bpigroup.com — AI Site Grade

Analysis

The Cold-Knowledge Gap

The most consequential finding is that a frontier LLM, queried cold, conflates BPI Group with the Bank of the Philippine Islands — a Philippine financial conglomerate with no connection to this site. The model described BPI as "a Philippine-based conglomerate primarily known as the holding company for the Bank of the Philippine Islands," mentioned a 2023 data breach, and listed retail banking products. The actual site is Bully Pulpit International, a global outcomes agency at the intersection of business, politics, and policy, with offices across the US and Europe, clients including Walmart, McDonald's, and Strava, and a leadership roster featuring former White House and Downing Street communications directors. The domain name bpigroup.com creates an identity collision that AI models resolve to the wrong entity entirely.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, anthropic-ai, Applebot-Extended — receive a 200 status with the same full HTML payload as a browser (104,704 bytes). Only Bytespider (ByteDance's crawler) is blocked with a 403 from Cloudflare. The robots.txt is minimal: a single User-agent: * rule disallowing only /wp-admin/, with no AI-bot-specific directives. No llms.txt exists (404). The site runs on WP Engine behind Cloudflare with nginx. Content is server-rendered HTML (not a JS shell), so AI crawlers see the full text.

Schema Posture

Every page carries a large JSON-LD block, but it is structurally flawed. The schema uses a @graph that declares a combined Person/Organization type at /#person as the publisher of every page, then wraps each page as an Article — even the homepage, the "Who We Are" page, and the "Where We Are" page. These are not articles. The Organization schema lacks sameAs links for the company's social profiles (Instagram, Facebook, X, LinkedIn), which are present in the page footer. No LocalBusiness or ProfessionalService schema exists for the 15+ physical office locations listed on the /where-we-are/ page. No FAQPage, Product, or ItemList schemas are used despite the site having FAQ-like service breakdowns and client lists.

Content and Structure

The site presents a clear, well-written value proposition: "an outcomes agency at the intersection of business, politics, and policy." The homepage has a strong H1 ("Lead change, don't be consumed by it."), a services grid, client logos (Walmart, McDonald's, Strava, Ad Council, Indra Nooyi), and a blog/news feed. The /what-we-do/ page lists detailed campaign services (strategic planning, niche audience development, message testing, etc.) and prominently displays awards: 2x Global Agency of the Year (PRovoke Media), Public Affairs Agency of the Year (PRWeek), Clio Awards, Telly Awards, Fast Company World Changing Ideas. The /leadership/ page lists 80+ team members across North America and Europe, including former White House Communications Director Ben LaBolt (President), former Downing Street Director of Communications Guto Harri (UK Senior Advisor), and former Gibbs advisor Robert Gibbs (Senior Advisor). The blog (/from-the-bully-pulpit/) is active with 40+ posts covering AI, EU regulation, political campaigning, and M&A announcements.

External Signals

DuckDuckGo searches for "Bully Pulpit International" returned zero results during testing, which is anomalous and may indicate a temporary search outage or the site's relatively low organic search footprint. The site references major industry awards from PRovoke Media, PRWeek, and Fast Company, but no external press coverage or third-party reviews were reachable through search during the audit window. The DNS records show Google-hosted email (smtp.google.com MX), OneTrust privacy, and multiple Google Search Console verification TXT records — indicating active SEO management intent.

Surprising Findings

The /heritage/ page, listed in the sitemap, returns a 404. The homepage's datePublished in schema is May 5, 2026 — a future date — suggesting a CMS misconfiguration or placeholder timestamp. The Organization schema's @id is set to /#person, conflating the organization entity with a person entity. The site has a full German-language mirror at /de/ but no hreflang annotations in the HTML or sitemap. Despite being a data-driven agency that sells AI-powered campaign tools (ChangeOS), the site itself has no llms.txt to guide AI crawlers to its most valuable content.

Findings

1. Domain name causes AI models to conflate BPI Group with a Philippine bank High

   A frontier LLM queried cold identifies BPI Group as a Philippine conglomerate holding company for the Bank of the Philippine Islands, with no connection to the actual agency Bully Pulpit International. The domain name bpigroup.com creates an identity collision that AI models resolve to the wrong entity entirely.

   What to change: Add structured data and content that explicitly disambiguates the brand from the Philippine bank, and consider a domain name that matches the brand name 'Bully Pulpit International' to reduce confusion.

2. Every page is marked as an Article in JSON-LD, including non-article pages High

   The site's JSON-LD schema wraps every page as an Article, even the homepage, 'Who We Are', and 'Where We Are' pages. These are not articles, which misleads AI crawlers about the content type.

