as a consumer-facing remittance app (formerly WorldRemit, rebranded 2021), serving 130+ countries with mobile-to-mobile transfers, and mentions 30% workforce layoffs in 2023 and a $5B valuation. The actual site tells a fundamentally different story: Zepz is a holding group that "powers two leading global remittance brands, WorldRemit and Sendwave," serving 9+ million customers across 5,000 corridors. The homepage has no product demo, no app download link, no pricing, no sign-up flow — it is a corporate brochure with investor relations, press releases, and a careers portal. The cold model knows nothing about the Sendwave Wallet, the stablecoin strategy with Circle/USDC, the $165M HSBC financing in 2025, or the $267M Series F. The gap between the model's prior (consumer app with layoff baggage) and the site's reality (B2B group with stablecoin infrastructure) is the single largest AI-visibility problem.

Schema Posture

Every page carries Yoast-generated JSON-LD with WebPage, BreadcrumbList, WebSite, and Organization types. The Organization schema includes a logo but no sameAs URLs, no foundingDate, no description, no knowsAbout — the organization's description field is empty string. Blog articles use Article schema with author, headline, and datePublished. No FAQPage, Product, SoftwareApplication, or FinancialService schema exists anywhere. The press release pages have no NewsArticle or PressRelease schema — they use generic WebPage. The homepage has no meta description (null).

Content Quality Issues

The press-releases/ archive page contains only 6 words of visible text ("Press Releases All Categories Partners Press") — a JavaScript-rendered filter UI with no server-side fallback. The news-and-insights/ archive is similarly thin (14 words). A press release about the Circle partnership has a duplicate slug (-copy suffix) that actually contains a different article (the HSBC $165M financing), creating a confusing canonical mismatch — the slug says "zepz-and-circle-join-forces" but the page content is about HSBC. The homepage links to https://zepz.io/Array which returns a 404. The add-article-here-copy/ slug is a published article titled "How migrants power the global economy" — the slug is clearly a placeholder that was never renamed.

Findings

1. LLMs describe Zepz as a consumer remittance app, not a B2B holding group High

   Cold LLM recall describes Zepz as a consumer-facing remittance app (formerly WorldRemit) with layoff baggage, while the site positions itself as a B2B payments group powering WorldRemit and Sendwave. The model knows nothing about the Sendwave Wallet, stablecoin strategy, or recent financing.

   What to change: Publish an llms.txt file and structured data that explicitly defines Zepz as a holding group, its subsidiaries, and key facts (funding, corridors, stablecoin partnerships).

2. Google-Extended and Bytespider blocked with HTTP 429 High

   Google's AI training crawler (Google-Extended) and Bytespider receive HTTP 429 rate-limit responses, preventing them from reading the site. This blocks Google's AI from using the site content for training.

   What to change: Remove the rate-limit blocking for Google-Extended and Bytespider, or explicitly allow them in robots.txt.

3. No llms.txt file published Medium

   The site returns a 404 for /llms.txt, missing an opportunity to provide AI crawlers with a curated summary of the site's content and key facts.

   What to change: Create an llms.txt file that describes the site as a corporate holding group, lists key pages, and provides a brief overview of Zepz's business.

4. Press releases archive page has only 6 words of visible text High

   The /press-releases/ page contains only 6 words of visible text, likely a JavaScript-rendered filter UI with no server-side fallback, making it nearly empty for crawlers.

   What to change: Ensure the archive page renders a full list of press releases server-side, with titles and dates, as a fallback for crawlers.

5. News and insights archive page has only 14 words of visible text High

   The /news-and-insights/ page contains only 14 words, similarly likely a JS-rendered filter with no server-side fallback.

   What to change: Ensure the archive page renders a full list of articles server-side.

6. Duplicate press release slug serves different article Medium

   The URL /press/zepz-and-circle-join-forces-to-reimagine-global-remittances-with-stablecoin-technology-copy/ contains an article about HSBC financing, not the Circle partnership. The slug suggests a duplicate that was never cleaned up.

   What to change: Remove the duplicate page or update the slug to match the content, and set a canonical URL to the correct article.

7. Published article uses placeholder slug 'add-article-here-copy' Medium

   The article 'How migrants power the global economy' is published at /add-article-here-copy/, a slug that was clearly a placeholder and never renamed.

   What to change: Rename the slug to a descriptive, keyword-rich URL and set a 301 redirect from the old slug.

8. Homepage links to /Array which returns 404 Low

   The homepage contains a link to https://zepz.io/Array, which returns a 404 page. This is likely a broken link or placeholder.

   What to change: Remove the broken link or point it to a valid page.

9. Homepage has no meta description Medium

   The homepage's meta description is null, which can hurt search snippet quality and AI understanding.

   What to change: Add a meta description that summarizes Zepz as a holding group for WorldRemit and Sendwave.

10. Organization schema lacks description, sameAs, and foundingDate Medium

    The JSON-LD Organization schema has an empty description field and no sameAs URLs or foundingDate, limiting how AI models understand the entity.

    What to change: Populate the Organization schema with a description, sameAs links (e.g., Wikipedia, Crunchbase), and foundingDate.

11. Press release pages lack NewsArticle or PressRelease schema Medium

    Press release pages use generic WebPage schema instead of NewsArticle or PressRelease, missing an opportunity for rich results and AI understanding.

    What to change: Add NewsArticle or PressRelease schema to press release pages with headline, datePublished, and author.

12. No FinancialService or Product schema for remittance offerings Low

    The site does not use FinancialService or Product schema to describe its remittance brands or services, which could help AI models understand the business.

    What to change: Consider adding FinancialService schema for WorldRemit and Sendwave on relevant pages.

What's working

• All pages serve full HTML content without JavaScript dependency — Every page returns complete HTML on first GET, making all content accessible to AI crawlers without requiring JavaScript rendering.
• Major AI crawlers (GPTBot, ClaudeBot, PerplexityBot) allowed with 200 responses — GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, ChatGPT-User, anthropic-ai, and Applebot-Extended all receive full 200 responses with identical HTML to a browser.
• Sitemap is accessible and contains 28 URLs — The sitemap at /sitemap_index.xml returns 200 and lists 28 URLs, helping crawlers discover content.
• Every page includes Yoast-generated JSON-LD schema — All pages carry WebPage, BreadcrumbList, WebSite, and Organization schema, providing structured data for search engines and AI.
• Blog articles use Article schema with author and date — Blog posts include Article schema with headline, author, and datePublished, aiding AI understanding of content.
• robots.txt allows all user agents with no disallowed paths — The robots.txt file has a single rule allowing all user agents, ensuring no accidental blocking of crawlers.
• Individual press release pages contain substantive content (500+ words) — Press release pages like the Circle partnership and HSBC financing contain 500+ words of detailed text, providing rich information for AI.
• Careers page provides extensive content (1809 words) — The /careers/ page contains 1809 words of detailed information about the company and job opportunities, useful for AI understanding.