afford.com.au — AI Site Grade

Analysis

I have enough data now. Let me compile the audit.

GPTBot and Bytespider are blocked at the Cloudflare edge while every other major AI crawler passes through — a selective blockade that creates an uneven AI-knowledge profile.

Crawler Access

The site runs on WordPress hosted on WP Engine behind Cloudflare. compare_bot_access on the homepage shows GPTBot returns 429 (rate-limited/blocked) and Bytespider returns 403 (blocked), while ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, and Applebot-Extended all return 200 with the full 1.2 MB payload — identical to a browser visit. The robots.txt contains no AI-bot directives at all (just a Yoast-generated User-agent: * Disallow: with a Crawl-delay: 10). No llms.txt exists (404). The sitemap at /sitemap_index.xml is valid and contains 8 sub-sitemaps covering ~125+ pages including property listings, blog posts, and service pages.

Cold-Knowledge Gap

The LLM's prior knowledge about Afford is partially outdated and partially inaccurate. It states Afford was "founded in 1992 by parents of children with disabilities" — the site's own history page says the organisation began in 1951 as the Poliomyelitis Society, founded by Dr Ross Williams (a polio survivor and medical practitioner). The LLM also references "Afford Lifestyle" and "Afford Works" as named product lines; the site uses "Explore community activity program" and "Inclusive Employment Australia" as current branding. The LLM mentions "scrutiny from the NDIS Quality and Safeguards Commission regarding restrictive practices reporting" — no evidence of this appears anywhere on the site or in search results. The site claims 75+ years of operation, 2,100+ clients supported in 2025, and an 80% workplace confidence rate — none of these figures appear in the LLM's cold knowledge.

Schema Posture

Every page fetched uses WebPage + BreadcrumbList + WebSite schema with SearchAction — a solid baseline. However, no LocalBusiness, Organization, FAQPage, Product, or Service schema is present anywhere. The FAQ page (/resources/faqs/) has 15+ question-answer pairs rendered as visible HTML but marked up only as generic WebPage — no FAQPage schema. The property pages (e.g., Ingleburn SIL) describe a specific vacancy with beds, bathrooms, accessibility features, and location data but use no Accommodation, LocalBusiness, or Place schema. The blog posts lack Article or BlogPosting schema (just WebPage).

Content Signals

The homepage has a FAQ section, tables, and lists — strong answer-format signals for AI snippet extraction. The content hub publishes regularly (articles dated as recently as May 2026). The site uses Easy Read versions of key pages and offers a National Relay Service integration — accessibility signals that AI engines value. The property pages include structured data like bed/bath counts, vacancy status, and proximity to amenities, but this is only in visible HTML, not in schema markup.

External Signals

Search results returned zero external mentions — no Reddit threads, no news articles, no review sites surfaced for "Afford disability services" in the current index. The LLM's cold knowledge references "inconsistent staff training and communication issues" in reviews, but no such reviews were findable via search. This creates a reputation vacuum: AI engines have no third-party signals to triangulate against, making the site's own content the sole source of truth — which amplifies the importance of schema accuracy and completeness.

Findings

1. GPTBot and Bytespider blocked at Cloudflare edge High

   GPTBot returns 429 (rate-limited) and Bytespider returns 403 (blocked) on the homepage, while ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, and Applebot-Extended all receive 200 with full content. This selective blockade creates an uneven AI-knowledge profile.

   What to change: Remove the Cloudflare WAF rules that block GPTBot and Bytespider, or ensure consistent access for all major AI crawlers.

2. No llms.txt file Medium

   The site returns a 404 for /llms.txt, missing an opportunity to guide AI crawlers to key content.

   What to change: Create an llms.txt file listing important pages like services, FAQs, and property listings.

3. LLM cold knowledge is outdated and inaccurate High

   The LLM's prior knowledge about Afford is partially incorrect: it states the organization was founded in 1992 by parents, but the site says it began in 1951 as the Poliomyelitis Society. It also references outdated brand names and unsubstantiated regulatory scrutiny.

   What to change: Publish accurate, structured data about the organization's history and current services to correct LLM knowledge.

4. FAQ page lacks FAQPage schema High

   The FAQ page at /resources/faqs/ contains 15+ question-answer pairs in visible HTML but uses only generic WebPage schema, missing the opportunity for rich snippet extraction.

   What to change: Add FAQPage schema markup to the FAQ page, wrapping each Q&A pair in the appropriate structured data.

5. No LocalBusiness or Organization schema on key pages High

   Despite having a physical presence and multiple service locations, the site does not use LocalBusiness or Organization schema on any fetched page.

   What to change: Add Organization and LocalBusiness schema to the homepage and contact pages, including name, address, phone, and service area.

6. Property pages lack Accommodation schema High

   Property pages like the Ingleburn SIL listing include bed/bath counts, vacancy status, and accessibility features in visible HTML but use no Accommodation, Place, or LocalBusiness schema.

   What to change: Add Accommodation or Apartment schema to property pages, including occupancy, amenities, and location data.

7. Blog posts lack Article schema Medium

   Blog posts use only WebPage schema instead of Article or BlogPosting, reducing their visibility in AI-driven news and content features.

   What to change: Add Article schema to all blog posts, including headline, datePublished, author, and image.

8. No external signals found in search results Medium

   Multiple web searches for Afford's reviews, news, and social mentions returned zero results, creating a reputation vacuum that makes the site's own content the sole source of truth.

   What to change: Encourage client reviews on Google and other platforms, and publish press releases or case studies to generate third-party mentions.

9. Robots.txt has no AI-bot directives Low

   The robots.txt file contains only a generic User-agent: * with a crawl-delay, and does not explicitly allow or disallow any AI crawlers.

   What to change: Add explicit directives for AI crawlers (e.g., GPTBot, ClaudeBot) to ensure intended access.

What's working

• Valid sitemap with 8 sub-sitemaps covering 125+ pages — The sitemap at /sitemap_index.xml is valid and includes sub-sitemaps for property listings, blog posts, and service pages, aiding crawler discovery.
• WebPage, BreadcrumbList, and WebSite schema present on all pages — Every fetched page includes WebPage, BreadcrumbList, and WebSite schema with SearchAction, providing a solid structured data foundation.
• Easy Read versions and National Relay Service integration — The site offers Easy Read versions of key pages and integrates the National Relay Service, providing accessibility signals that AI engines value.
• Homepage FAQ section and tables provide answer-format signals — The homepage includes a FAQ section, tables, and lists that are strong signals for AI snippet extraction, even without schema markup.
• Content hub publishes regularly with recent articles — The content hub has articles dated as recently as May 2026, indicating active content creation that can improve AI visibility.
• Most major AI crawlers receive full content access — ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User, and Applebot-Extended all return 200 with full page content, ensuring broad AI visibility.