water.ie — AI Site Grade

Analysis

AI crawlers see everything — and nothing useful

Every major AI crawler (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, Bytespider, Applebot-Extended) receives a 200 with identical full HTML as a browser — no UA-based blocking, no JS shell, no Cloudflare challenge. Yet the site is structurally invisible to AI engines because it ships zero structured data and has no AI-friendly content map.

Crawler Access

The robots.txt at water.ie/robots.txt contains a single catch-all User-agent: * rule with no AI-bot-specific directives. No GPTBot, no ClaudeBot, no PerplexityBot — nothing. The llms.txt returns 404. The sitemap (Drupal-generated, ~5,485 URLs across 3 sub-sitemaps) is well-formed and accessible. The site runs on Oracle Cloud via Akamai CDN (IPs 23.1.96.165/193) with HSTS, CSP, and X-Frame-Options: DENY. All bots get the same server-side rendered Drupal HTML — no JavaScript rendering risk.

Schema Posture

This is the critical gap. Across every page examined — homepage, about, help, conservation, business, contact, water-service-updates, projects, boil-water-notice — zero JSON-LD schema of any type was found. No Organization, no WebSite, no FAQPage, no LocalBusiness, no Service, no ContactPoint. The homepage has no schema at all. The contact page lists phone numbers (1800 278 278, 0818 778 778) in plain text with no ContactPoint markup. The boil-water-notice page is a rich FAQ-style document with clear question/answer patterns but no FAQPage schema. The site has hreflang annotations (en/ga) in the sitemap and xhtml:link elements, but no og:locale or schema-level language markup.

Cold-Knowledge Gap

The LLM prior on Uisce Eireann (formerly Irish Water) is dominated by historical controversy: the 2013 establishment amid protests over domestic water charges, the ~40% network leakage rate, persistent boil-water notices in Donegal and Kerry, and slow infrastructure progress. The actual site presents a polished, forward-looking brand — "Water Stewardship for SMEs," "Water for Growth," "The Story of Water" documentary, a seven-year Business Plan, and values-driven corporate messaging. The site does not acknowledge or address the controversy at all. No page explains the transition from Irish Water to Uisce Eireann, the history of water charges, or current leakage performance metrics. An AI engine retrieving the site cold would find aspirational content that contradicts the negative prior without any bridging narrative.

External Signals

The site links to social channels (X/Twitter @IWCare and @IrishWater, Facebook, YouTube, LinkedIn, Instagram, TikTok) but these are not discoverable via schema. The DNS TXT records reveal a complex vendor stack: Oracle Cloud DNS, Microsoft 365 (Exchange Online), Oracle email delivery, Brevo, TopSec, Cisco, TeamViewer SSO, and DocuSign. No external review aggregators, press mentions, or Reddit threads surfaced in search — the brand's external reputation is shaped almost entirely by news media coverage of water quality incidents and political history, none of which the site references.

Content Architecture

The site has strong informational content — 779 words on boil-water-notice procedures, detailed conservation guidance, a water conservation calculator, and a free leak repair scheme ("First Fix Free") — but none of it is surfaced to AI crawlers via structured data. The homepage delivers only 136 words of visible text (mostly navigation labels). Key pages like /water-service-updates are sign-up forms rather than live incident feeds. The sitemap reveals ~5,485 URLs including many stale nodes (e.g., /node/3808 last modified 2020, /404 as a sitemap entry). The Irish-language mirror (/ga) duplicates the entire site structure, doubling the crawl surface without schema differentiation.

Findings

1. Zero JSON-LD schema across all pages High

   No page on water.ie includes any JSON-LD structured data. Missing Organization, WebSite, FAQPage, LocalBusiness, Service, and ContactPoint schema prevents AI crawlers from understanding the site's content and entity relationships.

   What to change: Add Organization, WebSite, and relevant page-type schema (e.g., FAQPage for boil-water-notice, ContactPoint for contact page) using JSON-LD on all pages.

2. llms.txt returns 404 Medium

   The site does not provide an llms.txt file, which is a recommended way to guide AI crawlers to key content and provide context about the organization.

   What to change: Create an llms.txt file at the root that lists important pages (e.g., about, help, boil-water-notice) and provides a brief summary of the organization.

3. Cold-knowledge gap: site ignores historical controversy High

   LLM prior knowledge of Uisce Éireann is dominated by historical controversy (water charges protests, high leakage rates, boil-water notices), but the site presents only aspirational content without addressing or bridging this narrative. AI engines retrieving the site cold will find a disconnect that undermines trust.