   What to change: Use appropriate schema types (e.g., WebPage, AboutPage, ContactPage) for non-article pages, and reserve Article for blog posts.

3. Organization schema lacks sameAs links for social profiles Medium

   The Organization schema does not include sameAs links for the company's social media profiles (Instagram, Facebook, X, LinkedIn), which are present in the page footer. This reduces the entity's digital footprint for AI crawlers.

   What to change: Add sameAs properties to the Organization schema with URLs to all official social media profiles.

4. No LocalBusiness or ProfessionalService schema for 15+ office locations Medium

   The /where-we-are/ page lists 15+ physical office locations across the US and Europe, but no LocalBusiness or ProfessionalService schema is present. This limits local AI visibility and map integration.

   What to change: Add LocalBusiness or ProfessionalService schema for each office location with address, phone, and opening hours.

5. Homepage schema has a future datePublished of May 5, 2026 Medium

   The homepage's JSON-LD includes a datePublished value of May 5, 2026, which is in the future. This is likely a CMS misconfiguration and can confuse AI crawlers about content freshness.

   What to change: Correct the datePublished to the actual publication date or remove it for non-article pages.

6. Organization @id is set to /#person, conflating organization with a person entity Medium

   The Organization schema's @id is set to /#person, which incorrectly merges the organization entity with a person entity. This can cause AI models to misinterpret the entity type.

   What to change: Use a distinct @id for the Organization, such as /#organization, and ensure Person and Organization are separate entities.

7. No llms.txt file to guide AI crawlers to key content Medium

   The site returns a 404 for /llms.txt. Despite being a data-driven agency with AI-powered tools, the site does not provide a structured guide for AI crawlers to discover its most valuable content.

   What to change: Create an llms.txt file that lists key pages (e.g., /what-we-do/, /leadership/, /case-study/walmart/) and provides a brief summary of the site.

8. Heritage page listed in sitemap returns 404 Medium

   The /heritage/ page is included in the sitemap but returns a 404 error. This creates a broken user and crawler experience.

   What to change: Either restore the heritage page content or remove the URL from the sitemap and set up a proper redirect.

9. German-language mirror at /de/ lacks hreflang annotations Low

   The site has a full German-language version at /de/ but no hreflang tags in the HTML or sitemap. This prevents AI crawlers from properly associating the language variants.

   What to change: Add hreflang annotations to all pages linking to their German counterparts, and include them in the sitemap.

10. DuckDuckGo returns zero results for brand queries High

    Multiple DuckDuckGo searches for 'Bully Pulpit International' and related terms returned zero results during testing. This is anomalous and may indicate a temporary search outage or very low organic search footprint.

    What to change: Investigate search engine indexing status, ensure the site is not blocked by robots.txt or noindex tags, and improve off-site SEO signals.

11. Bytespider crawler is blocked with 403 Low

    The ByteDance crawler (Bytespider) receives a 403 Forbidden response from Cloudflare, while all other major AI crawlers are allowed. This may limit visibility on Chinese AI platforms.

    What to change: If the site targets Chinese audiences, allow Bytespider access; otherwise, this is acceptable.

What's working

• All major AI crawlers receive full HTML content — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and other major AI crawlers receive a 200 status with the same full HTML payload as a browser. Content is server-rendered, not a JS shell, so AI crawlers see the full text.
• Robots.txt only blocks /wp-admin/ with no AI-bot restrictions — The robots.txt file is minimal, disallowing only /wp-admin/ for all user-agents, with no AI-bot-specific directives. This ensures AI crawlers can access all public content.
• Clear value proposition and well-written content across key pages — The site presents a strong value proposition ('an outcomes agency at the intersection of business, politics, and policy') with detailed service descriptions, client logos, and awards. The blog is active with 40+ posts covering relevant topics.
• Leadership page lists 80+ team members with notable credentials — The /leadership/ page lists over 80 team members across North America and Europe, including former White House Communications Director Ben LaBolt and former Downing Street Director of Communications Guto Harri, providing strong entity signals.
• DNS records indicate active SEO management intent — DNS records show Google Search Console verification TXT records and Google-hosted email, indicating the site is actively managed for SEO.
• Case studies feature recognizable brands like Walmart — The site includes case studies with major clients such as Walmart, McDonald's, and Strava, providing credible external signals and rich content for AI crawlers.
• Full German-language version of the site exists at /de/ — The site has a complete German-language mirror at /de/, which can help reach German-speaking audiences and improve international AI visibility once hreflang is implemented.