headspace.org.au — AI Site Grade

Analysis

headspace.org.au — AI-Visibility Audit

The site's most consequential AI-visibility problem is not blocked bots or missing schema — it is that every content page shares a single canonical URL polluted with Google Ads click-tracking parameters, creating a massive canonical confusion risk for any AI crawler indexing the site's 30+ sitemaps worth of content.

Crawler Access

All major AI crawlers — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, Bytespider, Applebot-Extended, anthropic-ai, ChatGPT-User — receive a full 200 response with ~146KB of content, identical to a browser baseline. No UA-based blocking exists. The site runs on Cloudflare (NS records mitchell.ns.cloudflare.com / zainab.ns.cloudflare.com) with strong security headers (HSTS, CSP, X-Frame-Options DENY). The robots.txt uses a single User-agent: * rule with no AI-bot-specific directives — no bot is explicitly welcomed or restricted beyond the generic disallows. /llms.txt returns a 404 (serving the full HTML error page, ~101KB). The sitemap index contains 30 sub-sitemaps (SilverStripe CMS), signalling a large content surface that AI crawlers can reach but may struggle to navigate efficiently without an llms.txt guide.

Content & Schema Posture

The homepage carries a rich NGO schema block with tax ID, funder (Australian Government), 14 sameAs links, and potentialAction for Find/Search/Donate. However, deeper content pages are schema-thin: the explore-topics, services, and centre pages only carry BreadcrumbList schema. No FAQPage, Article, MedicalWebPage, HowTo, or QAPage markup exists on pages that contain answer-format content (lists, definitions, embedded videos). The online-and-phone-support page is the exception, carrying a WebPage schema with Offer and VideoObject sub-entities. The homepage has 1,489 words of visible text — no JS-rendering dependency. Content pages are text-rich (977–4,523 words) and server-rendered, making them fully parseable by AI crawlers.

Cold-Knowledge Gap

The LLM's prior knowledge of headspace is surprisingly accurate and current: it knows the 12–25 age range, the 150+ centres, the eheadspace online service, the 2006 founding by the Australian government, and even mentions 2023–2024 scrutiny over wait times and underfunding. The site itself does not address wait-time criticism or funding challenges anywhere on the homepage, services page, or "who we are" page — the site presents an unqualified positive narrative. This creates a trust gap: AI models that have ingested external news coverage will surface the wait-time and underfunding narrative alongside the site's own messaging, and the site offers no counter-narrative or acknowledgment.

Canonical Contamination

The most technically concerning finding: the canonical URL on the explore-topics/for-young-people/alcohol-and-other-drugs/ page points to https://headspace.org.au/explore-topics/for-young-people/alcohol-and-other-drugs/?gad_source=1&gad_campaignid=23062555560&gbraid=...&gclid=... — a URL littered with Google Ads tracking parameters. Worse, the page at /explore-topics/for-young-people/mental-ill-health/anxiety-0/ (the anxiety page) has the same canonical pointing to the alcohol-and-other-drugs page with the same ad-tracking URL. This means two different content topics (anxiety vs. alcohol) share a single canonical URL, and that canonical is a paid-ad tracking URL. AI crawlers indexing these pages will encounter conflicting signals about which URL represents the authoritative version of which content.

External Signals

The media releases page shows active, current news output (May 2026 items about the headspace Plus Model of Care, LGBTIQA+ data, Rainbow Tick accreditation, a Danish royal visit). The site references 14 external sameAs links including Wikipedia, health.gov.au, and state health departments. The DNS TXT records reveal a complex marketing stack: Microsoft Dynamics 365, Zendesk, Qualtrics, Campaign Monitor, and Amazon SES — indicating significant off-domain data collection that AI crawlers cannot access, creating a gap between the site's public content and the richer user-data picture held in these platforms.

Findings

1. Canonical URLs polluted with Google Ads tracking parameters High

   The canonical URL on the alcohol-and-other-drugs page points to a URL with Google Ads tracking parameters (gad_source, gad_campaignid, gbraid, gclid). The anxiety page shares the same canonical, pointing to the alcohol page's ad-tracking URL, causing two distinct topics to share a single canonical.

   What to change: Remove all Google Ads tracking parameters from canonical URLs. Ensure each content page has a unique, clean canonical URL that matches its own path.

2. Missing /llms.txt file Medium

   The /llms.txt endpoint returns a 404 with a full HTML error page (~101KB). AI crawlers have no machine-readable guide to the site's content surface, making navigation of the 30+ sitemaps inefficient.

   What to change: Create an /llms.txt file listing key content pages and their purposes, following the llms.txt standard.

3. Deeper content pages lack rich schema markup Medium

   Explore-topics, services, and centre pages only carry BreadcrumbList schema. No FAQPage, Article, MedicalWebPage, HowTo, or QAPage markup exists on pages with answer-format content, limiting AI understanding and eligibility for rich results.

   What to change: Add relevant schema types (FAQPage, Article, MedicalWebPage, HowTo) to content pages that contain structured information, lists, or definitions.

4. Site does not address wait-time criticism or funding challenges Medium

   The homepage, services page, and 'who we are' page present an unqualified positive narrative. LLM prior knowledge includes scrutiny over wait times and underfunding (2023–2024), creating a trust gap when AI models surface external criticism alongside the site's messaging.

   What to change: Add a page or section that acknowledges challenges (e.g., wait times, funding) and describes steps being taken, to provide a balanced narrative for AI consumption.

5. robots.txt lacks AI-bot-specific directives Low

   The robots.txt uses a single User-agent: * rule with no explicit welcome or restriction for AI crawlers. While no blocking occurs, the absence of directives leaves AI crawlers without guidance on crawl priorities or disallowed paths.

   What to change: Add explicit directives for AI crawlers (e.g., GPTBot, ClaudeBot) to welcome them and optionally disallow low-value paths.

6. Rich user data held in external platforms inaccessible to AI crawlers Low

   DNS TXT records reveal a complex marketing stack (Microsoft Dynamics 365, Zendesk, Qualtrics, Campaign Monitor, Amazon SES). AI crawlers cannot access this off-domain data, creating a gap between public content and the richer user-data picture.

What's working

• All major AI crawlers receive full 200 responses — GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and others receive the same content as a browser baseline, with no UA-based blocking.
• LLM prior knowledge of headspace is accurate and current — The LLM correctly knows the 12–25 age range, 150+ centres, eheadspace service, 2006 founding, and recent scrutiny over wait times and underfunding.
• Homepage carries rich NGO schema with sameAs links — The homepage includes NGO schema with tax ID, funder, 14 sameAs links, and potentialAction for Find/Search/Donate, providing strong structured data for AI understanding.
• Content pages are server-rendered and text-rich — Pages like explore-topics, services, and centres are server-rendered with 977–4,523 words of visible text, fully parseable by AI crawlers without JavaScript dependency.
• Media releases page shows current, active news output — The media releases page contains May 2026 items about the headspace Plus Model of Care, LGBTIQA+ data, Rainbow Tick accreditation, and a Danish royal visit, demonstrating ongoing public engagement.
• Sitemap index with 30 sub-sitemaps signals large content surface — The sitemap index contains 30 sub-sitemaps (SilverStripe CMS), indicating a large content surface that AI crawlers can discover.
• Online and phone support page carries WebPage schema with Offer and VideoObject — The online-and-phone-support page includes WebPage schema with Offer and VideoObject sub-entities, providing richer structured data than other content pages.