mhp.si — AI Site Grade

Analysis

mhp.si: A 50-Year-Old Agency Rendered Invisible to AI

The site's heavy JavaScript-rendered architecture means most AI crawlers receive a 229KB payload but extract only 5-9 words of visible text from the homepage and key service pages, creating a near-total content vacuum for LLM ingestion despite technically passing HTTP status 200.

Crawler Access

The robots.txt at https://mhp.si/robots.txt contains no AI-bot-specific rules — a single User-agent: * with Disallow: (empty) allows everything. The llms.txt returns 404, meaning no AI-friendly content map exists. compare_bot_access showed all major AI crawlers (GPTBot, ClaudeBot, Google-Extended, PerplexityBot, OAI-SearchBot, ChatGPT-User, anthropic-ai, Applebot-Extended) receiving 200 status with full 229KB payloads — except Bytespider (403) and ClaudeBot (429) which hit Cloudflare rate limits. However, individual re-fetches of GPTBot and ClaudeBot both returned 429 Too Many Requests, indicating Cloudflare WAF rate-throttling that makes bot access unreliable in practice. The site runs on WP Engine behind Cloudflare (nginx), with no security headers (no HSTS, no CSP, no X-Frame-Options).

JS-Rendering Risk

The homepage, SEO service page, banking success stories page, and the Taylor King Law case study all return 5-9 words of visible text from a plain GET. The <title> and <meta description> are present, but the actual body content is rendered client-side. AI crawlers that do not execute JavaScript (most of them) see essentially nothing. The /about/ page fares slightly better (389 words) and /blog/ returns 586 words of actual content. The /services/performance-media/ page is the exception — it returns 667 words of server-rendered text, suggesting some pages use SSR while most do not.

Cold-Knowledge Gap

An LLM queried cold about "mhp.si marketing agency" returned: "I do not have specific, verifiable information about a marketing agency operating under the domain mhp.si" — guessing it might be Slovenian (.si TLD). The site is actually a 52-year-old Arkansas agency (founded 1972), formerly MHP/Team SI / Mangan Holcomb Partners, with 150+ employees, offices in Little Rock, Bentonville, Memphis, and Dallas-Fort Worth, a Google Premier Partner badge, and an Inc. 5000 track record. The cold model knows nothing about any of this. The .si domain (Slovenia's ccTLD) actively misleads AI geographic inference.

Schema Posture

JSON-LD Organization schema is comprehensive and well-structured — includes alternate names (MHP/Team SI, Mangan Holcomb Partners, Team SI, Social Innovation), founding date (1972), 8 named executives, 3 office locations with LocalBusiness sub-types, 9 awards, 2 industry memberships (4A's, Google Premier Partners), and social profiles. Service pages use Service schema with OfferCatalog items. However, the schema is embedded in every page's global footer block and references a 404 URL (/about/full-service-marketing-advertising-agency-in-little-rock/) as the Little Rock location's URL. The /about/our-team/ page also returns 404.

External Signals

Web searches for "mhp.si", "MHP/Team SI", and "Mangan Holcomb Partners" returned zero indexed results on DuckDuckGo. The site has no detectable external press coverage, Reddit threads, or review mentions in search results. The Wayback Machine shows a snapshot from March 2026, confirming the site has been live for at least a year. Despite being a marketing agency that presumably builds SEO for clients, mhp.si has no organic search footprint for its own brand.

Findings

1. Most pages render as empty JavaScript shells for AI crawlers High

   The homepage, SEO service page, banking success stories page, and Taylor King Law case study return only 5-9 words of visible text from a plain GET. AI crawlers that do not execute JavaScript see essentially no content.

   What to change: Implement server-side rendering (SSR) or static generation for key pages so that AI crawlers receive full HTML content without requiring JavaScript execution.

2. Cloudflare WAF rate-limits GPTBot and ClaudeBot with 429 responses High

   While compare_bot_access showed 200 status for most AI bots, individual re-fetches of GPTBot and ClaudeBot returned 429 Too Many Requests, indicating unreliable access due to Cloudflare rate-throttling.

   What to change: Configure Cloudflare WAF to allow AI crawler user agents with appropriate rate limits, or whitelist known AI bot IP ranges.

3. llms.txt returns 404, no AI-friendly content map Medium

   The site does not provide an llms.txt file, which would help AI crawlers discover and prioritize content.

   What to change: Create an llms.txt file listing key pages and their summaries to guide AI crawlers.

4. LLMs have zero knowledge of mhp.si as a marketing agency High

   A cold query about 'mhp.si marketing agency' returned no verifiable information, and the .si TLD misleads geographic inference toward Slovenia. The agency is actually a 52-year-old Arkansas firm with 150+ employees and multiple offices.

   What to change: Increase external signals (press coverage, backlinks, social mentions) and consider using a .com domain to avoid geographic confusion.

5. No organic search presence for the brand or its legacy names High

   Web searches for 'mhp.si', 'MHP/Team SI', and 'Mangan Holcomb Partners' returned zero indexed results on DuckDuckGo. The site has no detectable external press coverage, Reddit threads, or review mentions.

   What to change: Launch a PR and link-building campaign to generate external mentions and backlinks from reputable sources.

6. Key about pages return 404 errors Medium

   The pages '/about/full-service-marketing-advertising-agency-in-little-rock/' and '/about/our-team/' return 404, breaking schema references and user navigation.

   What to change: Restore or redirect these pages to working URLs, and update the schema to point to valid URLs.

7. Organization schema references a 404 URL for Little Rock location Medium

   The JSON-LD Organization schema includes a URL for the Little Rock office that returns 404, which may confuse crawlers and degrade schema trust.

   What to change: Update the schema's location URL to a valid page or remove it if no equivalent page exists.

8. No security headers (HSTS, CSP, X-Frame-Options) present Low

   The site lacks HSTS, Content-Security-Policy, and X-Frame-Options headers, which are not directly related to AI visibility but indicate a lack of security best practices.

   What to change: Add HSTS, CSP, and X-Frame-Options headers to improve security posture.

What's working

• Comprehensive JSON-LD Organization schema with rich details — The site includes well-structured Organization schema with alternate names, founding date, executives, office locations, awards, and memberships, providing rich context for AI crawlers that can parse it.
• Service pages include Service schema with OfferCatalog — Service pages use structured data to describe offerings, which helps AI understand the services provided.
• Performance Media page returns 667 words of server-rendered text — Unlike most pages, the Performance Media page delivers substantial server-rendered content, making it accessible to AI crawlers.
• Blog and About pages contain moderate amounts of readable text — The /blog/ page (586 words) and /about/ page (389 words) provide some server-rendered content, offering a baseline for AI crawlers.
• robots.txt allows all crawlers with no disallowed paths — The robots.txt file has a permissive policy (Disallow: empty), ensuring no crawlers are blocked by robots.txt rules.
• Sitemap available with 80 URLs — The site has a sitemap listing 80 URLs, which helps crawlers discover pages.