   What to change: Add a page or section that transparently addresses the transition from Irish Water, current leakage metrics, and steps taken to improve water quality, using schema markup to highlight key facts.

4. No AI-bot-specific directives in robots.txt Low

   The robots.txt file contains only a catch-all rule with no explicit directives for GPTBot, ClaudeBot, PerplexityBot, or other AI crawlers. While not blocking them, this misses the opportunity to guide their crawl behavior.

   What to change: Add specific rules for AI bots (e.g., GPTBot, ClaudeBot) to allow access and optionally set crawl-delay.

5. Contact phone numbers in plain text without ContactPoint schema Medium

   The contact page lists phone numbers (1800 278 278, 0818 778 778) in plain text with no ContactPoint structured data, making it harder for AI assistants to extract and present this information.

   What to change: Add ContactPoint schema markup with telephone numbers, contact type, and available hours.

6. FAQ-style content on boil-water-notice lacks FAQPage schema Medium

   The boil-water-notice page contains clear question-and-answer patterns but no FAQPage schema, preventing AI crawlers from surfacing this content as rich results.

   What to change: Add FAQPage schema with mainEntity for each question/answer pair on the boil-water-notice page.

7. Homepage delivers only 136 words of visible text Medium

   The homepage contains very little substantive text (mostly navigation labels), which limits the semantic signal for AI crawlers trying to understand the site's purpose.

   What to change: Add a brief introductory paragraph about Uisce Éireann's mission and services on the homepage, marked up with Organization and WebSite schema.

8. Sitemap contains stale URLs including /node/3808 from 2020 and /404 Low

   The sitemap includes outdated URLs (e.g., /node/3808 last modified 2020) and a /404 page, which wastes crawl budget and may confuse AI crawlers.

   What to change: Remove stale and error-page URLs from the sitemap, and ensure only canonical, current pages are included.

9. Social media links not marked up with schema Low

   The site links to social channels (X/Twitter, Facebook, YouTube, LinkedIn, Instagram, TikTok) but does not use sameAs or other schema to associate them with the organization.

   What to change: Add sameAs properties to the Organization schema pointing to official social media profiles.

10. Hreflang annotations present but no schema-level language markup Low

    The site uses hreflang in sitemap and xhtml:link for English and Irish, but lacks og:locale or schema-level language specification, which could help AI crawlers understand content language.

    What to change: Add inLanguage property to schema markup for each page, and include og:locale meta tags.

11. Water service updates page is a sign-up form, not live incident feed Medium

    The /water-service-updates page presents a sign-up form for alerts rather than a live feed of current incidents, limiting its usefulness for AI crawlers seeking real-time information.

    What to change: Add a publicly accessible page listing current service updates (boil notices, outages) with structured data (e.g., Event or SpecialAnnouncement schema).

12. No external review or press mentions surfaced in search Medium

    Web searches for reviews, Reddit discussions, and press coverage returned no results, indicating the brand's external reputation is shaped entirely by news media coverage of controversies, which the site does not address.

    What to change: Proactively publish press releases, case studies, and customer testimonials on the site, and encourage reviews on third-party platforms. Link to these from the site with schema markup.

What's working

• All major AI crawlers receive full HTML content — Every tested AI bot (GPTBot, ClaudeBot, PerplexityBot, etc.) receives a 200 response with the same server-side rendered HTML as a browser, with no blocking or JavaScript dependency.
• Well-formed sitemap with ~5,485 URLs — The Drupal-generated sitemap is accessible, well-formed, and includes a large number of URLs across sub-sitemaps, providing good crawl coverage.
• Hreflang annotations for English and Irish — The sitemap and xhtml:link elements include hreflang annotations for en and ga, helping search engines serve the correct language version.
• Rich informational content on key pages — Pages like boil-water-notice (779 words), conservation (487 words), and contact (353 words) contain detailed, useful information that could be surfaced by AI if properly marked up.
• Server-side rendered Drupal HTML — The site uses server-side rendering, so all content is present in the initial HTML response, avoiding JavaScript rendering issues for AI crawlers.
• Active social media presence across multiple platforms — The site links to official accounts on X/Twitter, Facebook, YouTube, LinkedIn, Instagram, and TikTok, providing channels for public engagement and brand visibility.
• No JavaScript shell or Cloudflare challenges for bots — AI crawlers are not blocked by JavaScript rendering requirements or Cloudflare challenges, ensuring direct access to content.
• Comprehensive vendor stack with security measures — The site uses Oracle Cloud via Akamai CDN, HSTS, CSP, and X-Frame-Options: DENY, indicating a robust security posture.