martingroup.co — AI Site Grade

Analysis

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

The Martin Group (martingroup.co) — AI-Visibility Audit

The About page states the agency was founded in 2001, but the cold LLM knowledge (and many external sources) consistently cite 1991 — a decade-long discrepancy that creates a fragmented origin story across AI engines. The site launched a full redesign in spring 2026 yet still leaks its Pantheon dev environment (dev-tmg-2024.pantheonsite.io) in image URLs across multiple pages, including the About, Sports, and Healthcare sections.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, Applebot-Extended, Bytespider — receive a full 200 response with identical byte-size content (177,585 bytes) as a browser. No UA-based blocking exists. The robots.txt is a bare Yoast-generated file with a single User-agent: * Disallow: rule and no AI-specific directives whatsoever. No llms.txt exists (404). The site runs on Pantheon (nginx + Varnish) with a 60-second cache TTL, meaning AI crawlers hit cached copies, not stale origin renders.

Cold-Knowledge Gap

The LLM prior describes The Martin Group as founded in 1991 with offices in Buffalo, Rochester, and Albany, specializing in brand strategy, creative, digital, PR, and media. The actual site says founded in 2001 (the acquisition blog post explicitly states "Founded in 2001" and the About page says "Established in 2001"). The cold knowledge mentions CEO Todd Martin — but the site's leadership page lists Tod Martin (no second 'd') as the top team member, and the COO is Lisa Bellacicco. The cold knowledge also cites M&T Bank and University at Buffalo as notable clients; the site's work page does list M&T Bank but not UB, while featuring New Era Cap as the dominant client (appearing 15+ times in the work portfolio). The cold knowledge omits the Hearst minority investment (2020) entirely — a major structural fact the site prominently features.

Schema Posture

Every page carries a rich WebPage + Organization + WebSite JSON-LD graph with sameAs profiles, logo, and search action. However, no LocalBusiness or Service schema exists despite the agency having four physical offices (Buffalo, Rochester, Albany, NYC). The industry vertical pages (Sports, Healthcare) and service pages (Creative, Digital Marketing, Media) contain no Service schema to describe what the agency offers. The blog uses Article schema with wordCount and author — well-structured. The Privacy Policy page returns only 2 words of visible text ("Privacy Policy") despite being indexed, suggesting JS-rendered content that crawlers may not fully capture.

External Signals

The site references 107 employees and 200+ active clients including New Era Cap, M&T Bank, Wegmans, Kaleida Health, and the Ralph C. Wilson, Jr. Foundation. The agency recently acquired Mr. Smith Agency (May 2026) and has completed multiple prior acquisitions (Travers Collins, AMDG Public Affairs, Gramercy Communications, Spiral Design Studio). The blog is active with posts dated through May 2026, covering industry insights, agency news, and a "She's Got Next" women-in-sports series. The sitemap contains 786+ URLs across 6 sub-sitemaps, indicating substantial content depth — though many blog posts date back to 2017-2018 and may be stale for AI training recency signals.

Findings

1. Founding year mismatch between site (2001) and LLM knowledge (1991) High

   The About page and acquisition blog post state the agency was founded in 2001, but cold LLM knowledge and external sources consistently cite 1991. This decade-long discrepancy creates a fragmented origin story across AI engines.

   What to change: Align the founding year across all site pages and correct the LLM knowledge by publishing a consistent, authoritative statement (e.g., via a Wikipedia page or press release).

2. Production pages contain Pantheon dev environment URLs in images High

   Image URLs on the About, Sports, and Healthcare pages reference the dev domain `dev-tmg-2024.pantheonsite.io`, leaking internal infrastructure and potentially causing broken images if the dev site is decommissioned.

   What to change: Replace all dev environment URLs in production image sources with the live domain.

3. No LocalBusiness schema despite four physical offices Medium

   The site lacks `LocalBusiness` schema to represent its offices in Buffalo, Rochester, Albany, and NYC, reducing visibility in AI-generated local business answers.

   What to change: Add `LocalBusiness` JSON-LD for each office location with address, phone, and opening hours.

4. Industry vertical and service pages lack Service schema Medium

   Pages for Sports, Healthcare, Creative, Digital Marketing, and Media contain no `Service` schema to describe the agency's offerings, limiting AI understanding of service scope.

   What to change: Add `Service` schema to each vertical and service page describing the specific offerings.

5. Robots.txt is bare with no AI crawler directives Medium

   The `robots.txt` file contains only a single `User-agent: * Disallow:` rule and does not name any AI crawlers (GPTBot, ClaudeBot, etc.), missing an opportunity to guide AI bot behavior.

   What to change: Add explicit rules for AI crawlers (e.g., allow all) and consider adding a `llms.txt` file.

6. No llms.txt file published Low

   The site returns a 404 for `llms.txt`, missing a standard way to provide AI crawlers with a curated summary of the site's content.

   What to change: Create an `llms.txt` file with a brief site description and links to key pages.

7. Privacy Policy page returns only 2 words of visible text Medium

   The Privacy Policy page serves only the title "Privacy Policy" in the initial HTML, suggesting the full content is rendered client-side, which may not be fully indexed by AI crawlers.

   What to change: Ensure the full privacy policy content is server-side rendered or included in the initial HTML.

8. LLM cold knowledge omits Hearst minority investment Medium

   The cold LLM knowledge does not mention the 2020 Hearst minority investment, a major structural fact prominently featured on the site, leading to incomplete AI-generated company profiles.

   What to change: Publish a press release or Wikipedia page detailing the Hearst investment to improve LLM knowledge.

9. LLM cold knowledge lists CEO as 'Todd Martin' but site says 'Tod Martin' Low

   The cold knowledge mentions CEO Todd Martin, while the site's team page lists Tod Martin (one 'd'), creating confusion in AI-generated leadership summaries.

   What to change: Ensure the correct spelling (Tod Martin) is consistently used across the site and in external profiles.

10. Many blog posts date back to 2017-2018 Low

    The sitemap contains over 786 URLs, but a significant portion of blog posts are from 2017-2018, which may be considered stale for AI training recency signals.

    What to change: Consider updating or removing very old blog posts to maintain content freshness.

What's working

• All major AI crawlers receive full access with identical content — GPTBot, ClaudeBot, PerplexityBot, and other AI crawlers receive a 200 response with the same byte-size content as a browser, ensuring no blocking or cloaking.
• Every page includes rich WebPage, Organization, and WebSite JSON-LD — The site consistently embeds structured data with sameAs profiles, logo, and search action, aiding AI understanding of the brand.
• Blog posts use Article schema with wordCount and author — The blog implements proper `Article` schema, providing detailed metadata that helps AI crawlers index content accurately.
• Active blog with recent posts and comprehensive sitemap — The blog is updated through May 2026 and the sitemap contains over 786 URLs across 6 sub-sitemaps, indicating substantial content depth.
• 60-second cache TTL ensures AI crawlers hit fast cached pages — The Pantheon hosting with Varnish cache provides a 60-second TTL, so AI crawlers receive quickly served cached copies rather than waiting for dynamic generation.